{
 "meta": {
  "schema_note": "One record per company. Every claim in tools/metrics/guardrails/hiring_reviews_changes/results is an object {text, src} where src lists indexes into that record's sources[]. Empty arrays mean nothing could be verified from a public source. 'industry' and 'size' are editorial bands; size_src is filled only where a cited source states headcount.",
  "size_bands": [
   "Under 5k employees",
   "5k-50k employees",
   "50k+ employees"
  ],
  "last_build_note": "All source URLs checked 2026-10-25."
 },
 "companies": [
  {
   "name": "Shopify",
   "slug": "shopify",
   "industry": "E-commerce",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "yes",
   "mandate_summary": {
    "text": "CEO Tobi Lütke's April 2025 memo: \"Reflexive AI usage is now a baseline expectation at Shopify.\" It applies to everyone, including the executive team. In 2026, AI reflexiveness is assessed in Shopify's twice-yearly performance reviews.",
    "src": [
     0,
     1,
     4
    ]
   },
   "tools": [
    {
     "name": "Cursor",
     "src": [
      0,
      3,
      4
     ]
    },
    {
     "name": "GitHub Copilot",
     "src": [
      0
     ]
    },
    {
     "name": "Claude Code",
     "src": [
      0,
      4
     ]
    },
    {
     "name": "chat.shopify.io (internal)",
     "src": [
      0
     ]
    },
    {
     "src": [
      4
     ],
     "name": "OpenAI Codex"
    },
    {
     "name": "River (internal Slack agent)",
     "src": [
      6,
      7
     ]
    },
    {
     "name": "Aquifer (internal agent platform)",
     "src": [
      6
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "% code AI-generated",
    "PR throughput",
    "Quality / defects",
    "Agent autonomy",
    "Cost per PR"
   ],
   "metrics": [
    {
     "text": "Shopify said in October 2025 that it had hit universal adoption of AI code editors and was running thousands of Cursor licenses.",
     "src": [
      3
     ]
    },
    {
     "text": "President Harley Finkelstein said on the Q1 2026 call that AI writes 'well over 50%' of Shopify's code and that the share is rising.",
     "src": [
      5
     ]
    },
    {
     "text": "Shopify tracks code reversion rate, which it says has stayed roughly flat as AI use grew, and alerts on anyone spending more than $250 a day in tokens through its LLM proxy.",
     "src": [
      4
     ]
    },
    {
     "text": "In a 30-day window, River ran 59,918 sessions in 5,170 Slack channels and co-authored 3,536 merged pull requests (about one in eight merged PRs).",
     "src": [
      6
     ]
    }
   ],
   "guardrails": [
    {
     "text": "All AI requests route through an internal LLM proxy. A senior human reviewer is required on PRs, and AI cannot check in code automatically.",
     "src": [
      4
     ]
    },
    {
     "text": "Engineers are expected to understand two to three layers below the code they ship, to avoid 'comprehension debt'.",
     "src": [
      4
     ]
    },
    {
     "text": "River only works in public Slack channels (no DMs), so every session is a searchable transcript. Agent harnesses run outside the code sandbox.",
     "src": [
      6
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Performance and peer reviews include AI usage. Per BVP's April 2026 profile, employees are evaluated on how 'AI-reflexive' they are in twice-yearly reviews.",
     "src": [
      0,
      2,
      4
     ]
    },
    {
     "text": "Before asking for more headcount or resources, teams must show why they cannot get what they want done using AI.",
     "src": [
      0,
      1
     ]
    },
    {
     "text": "AI exploration is expected in the prototype phase of every GSD project.",
     "src": [
      0
     ]
    }
   ],
   "results": [
    {
     "text": "Shopify reports that its internal 'Quick' prototype-hosting sandbox has passed 5,000 sites since launching in mid-2025.",
     "src": [
      3
     ]
    },
    {
     "text": "Every team has unlimited access to top AI models.",
     "src": [
      3
     ]
    },
    {
     "text": "Engineering leadership estimates about a 20% productivity gain from AI tools.",
     "src": [
      4
     ]
    },
    {
     "text": "Shopify gave agents per-session remote dev environments, so agents run in parallel and overnight workflows fix bugs and open PRs.",
     "src": [
      7
     ]
    }
   ],
   "sources": [
    {
     "title": "Tobi Lütke memo, 'Reflexive AI usage is now a baseline expectation at Shopify' (shared copy)",
     "url": "https://readwise.io/reader/shared/01jr9218k2nf651ygvwygxrk0w/",
     "date": "2025-04-07",
     "type": "primary"
    },
    {
     "title": "Tom's Hardware: Shopify CEO says teams must prove AI can't do a task before asking for new hires",
     "url": "https://www.tomshardware.com/tech-industry/artificial-intelligence/shopify-ceo-teams-must-prove-ai-cant-do-a-task-before-asking-for-new-hires-resources",
     "date": "2025-04-07",
     "type": "press"
    },
    {
     "title": "Digital Commerce 360: Internal memo, Shopify CEO declares AI non-optional",
     "url": "https://www.digitalcommerce360.com/2025/04/08/internal-memo-shopify-ceo-declares-ai-non-optional/",
     "date": "2025-04-08",
     "type": "press"
    },
    {
     "title": "Shopify: Serious results, unserious methods: Shopify's AI playground",
     "url": "https://www.shopify.com/news/unserious-exploration",
     "date": "2025-10-28",
     "type": "company"
    },
    {
     "title": "BVP Atlas: Inside Shopify's AI-first engineering playbook",
     "url": "https://www.bvp.com/atlas/inside-shopifys-ai-first-engineering-playbook",
     "date": "2026-04-02",
     "type": "press"
    },
    {
     "title": "Motley Fool: Shopify (SHOP) Q1 2026 earnings call transcript",
     "url": "https://www.fool.com/earnings/call-transcripts/2026/05/05/shopify-shop-q1-2026-earnings-transcript/",
     "date": "2026-05-05",
     "type": "press"
    },
    {
     "title": "Shopify Engineering: Under the River",
     "url": "https://shopify.engineering/under-the-river",
     "date": "2026-05-28",
     "type": "company"
    },
    {
     "title": "Shopify Engineering: We gave AI agents a mobile dev environment",
     "url": "https://shopify.engineering/remote-dev",
     "date": "2026-10-22",
     "type": "company"
    }
   ],
   "notes": "Superseded: the October 2025 'universal adoption of AI code editors' figure is now joined by a 2026 share-of-code figure (well over 50%) and River PR data.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Coinbase",
   "slug": "coinbase",
   "industry": "Crypto / fintech",
   "size": "Under 5k employees",
   "size_src": [],
   "mandate": "yes",
   "mandate_summary": {
    "text": "CEO Brian Armstrong required engineers to onboard to AI coding assistants within about a week, and said he fired some who did not have a good reason for skipping it. In May 2026 Armstrong restructured the company around AI-native pods that manage fleets of agents.",
    "src": [
     0,
     1,
     3
    ]
   },
   "tools": [
    {
     "name": "GitHub Copilot",
     "src": [
      0,
      1
     ]
    },
    {
     "name": "Cursor",
     "src": [
      0,
      1
     ]
    }
   ],
   "metric_tags": [
    "% code AI-generated",
    "Adoption rate",
    "Agent autonomy"
   ],
   "metrics": [
    {
     "text": "Superseded (August 2025): Armstrong said about a third of Coinbase's code was AI-written, with a goal of 50% by the end of the quarter.",
     "src": [
      1,
      2
     ]
    },
    {
     "text": "In July 2026 Coinbase's Rob Witoff said 95-100% of code is written by or with LLMs, up from about 40% in February, and that effectively every employee uses AI daily. Most engineers run 5-10 agents.",
     "src": [
      4
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Armstrong: 'You don't want people vibe coding these systems moving money'; AI code must be code reviewed with appropriate checks and humans in the loop.",
     "src": [
      2
     ]
    },
    {
     "text": "Coinbase's Rob Witoff said core cryptography is still meticulously researched and verified by human cryptographers, core systems sit in the middle, and internal prototypes are effectively 100% automated.",
     "src": [
      4
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Coinbase holds monthly meetings where teams that use AI creatively share what they did.",
     "src": [
      0
     ]
    },
    {
     "text": "The May 2026 restructuring cut about 14% of staff, capped the org at five layers below the CEO/COO, removed pure-manager roles and introduced AI-native pods, including one-person teams.",
     "src": [
      3
     ]
    },
    {
     "text": "Cointelegraph reported that junior developer roles were among those affected by the layoffs.",
     "src": [
      4
     ]
    }
   ],
   "results": [],
   "sources": [
    {
     "title": "TechCrunch: Coinbase CEO explains why he fired engineers who didn't try AI immediately",
     "url": "https://techcrunch.com/2025/08/22/coinbase-ceo-explains-why-he-fired-engineers-who-didnt-try-ai-immediately/",
     "date": "2025-08-22",
     "type": "press"
    },
    {
     "title": "Fortune: Coinbase CEO Brian Armstrong on the AI coding assistants mandate",
     "url": "https://fortune.com/2025/08/25/coinbase-ceo-brian-armstrong-ai-coding-assistants-mandate-tech/",
     "date": "2025-08-25",
     "type": "press"
    },
    {
     "title": "CIO: 'Fire any developer who doesn't use AI', on the Coinbase CEO's tough message",
     "url": "https://www.cio.com/article/4050185/fire-any-developer-who-doesnt-use-ai-why-coinbase-ceos-tough-message.html",
     "date": "2025-09-03",
     "type": "press"
    },
    {
     "title": "Coinbase blog: Building a leaner and faster Coinbase (Brian Armstrong)",
     "url": "https://www.coinbase.com/blog/building-a-leaner-and-faster-coinbase",
     "date": "2026-05-05",
     "type": "primary"
    },
    {
     "title": "Cointelegraph: Over 95% of Coinbase's code is now written with help of AI",
     "url": "https://cointelegraph.com/news/over-95-of-coinbases-code-is-now-written-with-ai",
     "date": "2026-07-14",
     "type": "press"
    }
   ],
   "notes": "Fintech or crypto company. It is not a bank. The Coinbase blog returns HTTP 403 to automated checks (bot protection) but loads for readers.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Duolingo",
   "slug": "duolingo",
   "industry": "Edtech / consumer apps",
   "size": "Under 5k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "Von Ahn's April 2025 'AI-first' memo made AI use part of hiring and performance reviews. In April 2026 he said Duolingo had backtracked on the review metric: 'if it can't [help], I'm not going to force you to do that.' The policy changed, so the mandate is now unclear.",
    "src": [
     0,
     1,
     2,
     3
    ]
   },
   "tools": [],
   "metric_tags": [],
   "metrics": [],
   "guardrails": [
    {
     "text": "The memo says Duolingo will provide training, mentorship and tooling, and that getting AI to understand its codebase will take time.",
     "src": [
      0,
      1
     ]
    },
    {
     "text": "Von Ahn said in April 2026 that AI is 'not yet' better at coding than humans and that AI-written code can be hard to debug. A spokesperson said AI tools assist work but 'don't make decisions'.",
     "src": [
      3
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "AI use will be part of what Duolingo looks for in hiring.",
     "src": [
      0,
      1
     ]
    },
    {
     "text": "Superseded: AI use was to be part of performance reviews (April 2025). Duolingo dropped this metric by April 2026.",
     "src": [
      0,
      1,
      3
     ]
    },
    {
     "text": "Headcount will only be given if a team cannot automate more of its work.",
     "src": [
      0,
      1
     ]
    },
    {
     "text": "Duolingo will gradually stop using contractors for work AI can handle.",
     "src": [
      0
     ]
    }
   ],
   "results": [],
   "sources": [
    {
     "title": "The Verge: Duolingo will replace contract workers with AI",
     "url": "https://www.theverge.com/news/657594/duolingo-ai-first-replace-contract-workers",
     "date": "2025-04-28",
     "type": "press"
    },
    {
     "title": "Business Insider: Duolingo CEO on how AI will be used in performance reviews and headcount",
     "url": "https://www.businessinsider.com/duolingo-ceo-how-ai-will-be-used-performance-reviews-headcount-2025-4",
     "date": "2025-04-29",
     "type": "press"
    },
    {
     "title": "Duolingo on LinkedIn: note from CEO Luis von Ahn",
     "url": "https://www.linkedin.com/posts/duolingo_below-is-a-note-from-our-ceo-luis-von-activity-7322560534824865792-DLdu/",
     "date": "2025-04-28",
     "type": "primary"
    },
    {
     "title": "Fortune: Duolingo CEO backs off from evaluating employees on their AI usage",
     "url": "https://fortune.com/2026/04/13/duolingo-ceo-luis-von-ahn-ai-usage-requirement-employee-performance-evaluations/",
     "date": "2026-04-13",
     "type": "press"
    }
   ],
   "notes": "Mandate reclassified yes to unclear on 2026-10-25 after the April 2026 reversal of the performance-review requirement.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Fiverr",
   "slug": "fiverr",
   "industry": "Marketplace",
   "size": "Under 5k employees",
   "size_src": [
    1
   ],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "CEO Micha Kaufman wrote to all staff that AI is coming for everyone's jobs and urged them to master AI tools. The memo is a strong expectation, but no formal policy has been reported.",
    "src": [
     0,
     1
    ]
   },
   "tools": [
    {
     "name": "Cursor",
     "src": [
      0,
      1
     ]
    }
   ],
   "metric_tags": [],
   "metrics": [],
   "guardrails": [],
   "hiring_reviews_changes": [
    {
     "text": "Kaufman wrote that programmers should master tools such as Cursor, and that it doesn't make sense to hire more people before the company learns how to do more with what it has.",
     "src": [
      0
     ]
    }
   ],
   "results": [
    {
     "text": "GAAP research and development expense fell to $18.6M in Q2 2026 from $24.0M in Q2 2025. The company describes itself as running a lean organization; the filing does not attribute the change to AI coding.",
     "src": [
      2
     ]
    }
   ],
   "sources": [
    {
     "title": "Micha Kaufman on LinkedIn: email to Fiverr employees",
     "url": "https://www.linkedin.com/posts/michakaufman_before-it-gets-out-somewhere-else-this-is-activity-7315378462070853632-BT79",
     "date": "2025-04-08",
     "type": "primary"
    },
    {
     "title": "ITPro: Fiverr CEO memo on AI upskilling for staff",
     "url": "https://www.itpro.com/business/careers-and-training/fiverr-ceo-ai-upskilling-staff-memo",
     "date": "2025-04-10",
     "type": "press"
    },
    {
     "title": "Fiverr Q2 2026 results (SEC Form 6-K exhibit 99.1)",
     "url": "https://www.sec.gov/Archives/edgar/data/1762301/000117891326003624/exhibit_99-1.htm",
     "date": "2026-07-29",
     "type": "primary"
    }
   ],
   "notes": "No 2026 public update on engineering AI practice found as of 2026-10-25. The Q2 2026 results discuss AI's effect on demand, not internal engineering.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Box",
   "slug": "box",
   "industry": "Enterprise SaaS",
   "size": "Under 5k employees",
   "size_src": [
    1
   ],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "CEO Aaron Levie published 'AI-first' principles: encourage teams to automate with AI and reinvest savings ('keep what you kill'), pick common AI platforms, and upskill every employee. No usage requirement is stated. By 2026 Cursor was rolled out across the org as Box's 'core IDE and agent platform', still with no stated requirement.",
    "src": [
     0,
     2
    ]
   },
   "tools": [
    {
     "name": "Cursor",
     "src": [
      2
     ]
    },
    {
     "name": "Box AI",
     "src": [
      0
     ]
    },
    {
     "name": "AI agents for coding (unnamed)",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "PR throughput",
    "Migration throughput"
   ],
   "metrics": [
    {
     "text": "Cursor's case study reports Box's impact as product-roadmap throughput and migration effort.",
     "src": [
      2
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Strong data governance and security principles, with human-in-the-loop still required for most areas.",
     "src": [
      0
     ]
    },
    {
     "text": "Common platforms across the company (e.g., Box AI for corporate knowledge) plus core platforms per function (e.g., AI agents for coding).",
     "src": [
      0
     ]
    },
    {
     "text": "Teams encode agent guardrails and standards as Cursor rules in code (e.g., a frontend 'AI toolkit') rather than in documentation.",
     "src": [
      2
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "'Keep what you kill': teams that save money with AI primarily reinvest the savings back into the same org for more strategic work.",
     "src": [
      0
     ]
    },
    {
     "text": "A mentorship program paired Cursor power users with newer adopters for an hour a week. Cursor reports it raised usage 75% in six weeks.",
     "src": [
      2
     ]
    },
    {
     "text": "In May 2026 Box posted an 'AI Business Automation Engineer' role, which Levie described as a forward-deployed engineer for internal functions.",
     "src": [
      3
     ]
    }
   ],
   "results": [
    {
     "text": "Box says it ideated over 100 AI agents across every function, then focused on a few 'big bets' to roll out at scale across 2,800 employees.",
     "src": [
      1
     ]
    },
    {
     "text": "Per Cursor's case study, 800+ developers (85% of engineering) use Cursor daily, Levie estimates a 30-50% roadmap throughput increase, and React and design-system migrations finished 80-90% faster than expected.",
     "src": [
      2
     ]
    }
   ],
   "sources": [
    {
     "title": "Aaron Levie on LinkedIn: At Box, we are focused on building an AI-first company",
     "url": "https://www.linkedin.com/posts/boxaaron_at-box-we-are-focused-on-building-an-ai-first-activity-7323454310032232449-Jy_O",
     "date": "2025-04-30",
     "type": "primary"
    },
    {
     "title": "Box blog: AI-first, Part 1",
     "url": "https://blog.box.com/ai-first-part-1",
     "date": "2025-09-04",
     "type": "company"
    },
    {
     "title": "Cursor customer story: Box chooses Cursor for enterprise-grade quality, security, and control",
     "url": "https://cursor.com/blog/box",
     "date": "2026-02-13",
     "type": "vendor"
    },
    {
     "title": "Business Insider: Box is hiring an AI Business Automation Engineer",
     "url": "https://www.businessinsider.com/box-hiring-ai-business-automation-engineer-agent-role-palantir-2026-5",
     "date": "2026-05-11",
     "type": "press"
    }
   ],
   "notes": "The Box blog date comes from the page's metadata, because no visible byline date was found.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Meta",
   "slug": "meta",
   "industry": "Big tech",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "Meta announced 'AI-driven impact' as a core performance expectation starting in 2026. By September 2026 it had reportedly stepped back, saying it would not use AI adoption dashboards or token counts to evaluate impact.",
    "src": [
     0,
     1
    ]
   },
   "tools": [
    {
     "name": "Metamate (internal)",
     "src": [
      0
     ]
    },
    {
     "name": "Hatch (internal agent, encouraged not required)",
     "src": [
      1
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate"
   ],
   "metrics": [
    {
     "text": "Individual AI usage metrics were not part of 2025 performance reviews, though employees were encouraged to cite AI wins.",
     "src": [
      0
     ]
    },
    {
     "text": "An employee-built internal AI-usage leaderboard was taken down after details leaked; Meta later told staff it would not use AI adoption dashboards or token counts to evaluate impact.",
     "src": [
      1
     ]
    }
   ],
   "guardrails": [],
   "hiring_reviews_changes": [
    {
     "text": "Meta said 'AI-driven impact' would become a core expectation in performance reviews starting in 2026.",
     "src": [
      0
     ]
    },
    {
     "text": "Meta allows job candidates to use AI in coding interviews.",
     "src": [
      0
     ]
    }
   ],
   "results": [
    {
     "text": "WIRED, citing other outlets' reporting, described a rise in bugs linked to reliance on AI-generated code. This is reported, not a company-published figure.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "Business Insider: Meta's AI-driven performance review overhaul",
     "url": "https://www.businessinsider.com/meta-ai-employee-performance-review-overhaul-2025-11",
     "date": "2025-11-14",
     "type": "press"
    },
    {
     "title": "WIRED: Meta pushes its new AI agent on employees but eases off on 'tokenmaxxing'",
     "url": "https://www.wired.com/story/meta-pushes-its-new-ai-agent-on-employees-but-eases-off-on-tokenmaxxing/",
     "date": "2026-09-02",
     "type": "press"
    }
   ],
   "notes": "The policy has changed over time, so read the sources in date order.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Google (Alphabet)",
   "slug": "google",
   "industry": "Big tech",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "Leadership has publicly urged AI use. No formal requirement has been published.",
    "src": [
     0,
     1
    ]
   },
   "tools": [
    {
     "name": "Antigravity",
     "src": [
      2
     ]
    },
    {
     "name": "Gemini models (internal agentic workflows)",
     "src": [
      2
     ]
    }
   ],
   "metric_tags": [
    "% code AI-generated",
    "Migration throughput",
    "Agent autonomy"
   ],
   "metrics": [
    {
     "text": "Superseded (October 2024): Pichai said more than a quarter of all new code at Google is AI-generated, then reviewed and accepted by engineers.",
     "src": [
      0
     ]
    },
    {
     "text": "Pichai said in April 2026 that 75% of all new code at Google is AI-generated and approved by engineers, up from 50% the previous fall.",
     "src": [
      2
     ]
    }
   ],
   "guardrails": [
    {
     "text": "AI-generated code is reviewed and accepted by engineers before it lands.",
     "src": [
      0,
      2
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Business Insider reported that Pichai told employees at a mid-2025 all-hands that they need to use AI to stay competitive.",
     "src": [
      1
     ]
    }
   ],
   "results": [
    {
     "text": "Pichai said a complex code migration done by agents and engineers together finished six times faster than a comparable one a year earlier, and that engineers are moving to agentic workflows.",
     "src": [
      2
     ]
    }
   ],
   "sources": [
    {
     "title": "Google: Q3 2024 earnings call remarks from Sundar Pichai",
     "url": "https://blog.google/company-news/inside-google/message-ceo/alphabet-earnings-q3-2024/",
     "date": "2024-10-29",
     "type": "primary"
    },
    {
     "title": "Business Insider: Meta's AI-driven performance review overhaul (references Google all-hands)",
     "url": "https://www.businessinsider.com/meta-ai-employee-performance-review-overhaul-2025-11",
     "date": "2025-11-14",
     "type": "press"
    },
    {
     "title": "Google: Cloud Next '26, momentum and innovation at Google scale (Sundar Pichai)",
     "url": "https://blog.google/innovation-and-ai/infrastructure-and-cloud/google-cloud/cloud-next-2026-sundar-pichai/",
     "date": "2026-04-22",
     "type": "primary"
    }
   ],
   "notes": "",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Microsoft",
   "slug": "microsoft",
   "industry": "Big tech",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "yes",
   "mandate_summary": {
    "text": "Julia Liuson, president of the Developer Division, wrote in a June 2025 memo that 'using AI is no longer optional'. The memo covers her division and is not a company-wide policy.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "GitHub Copilot",
     "src": [
      0
     ]
    },
    {
     "name": "GitHub Copilot CLI",
     "src": [
      2
     ]
    },
    {
     "name": "Claude Code (being wound down in E+D)",
     "src": [
      2
     ]
    }
   ],
   "metric_tags": [
    "% code AI-generated",
    "Adoption rate"
   ],
   "metrics": [
    {
     "text": "Satya Nadella said in April 2025 that 20% to 30% of the code in some Microsoft repos and projects is written by software (AI).",
     "src": [
      1
     ]
    },
    {
     "text": "The Verge noted Microsoft said in 2025 that 91% of its engineering teams used GitHub Copilot. No updated 2026 share-of-code figure was found.",
     "src": [
      2
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Business Insider reported that some teams may use external AI tools if those tools meet security requirements.",
     "src": [
      0
     ]
    },
    {
     "text": "After opening Claude Code to thousands of developers in December 2025, the Experiences + Devices group is winding it down by June 30, 2026 and standardizing on GitHub Copilot CLI. Claude models stay available through Copilot CLI.",
     "src": [
      2
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Managers were told AI use should factor into holistic performance reflections. Business Insider reported that some teams were considering a more formal usage metric.",
     "src": [
      0
     ]
    }
   ],
   "results": [],
   "sources": [
    {
     "title": "Business Insider: Microsoft internal memo says using AI is no longer optional",
     "url": "https://www.businessinsider.com/microsoft-internal-memo-using-ai-no-longer-optional-github-copilot-2025-6",
     "date": "2025-06-27",
     "type": "press"
    },
    {
     "title": "CNBC: Satya Nadella says as much as 30% of Microsoft code is written by AI",
     "url": "https://www.cnbc.com/2025/04/29/satya-nadella-says-as-much-as-30percent-of-microsoft-code-is-written-by-ai.html",
     "date": "2025-04-29",
     "type": "press"
    },
    {
     "title": "The Verge: Microsoft starts canceling Claude Code licenses",
     "url": "https://www.theverge.com/tech/930447/microsoft-claude-code-discontinued-notepad",
     "date": "2026-05-14",
     "type": "press"
    }
   ],
   "notes": "The mandate is scoped to the Developer Division as reported. The 2026 Claude Code wind-down is a tool-standardization move, not a new usage mandate.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Amazon",
   "slug": "amazon",
   "industry": "Big tech",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "A November 2025 internal memo made Kiro Amazon's recommended AI development tool and said Amazon does not plan to support additional third-party AI development tools. That is a tool-standardization policy, not a reported usage mandate. In 2026 Stores engineering set AI-native targets for teams, but a spokesperson said Amazon does not 'centrally mandate that teams use AI tools'.",
    "src": [
     1,
     2,
     3
    ]
   },
   "tools": [
    {
     "name": "Amazon Q Developer",
     "src": [
      0
     ]
    },
    {
     "name": "Kiro",
     "src": [
      1,
      3
     ]
    },
    {
     "name": "AI Teammate (internal)",
     "src": [
      3
     ]
    },
    {
     "name": "Pippin (internal)",
     "src": [
      3
     ]
    }
   ],
   "metric_tags": [
    "Time saved",
    "Cost savings",
    "Adoption rate",
    "PR throughput"
   ],
   "metrics": [
    {
     "text": "Amazon estimated time saved from the number of Java dependencies migrated, and cost savings from hosts removed after the Java 17 upgrades.",
     "src": [
      0
     ]
    },
    {
     "text": "A February 2026 internal document asked 2,100+ Stores engineering teams to triple release velocity, with at least 25 teams aiming for 10x, and set a goal of 80% of teams on AI-native practices (about 60% were at the time).",
     "src": [
      3
     ]
    },
    {
     "text": "Tracked metrics include monthly active users, NPS, a 'Value Deriving Event' and deployments per engineer per week. The document flags Goodhart's Law as a risk.",
     "src": [
      3
     ]
    }
   ],
   "guardrails": [
    {
     "text": "The memo states: 'While we continue to support existing tools in use today, we do not plan to support additional third party, AI development tools.'",
     "src": [
      1
     ]
    },
    {
     "text": "According to a memo reviewed by Reuters, Amazon revised its internal guidance for OpenAI's Codex to 'Do Not Use' in October 2025.",
     "src": [
      1
     ]
    },
    {
     "text": "After retail outages, including one tied to Amazon Q, a 90-day policy for about 335 Tier-1 systems requires two-person review, documented approval through Modeled Change Management, and automated reliability checks. Amazon said only one reviewed incident was AI-related.",
     "src": [
      4
     ]
    },
    {
     "text": "Amazon's six AI-native engineering tenets include 'No black boxes'.",
     "src": [
      3
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "After employee pushback on top-down targets, the program shifted toward 'collaborative AI practices'.",
     "src": [
      3
     ]
    }
   ],
   "results": [
    {
     "text": "Amazon says Amazon Q Developer helped it migrate tens of thousands of production apps from Java 8/11 to Java 17. It puts the savings at more than 4,500 developer-years and $260M a year.",
     "src": [
      0
     ]
    },
    {
     "text": "AI Teammate, a Slack-integrated agent, has expanded to 700+ active teams, and some groups, including parts of AWS, have adopted Pippin (ideas to technical designs) more broadly.",
     "src": [
      3
     ]
    }
   ],
   "sources": [
    {
     "title": "AWS DevOps Blog: Amazon Q Developer just reached a $260 million dollar milestone",
     "url": "https://aws.amazon.com/blogs/devops/amazon-q-developer-just-reached-a-260-million-dollar-milestone/",
     "date": "2024-08-01",
     "type": "company"
    },
    {
     "title": "Reuters via CNA: Amazon pushes in-house AI coding tool Kiro over competitors, memo shows",
     "url": "https://www.channelnewsasia.com/business/amazon-pushes-in-house-ai-coding-tool-kiro-over-competitors-memo-shows-5488266",
     "date": "2025-11-25",
     "type": "press"
    },
    {
     "title": "Reuters (original; may be paywalled or block automated checks)",
     "url": "https://www.reuters.com/business/retail-consumer/amazon-pushes-in-house-ai-coding-tool-kiro-over-competitors-memo-shows-2025-11-25/",
     "date": "2025-11-25",
     "type": "press"
    },
    {
     "title": "Business Insider: Amazon pushes AI use and closely tracks adoption, as some employees push back",
     "url": "https://www.businessinsider.com/amazon-tracks-ai-use-engineers-internal-friction-2026-4",
     "date": "2026-04-27",
     "type": "press"
    },
    {
     "title": "Business Insider: Amazon orders 90-day reset after code mishaps cause millions of lost orders",
     "url": "https://www.businessinsider.com/amazon-tightens-code-controls-after-outages-including-one-ai-2026-3",
     "date": "2026-03-10",
     "type": "press"
    }
   ],
   "notes": "The Reuters original returned HTTP 401 to automated checks. Use the syndicated CNA copy, which loads.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Stripe",
   "slug": "stripe",
   "industry": "Fintech / payments",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "Stripe publishes how it built internal coding agents ('Minions'). No usage mandate has been reported.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Minions (internal, forked from Block's goose)",
     "src": [
      0
     ]
    },
    {
     "name": "Claude Code",
     "src": [
      0
     ]
    },
    {
     "name": "Cursor",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "PR throughput"
   ],
   "metrics": [
    {
     "text": "Stripe counts merged pull requests produced end to end by Minions.",
     "src": [
      0,
      1
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Minions run in isolated developer environments, and their pull requests are reviewed by humans.",
     "src": [
      0,
      1
     ]
    },
    {
     "text": "The agents get internal tools through a central MCP server ('Toolshed') and get a limited number of CI attempts.",
     "src": [
      0,
      1
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "Stripe says more than 1,000 pull requests merged each week are entirely produced by Minions and reviewed by humans.",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "Stripe Dev Blog: Minions, Stripe's one-shot, end-to-end coding agents",
     "url": "https://stripe.dev/blog/minions-stripes-one-shot-end-to-end-coding-agents",
     "date": "2026-02-09",
     "type": "company"
    },
    {
     "title": "InfoQ: Stripe's autonomous coding agents",
     "url": "https://www.infoq.com/news/2026/03/stripe-autonomous-coding-agents/",
     "date": "2026-03-20",
     "type": "press"
    }
   ],
   "notes": "Fintech. It is not a bank.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Zapier",
   "slug": "zapier",
   "industry": "SaaS / automation",
   "size": "Under 5k employees",
   "size_src": [],
   "mandate": "yes",
   "mandate_summary": {
    "text": "Since May 2025, 100% of new Zapier hires must meet an AI-fluency bar. AI fluency is also part of performance expectations for existing staff.",
    "src": [
     0,
     1
    ]
   },
   "tools": [
    {
     "name": "Cursor",
     "src": [
      2
     ]
    },
    {
     "name": "Claude Code",
     "src": [
      2
     ]
    },
    {
     "name": "GitHub Copilot",
     "src": [
      2
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "Self-reported productivity"
   ],
   "metrics": [
    {
     "text": "Zapier reports the share of employees using AI daily and changes in developer productivity.",
     "src": [
      2
     ]
    }
   ],
   "guardrails": [],
   "hiring_reviews_changes": [
    {
     "text": "All new hires are assessed against an AI-fluency rubric. Version 2 (March 2026) uses levels from 'Unacceptable' up to 'Transformative' and adds accountability as a component.",
     "src": [
      0,
      1
     ]
    },
    {
     "text": "Managers are expected to show they have led AI adoption on their teams.",
     "src": [
      1
     ]
    }
   ],
   "results": [
    {
     "text": "Zapier reports that 97% of employees use AI daily, and that developer productivity rose 11% in 2025 with Cursor, Claude Code and GitHub Copilot.",
     "src": [
      2
     ]
    }
   ],
   "sources": [
    {
     "title": "Wade Foster on LinkedIn: We're setting a new standard at Zapier",
     "url": "https://www.linkedin.com/posts/wadefoster_were-setting-a-new-standard-at-zapier-activity-7334216816732360705-rLZ1",
     "date": "2025-05-30",
     "type": "primary"
    },
    {
     "title": "Zapier blog: Raising the AI-fluency bar in hiring",
     "url": "https://zapier.com/blog/raising-ai-fluency-bar-in-hiring/",
     "date": "2026-03-31",
     "type": "company"
    },
    {
     "title": "Zapier blog: How to be an AI transformation leader",
     "url": "https://zapier.com/blog/ai-transformation-leader/",
     "date": "2025-07-30",
     "type": "company"
    }
   ],
   "notes": "The AI transformation post was published 2025-07-30 and last modified 2025-10-29, per page metadata.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Canva",
   "slug": "canva",
   "industry": "Design / SaaS",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "Canva requires AI use in engineering interviews. No usage mandate for current engineers has been reported.",
    "src": [
     0,
     1
    ]
   },
   "tools": [
    {
     "name": "GitHub Copilot",
     "src": [
      0
     ]
    },
    {
     "name": "Cursor",
     "src": [
      0
     ]
    },
    {
     "name": "Claude",
     "src": [
      0
     ]
    },
    {
     "name": "Claude for Work, ChatGPT Enterprise, Gemini, Glean",
     "src": [
      3
     ]
    },
    {
     "name": "Forge (internal parallel-agent manager)",
     "src": [
      3
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "Agent autonomy",
    "Code review"
   ],
   "metrics": [
    {
     "text": "Superseded (June 2025): ACS reported that almost half of Canva's frontend and backend engineers use AI coding tools daily.",
     "src": [
      1
     ]
    },
    {
     "text": "Canva says nine in ten employees use AI every week. Forge, an internal tool for managing parallel AI agents, has 475 monthly active users and is now supported DevX tooling.",
     "src": [
      3
     ]
    }
   ],
   "guardrails": [
    {
     "text": "CTO Brendan Humphreys said engineers apply a 'human touch' to meet the quality bar, and that domain mastery is needed to verify output and keep complexity out of Canva's ~70 million lines of code.",
     "src": [
      2
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Candidates for backend, frontend and ML engineering roles are expected to use AI tools in interviews. An 'AI-Assisted Coding' interview replaces the CS Fundamentals screen.",
     "src": [
      0,
      1
     ]
    },
    {
     "text": "Humphreys said senior engineers now describe their jobs as 'largely review': writing detailed instructions for agents that run in the background, sometimes overnight.",
     "src": [
      2
     ]
    },
    {
     "text": "Canva gave all 5,300 employees a week off normal work for its second AI Discovery Week in 2026.",
     "src": [
      3
     ]
    }
   ],
   "results": [
    {
     "text": "Humphreys said agents deliver 'hours and hours' of work done autonomously.",
     "src": [
      2
     ]
    }
   ],
   "sources": [
    {
     "title": "Canva Engineering: Yes, you can use AI in our interviews",
     "url": "https://www.canva.dev/blog/engineering/yes-you-can-use-ai-in-our-interviews/",
     "date": "2025-06-11",
     "type": "company"
    },
    {
     "title": "ACS Information Age: Canva insists job applicants use AI coding in interviews",
     "url": "https://ia.acs.org.au/article/2025/canva-insists-job-applicants-use-ai-coding-in-interviews.html",
     "date": "2025-06-13",
     "type": "press"
    },
    {
     "title": "Business Insider: AI agents are transforming what it's like to be a coder (Canva CTO Brendan Humphreys)",
     "url": "https://www.businessinsider.com/canva-ai-agents-are-changing-engineering-work-2026-2",
     "date": "2026-02-15",
     "type": "press"
    },
    {
     "title": "Canva Newsroom: Inside our second 'AI Discovery Week'",
     "url": "https://www.canva.com/newsroom/news/ai-discovery-week-2026/",
     "date": "2026-05-19",
     "type": "company"
    }
   ],
   "notes": "The Canva newsroom returns HTTP 403 to automated checks (bot protection) but loaded for readers on 2026-10-25. The post date is from page metadata.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Airbnb",
   "slug": "airbnb",
   "industry": "Travel / marketplace",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "Airbnb published a case study of an LLM-driven code migration. No mandate has been reported. In 2026 the CEO reported that most new code is AI-written, again without a stated requirement.",
    "src": [
     0,
     2,
     4
    ]
   },
   "tools": [
    {
     "name": "Frontier LLMs (unnamed)",
     "src": [
      0
     ]
    },
    {
     "name": "Claude Code",
     "src": [
      3
     ]
    }
   ],
   "metric_tags": [
    "Time saved",
    "Migration throughput",
    "% code AI-generated"
   ],
   "metrics": [
    {
     "text": "Airbnb compared the migration's duration with an estimate of how long it would take by hand, and tracked the share of files migrated per run.",
     "src": [
      0
     ]
    },
    {
     "text": "Brian Chesky said on the Q1 2026 call that nearly 60% of the code Airbnb engineers produce is now written by AI, which he estimated at about twice the industry average.",
     "src": [
      2,
      4
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Each file went through automated validation steps with retries, and the files the pipeline could not finish were completed by hand.",
     "src": [
      0,
      1
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Chesky said there is no room for 'pure people managers'. Many design and engineering managers are going back to coding or using Claude Code.",
     "src": [
      3,
      4
     ]
    }
   ],
   "results": [
    {
     "text": "Airbnb migrated nearly 3.5K test files from Enzyme to React Testing Library in 6 weeks. It had estimated the work at 1.5 years by hand. The first bulk run migrated 75% of files in four hours, and the pipeline reached 97% before the rest were finished manually.",
     "src": [
      0,
      1
     ]
    },
    {
     "text": "Chesky said teams are shipping more features and iterating more quickly as a result.",
     "src": [
      4
     ]
    }
   ],
   "sources": [
    {
     "title": "Airbnb Tech: Accelerating large-scale test migration with LLMs",
     "url": "https://airbnb.tech/infrastructure/accelerating-large-scale-test-migration-with-llms/",
     "date": "2025-03",
     "type": "company"
    },
    {
     "title": "InfoQ: Airbnb uses LLMs to migrate tests",
     "url": "https://www.infoq.com/news/2025/03/airbnb-llm-test-migration/",
     "date": "2025-03-28",
     "type": "press"
    },
    {
     "title": "TechCrunch: Airbnb says AI now writes 60% of its new code",
     "url": "https://techcrunch.com/2026/05/08/airbnb-says-ai-now-writes-60-of-its-new-code/",
     "date": "2026-05-08",
     "type": "press"
    },
    {
     "title": "Business Insider: Airbnb CEO says AI writes majority of code, managers going hands-on",
     "url": "https://www.businessinsider.com/airbnb-ceo-ai-write-majority-company-code-managers-hands-on-2026-5",
     "date": "2026-05-08",
     "type": "press"
    },
    {
     "title": "Motley Fool: Airbnb (ABNB) Q1 2026 earnings call transcript",
     "url": "https://www.fool.com/earnings/call-transcripts/2026/05/07/airbnb-abnb-q1-2026-earnings-call-transcript/",
     "date": "2026-05-07",
     "type": "press"
    }
   ],
   "notes": "The Airbnb post shows no visible date. The month is taken from contemporaneous InfoQ coverage.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Intercom",
   "slug": "intercom",
   "industry": "SaaS / customer service AI",
   "size": "Under 5k employees",
   "size_src": [
    1
   ],
   "mandate": "yes",
   "mandate_summary": {
    "text": "Intercom publicly committed to doubling R&D productivity within 12 months. It set targets for the share of pull requests driven by agents (Claude Code) and approved by AI.",
    "src": [
     0,
     1
    ]
   },
   "tools": [
    {
     "name": "Claude Code",
     "src": [
      0,
      1,
      2
     ]
    }
   ],
   "metric_tags": [
    "PR throughput",
    "Quality / defects",
    "Adoption rate",
    "Cost per PR"
   ],
   "metrics": [
    {
     "text": "Intercom tracks PRs per R&D person per month, the product defect backlog, idea-to-ship speed, downtime from breaking changes, cost per PR, the share of agent-driven PRs and the share of AI-approved PRs.",
     "src": [
      0,
      1
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Intercom says it has built shared Claude Code plugins, skills and hooks to encode its standards.",
     "src": [
      2
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Code review changed: 19% of merged PRs are now auto-approved by AI with no human review.",
     "src": [
      1
     ]
    }
   ],
   "results": [
    {
     "text": "PRs per R&D person rose from 16 to 33 per month, and the product defect backlog fell 54%.",
     "src": [
      1
     ]
    },
    {
     "text": "Product changes ship 39% faster, and downtime from breaking changes fell 35% even as deployments doubled.",
     "src": [
      1
     ]
    },
    {
     "text": "Fully loaded cost per PR was cut in half, and 93.6% of PRs are agent-driven via Claude Code.",
     "src": [
      1
     ]
    },
    {
     "text": "1,100 of Intercom's 1,305 employees use Claude Code.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "Ideas by Fin/Intercom: 2x, nine months later",
     "url": "https://ideas.fin.ai/p/2x-nine-months-later",
     "date": "2026-04-16",
     "type": "company"
    },
    {
     "title": "Darragh Curran (Intercom CTO) on LinkedIn: Nine months ago, we publicly committed to 2x...",
     "url": "https://www.linkedin.com/feed/update/urn:li:activity:7450587155372015616",
     "date": "2026-04-16",
     "type": "primary"
    },
    {
     "title": "Ideas by Fin/Intercom: How we use Claude Code today at Intercom",
     "url": "https://ideas.fin.ai/p/how-we-use-claude-code-today-at-intercom",
     "date": "2026-03-19",
     "type": "company"
    }
   ],
   "notes": "The mandate is classed as 'yes' because leadership set explicit, organization-wide adoption targets.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Spotify",
   "slug": "spotify",
   "industry": "Consumer media",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "Spotify publishes engineering posts about its background coding agent (Honk). No mandate has been reported.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Honk (internal background coding agent)",
     "src": [
      0,
      1
     ]
    },
    {
     "name": "Fleet Management / Fleetshift (internal)",
     "src": [
      0,
      1
     ]
    },
    {
     "name": "Backstage",
     "src": [
      1
     ]
    }
   ],
   "metric_tags": [
    "PR throughput",
    "Time saved"
   ],
   "metrics": [
    {
     "text": "Spotify counts merged agent-generated PRs and estimates the engineering weeks saved on migrations.",
     "src": [
      0,
      1
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Agent changes go through verification loops before PRs are opened, including an LLM-as-judge check on diffs.",
     "src": [
      0
     ]
    },
    {
     "text": "Downstream owning teams still did their own manual testing before merging the automated migration PRs.",
     "src": [
      1
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "Spotify says the background agent has merged 1,500+ PRs, and that Fleet Management has automated roughly half of its PRs since mid-2024.",
     "src": [
      0
     ]
    },
    {
     "text": "Spotify rolled out 240 automated dataset-migration PRs, saving an estimated 10 engineering weeks.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "Spotify Engineering: Spotify's background coding agent (Part 1)",
     "url": "https://engineering.atspotify.com/2025/11/spotifys-background-coding-agent-part-1",
     "date": "2025-11",
     "type": "company"
    },
    {
     "title": "Spotify Engineering: Background coding agents and dataset migrations (Honk, Part 4)",
     "url": "https://engineering.atspotify.com/2026/4/background-coding-agents-dataset-migrations-honk-part-4",
     "date": "2026-04",
     "type": "company"
    }
   ],
   "notes": "Dates are month-level, taken from the URL paths. The day was not found on the pages.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Block",
   "slug": "block",
   "industry": "Fintech / payments",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "Block describes broad tool access and an AI Champions program. It does not describe a formal usage mandate.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "goose (Block, open source)",
     "src": [
      0,
      1
     ]
    },
    {
     "name": "Claude Code",
     "src": [
      0
     ]
    },
    {
     "name": "Cursor",
     "src": [
      0
     ]
    },
    {
     "name": "GitHub Copilot",
     "src": [
      0
     ]
    },
    {
     "name": "Cline",
     "src": [
      0
     ]
    },
    {
     "name": "Amp",
     "src": [
      0
     ]
    },
    {
     "name": "Firebender",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "Time saved",
    "PR throughput",
    "AI-authored code"
   ],
   "metrics": [
    {
     "text": "Block tracks AI's contribution through PR labels and Co-authored-by attribution, and measures time savings and automated PRs.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Repositories carry agent guidance files (e.g. AGENTS.md), and AI-assisted changes are attributed in commits and PRs.",
     "src": [
      0
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Block runs an AI Champions program in which engineers spend part of their time driving AI adoption within teams.",
     "src": [
      0
     ]
    }
   ],
   "results": [
    {
     "text": "Block reports that about 95% of engineers regularly use AI. Within three months of the Champions program, AI-authored code rose 69%, time savings rose 37%, and automated PRs grew 21x.",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "Block Engineering: AI-assisted development at Block",
     "url": "https://engineering.block.xyz/blog/ai-assisted-development-at-block",
     "date": "2026-01-18",
     "type": "company"
    },
    {
     "title": "Block: Block Open Source introduces codename goose",
     "url": "https://block.xyz/inside/block-open-source-introduces-codename-goose",
     "date": "2025-01-28",
     "type": "company"
    }
   ],
   "notes": "Fintech. It is not a bank.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Atlassian",
   "slug": "atlassian",
   "industry": "Enterprise SaaS",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "Atlassian published internal results from its own AI code reviewer. No mandate has been reported.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Rovo Dev Code Reviewer (Atlassian)",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Cycle time",
    "Code review"
   ],
   "metrics": [
    {
     "text": "Atlassian measures median PR cycle time, human review comments per PR, and the resolution rate of AI comments versus human comments.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [
    {
     "text": "AI review comments are filtered for quality before posting, and humans make the final review decision.",
     "src": [
      0
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "Atlassian reports a 30.8% reduction in median PR cycle time and 35.6% fewer human review comments across 1,900+ repos.",
     "src": [
      0
     ]
    },
    {
     "text": "AI review comments were resolved 38.7% of the time, versus 44.45% for human comments.",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "Atlassian: Developer productivity improved with Rovo Dev",
     "url": "https://www.atlassian.com/blog/ai-at-work/developer-productivity-improved-with-rovo-dev",
     "date": "2026-04-07",
     "type": "vendor"
    }
   ],
   "notes": "Vendor-authored. Atlassian is reporting on its own commercial product.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Robinhood",
   "slug": "robinhood",
   "industry": "Fintech / brokerage",
   "size": "Under 5k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "The CEO reported near-universal adoption. No formal mandate has been reported.",
    "src": [
     0,
     1
    ]
   },
   "tools": [
    {
     "name": "GitHub Copilot",
     "src": [
      0
     ]
    },
    {
     "name": "Cursor",
     "src": [
      0
     ]
    },
    {
     "name": "Windsurf",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "% code AI-generated",
    "PR throughput"
   ],
   "metrics": [
    {
     "text": "Superseded (July 2025): CEO Vlad Tenev said close to 100% of engineers use AI code editors and about half of new code is AI-generated, a measurement he called imprecise.",
     "src": [
      0
     ]
    },
    {
     "text": "On the Q1 2026 call Robinhood said more than 90% of employees use AI tools and that it tracks commits per engineer deployed to production.",
     "src": [
      1
     ]
    }
   ],
   "guardrails": [],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "Commits per engineer hit a new high in Q1 2026, up 50% since the start of 2025, which management attributed to AI tools.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "Business Insider: Robinhood CEO says majority of new code is AI-generated",
     "url": "https://www.businessinsider.com/robinhood-ceo-majority-new-code-ai-generated-engineer-adoption-2025-7",
     "date": "2025-07-17",
     "type": "press"
    },
    {
     "title": "Motley Fool: Robinhood (HOOD) Q1 2026 earnings call transcript",
     "url": "https://www.fool.com/earnings/call-transcripts/2026/04/28/robinhood-hood-q1-2026-earnings-transcript/",
     "date": "2026-04-28",
     "type": "press"
    }
   ],
   "notes": "Fintech or brokerage. It is not a bank.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Salesforce",
   "slug": "salesforce",
   "industry": "Enterprise SaaS",
   "size": "50k+ employees",
   "size_src": [
    0
   ],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "The company announced a hiring freeze for software engineers, citing AI productivity. No usage mandate has been reported. By 2026 Cursor was offered alongside internal tools as an option, with no requirement reported.",
    "src": [
     0,
     1,
     2,
     3
    ]
   },
   "tools": [
    {
     "name": "Agentforce (Salesforce)",
     "src": [
      0
     ]
    },
    {
     "name": "Cursor",
     "src": [
      3
     ]
    }
   ],
   "metric_tags": [
    "Self-reported productivity",
    "Adoption rate",
    "Cycle time",
    "PR throughput",
    "Quality / defects"
   ],
   "metrics": [
    {
     "text": "Marc Benioff said engineering productivity rose by more than 30% with Agentforce and other AI technology.",
     "src": [
      0
     ]
    },
    {
     "text": "Salesforce tracks cycle time, quality (bug count) and throughput on existing dashboards, per Cursor's case study.",
     "src": [
      3
     ]
    }
   ],
   "guardrails": [],
   "hiring_reviews_changes": [
    {
     "text": "Benioff said Salesforce would not add any more software engineers in 2025.",
     "src": [
      0,
      1
     ]
    },
    {
     "text": "In January 2026 Benioff said he had held engineering headcount (about 15,000) mostly flat because of AI productivity, while hiring 20% more account executives.",
     "src": [
      2
     ]
    }
   ],
   "results": [
    {
     "text": "Per Cursor, over 75% of Salesforce developers use Cursor, PR velocity rose 30%+, and one team cut legacy test-coverage time by 85%.",
     "src": [
      3
     ]
    }
   ],
   "sources": [
    {
     "title": "Salesforce Ben: Salesforce will hire no more software engineers in 2025, says Marc Benioff",
     "url": "https://www.salesforceben.com/salesforce-will-hire-no-more-software-engineers-in-2025-says-marc-benioff/",
     "date": "2024-12-18",
     "type": "press"
    },
    {
     "title": "The Register: Salesforce won't hire more engineers in 2025 due to AI tools",
     "url": "https://www.theregister.com/software/2025/02/27/salesforce-wont-hire-more-engineers-in-2025-due-to-ai-tools/462789",
     "date": "2025-02-27",
     "type": "press"
    },
    {
     "title": "ITPro: Marc Benioff says hiring in software engineering is 'mostly flat' at Salesforce because of AI",
     "url": "https://www.itpro.com/business/business-strategy/marc-benioff-salesforce-software-engineering-hiring-flat-ai",
     "date": "2026-01-15",
     "type": "press"
    },
    {
     "title": "Cursor customer story: Salesforce accelerates velocity by over 30% with Cursor",
     "url": "https://cursor.com/blog/salesforce",
     "date": "2026-01-21",
     "type": "vendor"
    }
   ],
   "notes": "Salesforce Ben returns 403 to automated checks but loads for readers.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Uber",
   "slug": "uber",
   "industry": "Mobility / marketplace",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "Uber published details of its in-house AI code reviewer (uReview). In 2026 the CTO said the strongest adoption came from engineers experimenting, not from a top-down push. No mandate has been reported.",
    "src": [
     0,
     1
    ]
   },
   "tools": [
    {
     "name": "uReview (internal)",
     "src": [
      0
     ]
    },
    {
     "name": "Claude 4 Sonnet (comment generator)",
     "src": [
      0
     ]
    },
    {
     "name": "OpenAI o4-mini-high (grader)",
     "src": [
      0
     ]
    },
    {
     "name": "Internal coding agent",
     "src": [
      1
     ]
    },
    {
     "name": "Claude Code",
     "src": [
      2
     ]
    }
   ],
   "metric_tags": [
    "Code review",
    "Time saved",
    "Adoption rate",
    "AI-authored code",
    "Agent autonomy"
   ],
   "metrics": [
    {
     "text": "Uber measures the usefulness rate (developer ratings), the share of comments addressed in the same changeset, and estimated review hours saved.",
     "src": [
      0
     ]
    },
    {
     "text": "CTO Praveen Neppalli Naga said in March 2026 that 95% of engineers use AI monthly, and that Uber's internal agent writes about 1,800 code changes a week entirely, up from under 1% to 8% of all changes.",
     "src": [
      1
     ]
    },
    {
     "text": "CEO Dara Khosrowshahi said in May 2026 that roughly 10% of code changes come from autonomous agents.",
     "src": [
      2
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Comments go through multi-stage filtering (confidence grading, deduplication and category suppression) before posting. The tool was rolled out gradually with precision and recall dashboards.",
     "src": [
      0
     ]
    },
    {
     "text": "Agent-written changes have 'zero human authoring', but engineers review and approve them before they merge.",
     "src": [
      1,
      2
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Code review changed: an AI reviewer now runs on every commit in CI across all six of Uber's monorepos, in addition to human review.",
     "src": [
      0
     ]
    },
    {
     "text": "Uber said in May 2026 that it is slowing hiring to fund AI investment. The CTO said it had already spent its 2026 Claude Code budget.",
     "src": [
      2
     ]
    }
   ],
   "results": [
    {
     "text": "uReview analyzes over 90% of Uber's ~65,000 weekly diffs. Engineers mark 75% of its comments useful, and 65%+ are addressed, versus 51% for human comments.",
     "src": [
      0
     ]
    },
    {
     "text": "Uber estimates uReview saves about 1,500 hours each week, nearly 39 developer-years a year.",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "Uber Engineering: uReview, scalable, trustworthy GenAI for code review at Uber",
     "url": "https://www.uber.com/us/en/blog/ureview/",
     "date": "2025-08-12",
     "type": "company"
    },
    {
     "title": "Business Insider: Uber CTO says it's going big on AI coding",
     "url": "https://www.businessinsider.com/uber-cto-ai-coding-agentic-software-engineers-2026-3",
     "date": "2026-03-18",
     "type": "press"
    },
    {
     "title": "Business Insider: Uber CEO says the company is slowing hiring as it invests in AI",
     "url": "https://www.businessinsider.com/uber-slowing-hiring-fund-ai-investment-2026-5",
     "date": "2026-05-06",
     "type": "press"
    }
   ],
   "notes": "",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Nvidia",
   "slug": "nvidia",
   "industry": "Semiconductors / AI",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "yes",
   "mandate_summary": {
    "text": "Cursor's February 2026 case study says Nvidia 'set a new engineering mandate' to use Cursor to embed AI across every phase of the SDLC. Jensen Huang has said 100% of engineers use Cursor and pitched token budgets worth about half an engineer's salary.",
    "src": [
     0,
     1,
     2
    ]
   },
   "tools": [
    {
     "name": "Cursor",
     "src": [
      0,
      1
     ]
    },
    {
     "name": "OpenAI Codex (GPT-5.5)",
     "src": [
      3
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "PR throughput",
    "Quality / defects",
    "Code review"
   ],
   "metrics": [
    {
     "text": "Jensen Huang said 100% of Nvidia's software engineers and chip designers use Cursor.",
     "src": [
      0
     ]
    },
    {
     "text": "Nvidia measures adoption, coding velocity (code committed) and code quality (bug rates), per Cursor's case study.",
     "src": [
      1
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Teams use custom Cursor rules to automate workflows such as git flow, CI debugging and bug fixes, with Cursor running tests for validation.",
     "src": [
      1
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Huang proposed giving engineers AI tokens worth about half their base pay on top of salary, because 'every engineer that has access to tokens will be more productive'.",
     "src": [
      2
     ]
    }
   ],
   "results": [
    {
     "text": "Per Cursor, 30,000+ Nvidia developers use Cursor daily and commit three times more code, with bug rates flat.",
     "src": [
      1
     ]
    },
    {
     "text": "In April 2026 Huang emailed staff that OpenAI's Codex was available to every employee after early access for about 10,000.",
     "src": [
      3
     ]
    }
   ],
   "sources": [
    {
     "title": "Business Insider: Jensen Huang name-checks 6 AI companies, says 100% of Nvidia engineers use one of them",
     "url": "https://www.businessinsider.com/jensen-huang-hot-ai-startups-nvidia-vibe-coding-cursor-openai-2025-10",
     "date": "2025-10-15",
     "type": "press"
    },
    {
     "title": "Cursor customer story: NVIDIA commits 3x more code across 30,000 developers",
     "url": "https://cursor.com/blog/nvidia",
     "date": "2026-02-06",
     "type": "vendor"
    },
    {
     "title": "CNBC: Nvidia's Huang pitches AI tokens on top of salary",
     "url": "https://www.cnbc.com/2026/03/20/nvidia-ai-agents-tokens-human-workers-engineer-jobs-unemployment-jensen-huang.html",
     "date": "2026-03-20",
     "type": "press"
    },
    {
     "title": "Economic Times: Nvidia rolls out OpenAI's Codex AI agent to all employees",
     "url": "https://economictimes.indiatimes.com/tech/artificial-intelligence/nvidia-rolls-out-openais-codex-ai-agent-to-all-employees/articleshow/130502339.cms",
     "date": "2026-04-25",
     "type": "press"
    }
   ],
   "notes": "Mandate reclassified unclear to yes on 2026-10-25 on the vendor-reported 'engineering mandate' plus the CEO's universal-use statements. The mandate wording comes from Cursor, the tool vendor.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Anthropic",
   "slug": "anthropic",
   "industry": "AI lab",
   "size": "Under 5k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "Anthropic published internal research on its own engineers' use of Claude. No mandate is described.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Claude Code",
     "src": [
      0
     ]
    },
    {
     "name": "Automated Claude code reviewer",
     "src": [
      1
     ]
    }
   ],
   "metric_tags": [
    "Self-reported productivity",
    "PR throughput",
    "Agent autonomy",
    "AI-authored code",
    "% code AI-generated",
    "Code review",
    "Quality / defects"
   ],
   "metrics": [
    {
     "text": "Anthropic surveyed 132 engineers and researchers, interviewed 53, and analyzed 200,000 internal Claude Code transcripts (August 2025).",
     "src": [
      0
     ]
    },
    {
     "text": "As of May 2026, more than 80% of code merged into Anthropic's codebase was attributed to Claude, up from low single digits before February 2025. Leadership estimates 90%+ including scripts.",
     "src": [
      1
     ]
    },
    {
     "text": "The typical engineer merged about 8x as much code per day in Q2 2026 as in 2024. Anthropic says lines of code overstate the true gain.",
     "src": [
      1
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Most respondents said they can 'fully delegate' only 0-20% of their work. Work generally involves active supervision and validation.",
     "src": [
      0
     ]
    },
    {
     "text": "Every proposed change is read by an automated Claude reviewer before merge. A retrospective found it would have caught about a third of the bugs behind past claude.ai incidents.",
     "src": [
      1
     ]
    },
    {
     "text": "Lesse said agents already fix tests autonomously with no human interaction, but leaders are moving carefully on letting agents change the actual codebase on their own.",
     "src": [
      2
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Anthropic says human code review has become a bottleneck as code volume grows.",
     "src": [
      1
     ]
    },
    {
     "text": "Business Insider reported that an Anthropic report from October 2026 said it has hired for a smaller share of junior technical positions as agents took on more work.",
     "src": [
      2
     ]
    }
   ],
   "results": [
    {
     "text": "Respondents self-reported using Claude in 59% of their work, for a +50% productivity gain. Anthropic saw a 67% increase in merged PRs per engineer per day after adopting Claude Code.",
     "src": [
      0
     ]
    },
    {
     "text": "Claude Code went from about 9.8 consecutive tool calls per transcript to 21.2, and human turns fell from 6.2 to 4.1.",
     "src": [
      0
     ]
    },
    {
     "text": "Employees raised concerns about skill atrophy and reduced mentorship.",
     "src": [
      0
     ]
    },
    {
     "text": "Head of platform engineering Katelyn Lesse said in October 2026 that Claude writes 'literally all' of her team's code, while people still design systems and verifications.",
     "src": [
      2
     ]
    },
    {
     "text": "Superseded (2025 survey): staff now say Claude-written code went from worse than human code in late 2025 to roughly at parity by mid-2026.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "Anthropic: How AI is transforming work at Anthropic",
     "url": "https://www.anthropic.com/research/how-ai-is-transforming-work-at-anthropic",
     "date": "2025-12-02",
     "type": "vendor"
    },
    {
     "title": "Anthropic Institute: When AI builds itself",
     "url": "https://www.anthropic.com/institute/recursive-self-improvement",
     "date": "2026-06-04",
     "type": "vendor"
    },
    {
     "title": "Business Insider: Top Anthropic engineer says Claude now writes 'literally all' her team's code",
     "url": "https://www.businessinsider.com/anthropic-katelyn-lesse-team-shifted-to-ai-coding-entirely-2026-10",
     "date": "2026-10-22",
     "type": "press"
    }
   ],
   "notes": "Vendor-authored. Anthropic studies its own product. The 2025 survey figures are superseded by the June 2026 report.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Dropbox",
   "slug": "dropbox",
   "industry": "SaaS / cloud storage",
   "size": "Under 5k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "AI adoption was made a leadership-aligned engineering priority with a change-management plan. No formal mandate is described. Dropbox said in 2026 that the goal is 'not to force every workflow through an agent'.",
    "src": [
     0,
     2
    ]
   },
   "tools": [
    {
     "name": "GitHub Copilot",
     "src": [
      0
     ]
    },
    {
     "name": "Internal AI tools (code review, migrations, incident debugging)",
     "src": [
      0
     ]
    },
    {
     "name": "Nova (internal coding agent platform)",
     "src": [
      1,
      2
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "Developer experience",
    "AI-authored code",
    "Agent autonomy",
    "Migration throughput",
    "Code review",
    "Quality / defects"
   ],
   "metrics": [
    {
     "text": "Dropbox evaluates AI tools on developer velocity and developer-experience sentiment, and tracks and shares adoption metrics.",
     "src": [
      0
     ]
    },
    {
     "text": "Dropbox uses a four-stage model (Fuel, Adoption, Output, Impact) and tracks review turnaround, first-run test pass rate, defect ratio and rework rate.",
     "src": [
      2
     ]
    }
   ],
   "guardrails": [
    {
     "text": "CTO Ali Dasdan cautions that the volume of AI-generated code does not equal business value.",
     "src": [
      0
     ]
    },
    {
     "text": "Nova runs agents in isolated environments with validation commands, one branch per session and publication kept outside the agent. A human makes the final call before code reaches production.",
     "src": [
      1,
      2
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "More than 90% of Dropbox engineering uses AI tools. Copilot adoption sat in the low double digits from late 2023 through 2024, until leadership alignment and a change-management plan.",
     "src": [
      0
     ]
    },
    {
     "text": "Nova accounts for roughly 1 in 12 pull requests at Dropbox (May 2026). It handles flaky-test fixes validated with 100+ CI runs, migrations and dependency upgrades.",
     "src": [
      1,
      2
     ]
    }
   ],
   "sources": [
    {
     "title": "Dropbox.tech: AI adoption and productivity at Dropbox (CTO Ali Dasdan)",
     "url": "https://dropbox.tech/culture/ai-adoption-productivity-dropbox-cto-ali-dasdan",
     "date": "2025-08-19",
     "type": "company"
    },
    {
     "title": "Dropbox.tech: Introducing Nova, our internal platform for coding agents",
     "url": "https://dropbox.tech/machine-learning/introducing-nova-our-internal-platform-for-coding-agents",
     "date": "2026-05-21",
     "type": "company"
    },
    {
     "title": "Dropbox.tech: Beyond code generation, rethinking engineering productivity in the age of AI agents",
     "url": "https://dropbox.tech/culture/beyond-code-generation-rethinking-engineering-productivity-in-the-age-of-ai-agents",
     "date": "2026-05-28",
     "type": "company"
    }
   ],
   "notes": "",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Mercado Libre",
   "slug": "mercado-libre",
   "industry": "E-commerce / fintech (LatAm)",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "GitHub's customer story describes a broad Copilot rollout. No mandate is described. In 2026 management said most code is AI-written, with no mandate described.",
    "src": [
     0,
     1
    ]
   },
   "tools": [
    {
     "name": "GitHub Copilot",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "Time saved",
    "% code AI-generated"
   ],
   "metrics": [
    {
     "text": "The case study reports time spent writing code and the number of developers using Copilot.",
     "src": [
      0
     ]
    },
    {
     "text": "On the Q2 2026 call management said 20,000 developers use AI and that 'human written code is the exception'; the majority of code is now written by AI.",
     "src": [
      1
     ]
    }
   ],
   "guardrails": [],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "Superseded (2023): GitHub reported 9,000+ Mercado Libre developers using Copilot, with about a 50% reduction in time spent writing code.",
     "src": [
      0
     ]
    },
    {
     "text": "Management said product development costs fell from 8.4% to 7.2% of revenue, and pointed to AI productivity as a factor.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "GitHub customer story: Mercado Libre",
     "url": "https://github.com/customer-stories/mercado-libre",
     "date": "2023-05-23",
     "type": "vendor"
    },
    {
     "title": "Motley Fool: MercadoLibre (MELI) Q2 2026 earnings call transcript",
     "url": "https://www.fool.com/earnings/call-transcripts/2026/08/12/mercadolibre-meli-q2-2026-earnings-call-transcript/",
     "date": "2026-08-12",
     "type": "press"
    }
   ],
   "notes": "The 2023 GitHub case study is superseded by the Q2 2026 call.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Ramp",
   "slug": "ramp",
   "industry": "Fintech / spend management",
   "size": "Under 5k employees",
   "size_src": [],
   "mandate": "no",
   "mandate_summary": {
    "text": "Ramp says it did not force anyone to use its in-house agent, Inspect, and that adoption was driven by building to people's needs and by visible use in shared spaces.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Inspect (internal background agent)",
     "src": [
      0,
      1
     ]
    },
    {
     "name": "OpenCode",
     "src": [
      0
     ]
    },
    {
     "name": "Modal sandboxes",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "PR throughput",
    "Adoption rate"
   ],
   "metrics": [
    {
     "text": "Ramp tracks the share of merged PRs written by Inspect, and sessions that result in a merged PR, which it calls the most important metric.",
     "src": [
      0,
      1
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Each session runs in a sandboxed VM. PRs are opened with the user's own GitHub token so that people cannot approve their own agent-written changes.",
     "src": [
      0
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "Inspect wrote about 30% of merged PRs to Ramp's frontend and backend repos as of January 2026. It wrote about 70% of all merged PRs by late April 2026.",
     "src": [
      0,
      1
     ]
    },
    {
     "text": "More than 130 people at Ramp have contributed to Inspect.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "Ramp Builders: Why we built our own background agent",
     "url": "https://builders.ramp.com/post/why-we-built-our-background-agent",
     "date": "2026-01-12",
     "type": "company"
    },
    {
     "title": "Zach Bruggeman on LinkedIn: When we first shared Inspect... ~70% of all merged PRs",
     "url": "https://www.linkedin.com/posts/zachbruggeman_when-we-first-shared-inspect-ramps-in-house-activity-7455684655460028417-0W8R",
     "date": "2026-04-30",
     "type": "primary"
    }
   ],
   "notes": "Fintech. It is not a bank. The LinkedIn post date was derived from the post's activity ID.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "OpenAI",
   "slug": "openai",
   "industry": "AI lab",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No formal AI-use mandate has been published. OpenAI describes near-total internal use of its own Codex agent, with engineers moving most of their work to it by December 2025.",
    "src": [
     1
    ]
   },
   "tools": [
    {
     "name": "Codex",
     "src": [
      0,
      1,
      2
     ]
    },
    {
     "name": "Bespoke AI code-review model",
     "src": [
      2
     ]
    },
    {
     "name": "AGENTS.md and skills",
     "src": [
      2
     ]
    }
   ],
   "metric_tags": [
    "% code AI-generated",
    "Adoption rate",
    "Agent autonomy",
    "Code review",
    "AI-authored code"
   ],
   "metrics": [
    {
     "text": "Codex accounts for 99.8% of weekly output tokens inside OpenAI; the average engineer now generates 99% of output tokens with Codex, and engineering usage is 27x its November 2025 level.",
     "src": [
      1
     ]
    },
    {
     "text": "Codex writes nearly 80% of OpenAI's code, according to its India strategy head.",
     "src": [
      0
     ]
    },
    {
     "text": "Codex writes more than 90% of the Codex app's own code; engineers typically run 4-8 agents in parallel.",
     "src": [
      2
     ]
    }
   ],
   "guardrails": [
    {
     "text": "A bespoke AI code-review model runs on every PR; about 9 in 10 of its comments are judged valid. Non-critical code can merge after AI review alone, while core agent and open-source code still needs a human reviewer.",
     "src": [
      2
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "The team says the traditional PR flow is 'starting to crack' under the volume of agent-written code.",
     "src": [
      2
     ]
    },
    {
     "text": "Legal, finance and recruiting staff switched to agent-based work around April 2026.",
     "src": [
      1
     ]
    }
   ],
   "results": [
    {
     "text": "The heaviest (99th-percentile) users run more than 60 hours of agent turns per day across parallel agents.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "Outlook Business: OpenAI's strategy head in India reveals Codex writes 80% of company's code",
     "url": "https://www.outlookbusiness.com/deeptech/openais-strategy-head-in-india-reveals-codex-writes-80-of-companys-code",
     "date": "2026-05-15",
     "type": "press"
    },
    {
     "title": "OpenAI: How agents are transforming work",
     "url": "https://openai.com/index/how-agents-are-transforming-work/",
     "date": "2026-06-25",
     "type": "vendor"
    },
    {
     "title": "The Pragmatic Engineer: How Codex is built",
     "url": "https://newsletter.pragmaticengineer.com/p/how-codex-is-built",
     "date": "2026-02-17",
     "type": "press"
    }
   ],
   "notes": "Vendor-authored: OpenAI sells Codex and is reporting on its own use of it. openai.com returns a bot challenge to automated checks but loads for readers. Size band is an editorial estimate.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Apple",
   "slug": "apple",
   "industry": "Big tech",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No AI-use mandate or internal metrics have been published. Evidence of internal use comes from press reports and from Claude Code project files accidentally shipped in an Apple app.",
    "src": [
     1,
     2
    ]
   },
   "tools": [
    {
     "name": "Claude Code",
     "src": [
      1,
      2
     ]
    },
    {
     "name": "Xcode 26.3 agentic coding (Claude Agent and Codex integrations, shipped as a product)",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate"
   ],
   "metrics": [],
   "guardrails": [],
   "hiring_reviews_changes": [
    {
     "text": "A large share of Siri engineers were sent to a multi-week AI coding bootcamp, and some teams put large parts of their budgets toward Claude Code, MacRumors reported, relaying The Information.",
     "src": [
      1
     ]
    }
   ],
   "results": [
    {
     "text": "Apple's Support app version 5.13 shipped with internal CLAUDE.md files (Claude Code project instructions), which were removed in the next update.",
     "src": [
      2
     ]
    }
   ],
   "sources": [
    {
     "title": "Apple Newsroom: Xcode 26.3 unlocks the power of agentic coding",
     "url": "https://www.apple.com/newsroom/2026/02/xcode-26-point-3-unlocks-the-power-of-agentic-coding/",
     "date": "2026-02-03",
     "type": "company"
    },
    {
     "title": "MacRumors: Siri engineers sent to AI coding bootcamp",
     "url": "https://www.macrumors.com/2026/04/15/siri-engineers-ai-coding-bootcamp/",
     "date": "2026-04-15",
     "type": "press"
    },
    {
     "title": "Yahoo Tech: Apple may be using Claude inside the company",
     "url": "https://tech.yahoo.com/ai/claude/articles/apple-using-claude-inside-company-114500152.html",
     "date": "2026-05-01",
     "type": "press"
    }
   ],
   "notes": "Apple has published no internal adoption or productivity figures. The bootcamp report relays The Information, which relies on unnamed sources. Xcode features are products for outside developers, not evidence of internal use on their own.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Netflix",
   "slug": "netflix",
   "industry": "Consumer media",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. Netflix describes a maturity model for moving its engineers up from basic AI assistance to running coding agents, and is hiring for an 'AI Native' agent platform team.",
    "src": [
     0,
     3
    ]
   },
   "tools": [
    {
     "name": "Claude Code",
     "src": [
      0
     ]
    },
    {
     "name": "JetBrains AI coding agents",
     "src": [
      1
     ]
    },
    {
     "name": "Internal agent platform",
     "src": [
      3
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "Self-reported productivity",
    "Code review",
    "Agent autonomy"
   ],
   "metrics": [
    {
     "text": "More than 3,000 Netflix developers use Claude Code, tracked against a maturity ladder, according to a session at Anthropic's Code with Claude event.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [
    {
     "text": "A pre-review agent flags drift from the team's architecture decisions before human reviewers open the diff, and a separate agent updates documentation after merge.",
     "src": [
      2
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "Netflix engineers became 20-30% more efficient with AI coding agents, and up to 65% on some tasks, according to JetBrains.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "Claude (Code with Claude SF): The AI-native engineer, how Netflix moves engineers up the Claude Code maturity ladder",
     "url": "https://claude.com/code-with-claude/session/sf-the-ai-native-engineer-how-netflix-moves-engineers-up-the-claude-code-maturity-ladder",
     "date": "2026-05-06",
     "type": "vendor"
    },
    {
     "title": "JetBrains: How Netflix made AI coding agents more efficient",
     "url": "https://blog.jetbrains.com/ai/2026/10/how-netflix-made-ai-coding-agents-more-efficient/",
     "date": "2026-10-01",
     "type": "vendor"
    },
    {
     "title": "QCon San Francisco 2026: How Netflix retrofitted its SDLC for AI coding agents",
     "url": "https://qconsf.com/presentation/nov2026/how-netflix-retrofitted-its-sdlc-ai-coding-agents",
     "date": "2026-10-01",
     "type": "company"
    },
    {
     "title": "Netflix careers: Software Engineer 5, Agent Platform (AI Platform)",
     "url": "https://netflix.wd108.myworkdayjobs.com/en-US/Netflix/job/Software-Engineer-5---Agent-Platform--AI-Platform_JR41100",
     "date": "2026-10-01",
     "type": "company"
    }
   ],
   "notes": "Both headline figures are vendor-reported (Anthropic and JetBrains). The QCon abstract is a talk by Netflix engineers. Dates for the JetBrains post, the QCon abstract and the job post are month-level.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Oracle",
   "slug": "oracle",
   "industry": "Big tech",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No AI-use mandate reported. Oracle leadership says AI lets it build software with smaller teams, while Oracle-led OpenJDK bans LLM-written contributions.",
    "src": [
     1,
     0
    ]
   },
   "tools": [],
   "metric_tags": [],
   "metrics": [],
   "guardrails": [
    {
     "text": "Oracle's OpenJDK project bans contributions written by large language models, even as Larry Ellison promotes AI-written code elsewhere in the company.",
     "src": [
      0
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Oracle cut about 21,000 roles; its annual report links the restructuring to AI.",
     "src": [
      2
     ]
    },
    {
     "text": "Oracle said it will need fewer developers as it brings in AI coding tools, with smaller teams building software.",
     "src": [
      1
     ]
    }
   ],
   "results": [],
   "sources": [
    {
     "title": "The Register: As Larry Ellison bets the farm, Oracle says it loves AI-written code, just not in OpenJDK",
     "url": "https://www.theregister.com/ai-and-ml/2026/08/03/as-larry-ellison-bets-the-farm-oracle-says-it-loves-ai-written-code-just-not-in-openjdk/5281851",
     "date": "2026-08-03",
     "type": "press"
    },
    {
     "title": "CIO: Oracle to shed developers as it brings in AI tools",
     "url": "https://www.cio.com/article/4144164/oracle-to-shed-developers-as-it-brings-in-ai-tools.html",
     "date": "2026-03-12",
     "type": "press"
    },
    {
     "title": "BBC: Oracle job cuts",
     "url": "https://www.bbc.com/news/articles/c4gy0x0j5deo",
     "date": "2026-06-23",
     "type": "press"
    }
   ],
   "notes": "Oracle has not published internal adoption or code-share figures.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "IBM",
   "slug": "ibm",
   "industry": "Big tech",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No AI-use mandate reported. IBM says more than 80,000 employees use its own Bob development tool, and its CEO says agents now write about half of IBM's code.",
    "src": [
     0,
     2
    ]
   },
   "tools": [
    {
     "name": "IBM Bob (IBM's own AI development partner; uses Claude, Mistral and Granite models)",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "Self-reported productivity",
    "% code AI-generated",
    "Time saved"
   ],
   "metrics": [
    {
     "text": "IBM Bob is used by more than 80,000 IBM employees, who self-report an average 45% productivity gain.",
     "src": [
      0
     ]
    },
    {
     "text": "CEO Arvind Krishna: 'We use agents and coding agents now for probably half the code that is written inside our place. And it is superior to what was written before.'",
     "src": [
      2
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Bob includes approval checkpoints, sensitive-data scanning and an audit trail.",
     "src": [
      0
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "IBM is tripling US entry-level hiring in 2026 and rewriting junior roles around AI.",
     "src": [
      1
     ]
    }
   ],
   "results": [
    {
     "text": "The Instana team reports a 70% gain (about 10 hours a week per developer), and the Maximo team 69%.",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "IBM Newsroom: Introducing IBM Bob",
     "url": "https://newsroom.ibm.com/2026-04-28-introducing-ibm-bob-ai-development-partner-that-takes-enterprises-from-ai-assisted-coding-to-production-ready-software",
     "date": "2026-04-28",
     "type": "vendor"
    },
    {
     "title": "Fortune: IBM tripling Gen Z entry-level hiring",
     "url": "https://fortune.com/2026/02/13/tech-giant-ibm-tripling-gen-z-entry-level-hiring-according-to-chro-rewriting-jobs-ai-era/",
     "date": "2026-02-13",
     "type": "press"
    },
    {
     "title": "All-In podcast transcript: Arvind Krishna and Nikesh Arora",
     "url": "https://smashnotes.com/p/all-in/e/arvind-krishna-and-nikesh-arora-quantum-threat-open-vs-frontier-ai-ai-cyber-warfare-and-saas-collapse/transcript",
     "date": "2026-10-21",
     "type": "primary"
    }
   ],
   "notes": "Bob figures are vendor-reported: IBM sells Bob and is reporting on its own internal use. The 'half the code' figure is the CEO's spoken estimate.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Adobe",
   "slug": "adobe",
   "industry": "Design / SaaS",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No formal mandate reported. Adobe is training about 10,000 engineers and other builders on AI coding agents, with a stated goal of 10x productivity.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Claude Code",
     "src": [
      0
     ]
    },
    {
     "name": "Codex",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "Agent autonomy"
   ],
   "metrics": [
    {
     "text": "About 10,000 builders are being trained on Claude Code and Codex, across more than 170 pilots and 'AI factories'.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [],
   "hiring_reviews_changes": [
    {
     "text": "Engineering headcount is being held flat while output targets rise.",
     "src": [
      0
     ]
    }
   ],
   "results": [],
   "sources": [
    {
     "title": "Time Under Tension: How Adobe are teaching their 10,000 builders",
     "url": "https://timeundertension.substack.com/p/how-adobe-are-teaching-their-10000",
     "date": "2026-05-06",
     "type": "press"
    }
   ],
   "notes": "Single source, written by a guest of Adobe based on access Adobe provided.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Tencent",
   "slug": "tencent",
   "industry": "Big tech (China)",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. Tencent says more than 12,000 of its engineers use its own CodeBuddy assistant.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "CodeBuddy (Tencent's own coding assistant)",
     "src": [
      0,
      1
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "% code AI-generated",
    "Time saved"
   ],
   "metrics": [
    {
     "text": "CodeBuddy is used by more than 12,000 Tencent engineers and helps generate over half of new code.",
     "src": [
      0,
      1
     ]
    }
   ],
   "guardrails": [],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "Coding time has been cut by more than 40%, Tencent says.",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "PR Newswire: Tencent Cloud at Davos",
     "url": "https://www.prnewswire.com/news-releases/tencent-cloud-at-davos-unlocking-ai-for-all-bringing-value-within-reach-302667443.html",
     "date": "2026-01-21",
     "type": "vendor"
    },
    {
     "title": "Sina Finance: Tencent CodeBuddy internal use",
     "url": "https://finance.sina.com.cn/roll/2026-01-23/doc-inhihsry1008167.shtml",
     "date": "2026-01-23",
     "type": "press"
    }
   ],
   "notes": "Vendor-reported: Tencent sells CodeBuddy and is describing its own internal use.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "ByteDance",
   "slug": "bytedance",
   "industry": "Big tech (China)",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. A ByteDance VP says AI's share of the company's code grew more than 6x in a year, but warns against chasing that single metric.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "TRAE (ByteDance's own coding agent)",
     "src": [
      0,
      1
     ]
    }
   ],
   "metric_tags": [
    "% code AI-generated",
    "Adoption rate",
    "Quality / defects",
    "PR throughput"
   ],
   "metrics": [
    {
     "text": "Over the past year, AI's code contribution rate grew more than 6x, AI-coding token use 5x, and the AI-code merge rate more than 2x.",
     "src": [
      0
     ]
    },
    {
     "text": "TRAE is used by 92% of ByteDance engineers.",
     "src": [
      1
     ]
    }
   ],
   "guardrails": [
    {
     "text": "ByteDance found AI-generated output scored only 40-60 out of 100 on usability, reliability and maintainability; adding harness work (context, architecture constraints, team knowledge) raised correctness from about 80% to nearly 90% and deliverability to about 80.",
     "src": [
      0
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "The TRAE team had more than 90% of its code written by AI over six months, but per-person throughput rose only 1.6x, which ByteDance cites as evidence that code share is a misleading metric.",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "163.com (IT Times): Hong Dingkun on the challenges behind ByteDance's 6x growth in AI code contribution",
     "url": "https://www.163.com/dy/article/L07IQN5P05527PMU.html",
     "date": "2026-06-24",
     "type": "press"
    },
    {
     "title": "Longbridge: TRAE used by 92% of ByteDance engineers",
     "url": "https://longbridge.com/en/news/270106977",
     "date": "2025-12-18",
     "type": "press"
    }
   ],
   "notes": "ByteDance sells TRAE, so its internal figures are self-reported by the vendor. The 2026 figures come from a talk by VP of Technology Hong Dingkun.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Alibaba",
   "slug": "alibaba",
   "industry": "Big tech (China)",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No company-wide mandate reported. Alibaba's Amap team describes moving from 'vibe coding' to spec-driven development with harness constraints after a high AI code share did not translate into obvious efficiency gains.",
    "src": [
     0
    ]
   },
   "tools": [],
   "metric_tags": [
    "% code AI-generated",
    "Quality / defects"
   ],
   "metrics": [
    {
     "text": "The Amap team's AI-generated code share rose from 53% to 80-90% or more.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [
    {
     "text": "The team adopted spec-driven development (SDD) and harness governance to constrain AI-written code.",
     "src": [
      0
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "Despite the higher code share, 'efficiency gains were not obvious' until the team changed its process.",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "Alibaba Cloud blog: Farewell to vibe coding, team-level AI development powered by harness governance and SDD",
     "url": "https://www.alibabacloud.com/blog/farewell-to-vibe-coding-the-evolution-and-practice-of-team-level-ai-development-powered-by-harness-governance-and-sdd_603448",
     "date": "2026-08-12",
     "type": "company"
    }
   ],
   "notes": "Covers one team (Amap), not all of Alibaba.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Baidu",
   "slug": "baidu",
   "industry": "Big tech (China)",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. Press reports citing an unnamed source say AI-generated code overtook human-written code inside Baidu by the end of 2025.",
    "src": [
     1
    ]
   },
   "tools": [
    {
     "name": "Comate (Baidu's coding assistant, merged into DuMate in 2026)",
     "src": [
      1
     ]
    },
    {
     "name": "Miaoda",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "% code AI-generated",
    "Adoption rate"
   ],
   "metrics": [
    {
     "text": "By the end of 2025, AI-generated code exceeded human-written code inside Baidu, and AI coding supports more than 99% of engineers' daily work, Guancha reported, citing an unnamed source.",
     "src": [
      1
     ]
    },
    {
     "text": "CEO Robin Li said Baidu's Miaoda app builder generates about 90% of its own code.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [],
   "hiring_reviews_changes": [],
   "results": [],
   "sources": [
    {
     "title": "TechNode: Baidu Create 2026, CEO says AI is moving to the agent era",
     "url": "https://technode.com/2026/05/14/baidu-create-2026-ceo-says-ai-is-moving-from-model-competition-to-ai-agent-era-foresees-rise-of-super-individuals/",
     "date": "2026-05-14",
     "type": "press"
    },
    {
     "title": "Guancha: Baidu Comate merged into DuMate",
     "url": "https://www.guancha.cn/economy/2026_09_08_830369.shtml",
     "date": "2026-09-08",
     "type": "press"
    }
   ],
   "notes": "The company-wide code-share figure rests on an unnamed source. Baidu sells Comate and Miaoda.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "AMD",
   "slug": "amd",
   "industry": "Semiconductors / AI",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported, but AMD set an internal productivity target for AI and says the goal is not to reduce headcount.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Claude (rolled out broadly)",
     "src": [
      1
     ]
    },
    {
     "name": "Agent swarms for software and hardware workflows",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Self-reported productivity",
    "% code AI-generated",
    "Quality / defects"
   ],
   "metrics": [
    {
     "text": "AMD measured a 30% productivity gain against a 25% target.",
     "src": [
      0
     ]
    },
    {
     "text": "AI-generated code was about 20% of shipped code at the start of 2026 and heading toward 50%.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [],
   "hiring_reviews_changes": [
    {
     "text": "AMD says the aim is 'not to reduce headcount'.",
     "src": [
      0
     ]
    }
   ],
   "results": [
    {
     "text": "An automated fix agent's success rate rose from 6% to more than 75%.",
     "src": [
      0
     ]
    },
    {
     "text": "AMD set a goal of making engineers 10x more productive.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "IEEE Spectrum: AMD agent swarms",
     "url": "https://spectrum.ieee.org/amd-agent-swarms",
     "date": "2026-08-17",
     "type": "press"
    },
    {
     "title": "AMD blog: Becoming a 10x productivity engineer",
     "url": "https://www.amd.com/en/blogs/2026/becoming-a-10x-productivity-engineer.html",
     "date": "2026-08-13",
     "type": "company"
    }
   ],
   "notes": "",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Intel",
   "slug": "intel",
   "industry": "Semiconductors / AI",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate or internal metrics reported. Intel is rolling out Google's Gemini Enterprise, including agentic coding assistance, as part of an internal AI transformation.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Gemini Enterprise with agentic coding assistance",
     "src": [
      0
     ]
    },
    {
     "name": "GitHub Copilot, Claude Code and Cursor (listed as preferred skills in an Intel job post)",
     "src": [
      1
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate"
   ],
   "metrics": [],
   "guardrails": [],
   "hiring_reviews_changes": [
    {
     "text": "An Intel chip-design job post lists experience with GitHub Copilot, Claude Code and Cursor as preferred.",
     "src": [
      1
     ]
    }
   ],
   "results": [],
   "sources": [
    {
     "title": "Intel Newsroom: Intel and Google Cloud announce collaboration to accelerate Intel's AI-enabled enterprise transformation",
     "url": "https://newsroom.intel.com/artificial-intelligence/intel-google-cloud-announce-collaboration-to-accelerate-intel-ai-enabled-enterprise-transformation",
     "date": "2026-07-16",
     "type": "company"
    },
    {
     "title": "Intel careers: AI/ML CAD Engineer (RTL / Design Verification), JR0287395",
     "url": "https://intel.wd1.myworkdayjobs.com/en-US/External/job/XMLNAME---AI-ML-CAD-Engineer--RTL---Design-Verification_JR0287395",
     "date": "2026-10-01",
     "type": "company"
    }
   ],
   "notes": "Thin record: Intel has published no internal adoption or productivity figures. Job post date is month-level.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Cisco",
   "slug": "cisco",
   "industry": "Networking / security",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No formal mandate reported. Cisco's product chief says the company's first product is now 100% AI-written and expects 60-70% of the portfolio to be built that way by 2027, and has urged engineers to upskill.",
    "src": [
     1,
     2
    ]
   },
   "tools": [
    {
     "name": "Codex",
     "src": [
      1
     ]
    }
   ],
   "metric_tags": [
    "AI-authored code",
    "Adoption rate",
    "Code review"
   ],
   "metrics": [
    {
     "text": "Cisco AI Defense is the company's first product written 100% with AI; about six products are expected to follow by the end of 2026, and 70% of the portfolio by 2027.",
     "src": [
      0,
      1
     ]
    },
    {
     "text": "85% of Cisco engineers use AI.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Code reviews are still done by humans, and Cisco requires 'provenance of a human' on AI-written code.",
     "src": [
      0,
      1
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Jeetu Patel said engineers spend only 30-35% of their time actually writing code, and urged employees to upskill to stay relevant.",
     "src": [
      1,
      2
     ]
    }
   ],
   "results": [],
   "sources": [
    {
     "title": "SDxCentral: Cisco CPO predicts AI will have built majority of the vendor's products by end of 2027",
     "url": "https://www.sdxcentral.com/analysis/cisco-cpo-predicts-ai-will-have-built-majority-of-the-vendors-products-by-end-of-2027/",
     "date": "2026-05-06",
     "type": "press"
    },
    {
     "title": "Cisco Newsroom: AI is writing the code now, Cisco's vision for the agent era (Six Five On the Road)",
     "url": "https://newsroom.cisco.com/c/r/newsroom/en/us/a/y2026/m03/ai-is-writing-the-code-now-cisco-vision-for-the-agent-era-six-five-on-the-road.html",
     "date": "2026-03-01",
     "type": "primary"
    },
    {
     "title": "India Today: Jeetu Patel says AI will build 70% of Cisco products by 2027",
     "url": "https://www.indiatoday.in/technology/news/story/jeetu-patel-says-ai-will-build-70-percent-of-cisco-products-by-2027-employees-must-upskill-to-stay-relevant-2873539-2026-02-24",
     "date": "2026-02-24",
     "type": "press"
    }
   ],
   "notes": "Cisco newsroom date is month-level (March 2026).",
   "last_verified": "2026-10-25"
  },
  {
   "name": "SAP",
   "slug": "sap",
   "industry": "Enterprise SaaS",
   "size": "50k+ employees",
   "size_src": [
    0
   ],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No AI-use mandate reported. SAP's CEO attributes a roughly 70% rise in developer code contributions this year to its AI-first principles.",
    "src": [
     0
    ]
   },
   "tools": [],
   "metric_tags": [
    "PR throughput",
    "Self-reported productivity"
   ],
   "metrics": [
    {
     "text": "SAP developers completed approximately 70% more code contributions since the start of 2026, which CEO Christian Klein attributes to AI-first working principles.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "SAP reports about 20% efficiency gains in central functions from AI.",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "Motley Fool: SAP Q3 2026 earnings call transcript",
     "url": "https://www.fool.com/earnings/call-transcripts/2026/10/21/sap-sap-q3-2026-earnings-call-transcript/",
     "date": "2026-10-22",
     "type": "primary"
    }
   ],
   "notes": "",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Samsung Electronics",
   "slug": "samsung",
   "industry": "Electronics / semiconductors",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. More than 10,000 developers in Samsung's chip (DS) division use Claude Code through a company-managed gateway, according to AWS.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Claude Code",
     "src": [
      0,
      1
     ]
    },
    {
     "name": "Internal agent gateway",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "Time saved"
   ],
   "metrics": [
    {
     "text": "More than 10,000 DS-division developers use Claude Code, starting from a pilot, an AWS speaker said at Samsung AI Forum 2026.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [
    {
     "text": "An agent gateway in front of the coding agents issues short-lived keys, checks each team's remaining budget, sets per-team call limits and allowed models, switches to smaller models near budget limits, and logs token use and cost.",
     "src": [
      0
     ]
    },
    {
     "text": "Engineers must still review outputs because the AI makes mistakes in hardware design work.",
     "src": [
      1
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "In the System LSI business, some tasks such as SoC verification and USB driver and emulator work went from weeks to days.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "Digital Daily: 10,000 Samsung DS developers use Claude Code, with permissions and costs managed via AWS",
     "url": "https://www.ddaily.co.kr/page/view/2026093016340947942",
     "date": "2026-09-30",
     "type": "press"
    },
    {
     "title": "SamMobile: Samsung's chip division using Claude AI to speed up development",
     "url": "https://www.sammobile.com/news/samsungs-chip-division-using-claude-ai-speed-up-development/",
     "date": "2026-08-12",
     "type": "press"
    }
   ],
   "notes": "The 10,000-developer figure was given by an AWS speaker (a vendor), relayed by Digital Daily.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "LinkedIn",
   "slug": "linkedin",
   "industry": "Social / professional network",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. LinkedIn (owned by Microsoft) has published how it gives coding agents organizational context and runs AI code review at scale.",
    "src": [
     0,
     1
    ]
   },
   "tools": [
    {
     "name": "Contextual agent playbooks and tools (CAPT, MCP-based)",
     "src": [
      0
     ]
    },
    {
     "name": "Multi-agent AI code reviewer",
     "src": [
      1
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "Code review",
    "Time saved"
   ],
   "metrics": [
    {
     "text": "More than 1,000 engineers use CAPT, with more than 500 playbooks.",
     "src": [
      0
     ]
    },
    {
     "text": "The AI code reviewer covers about 10,000 repositories and about 79,000 reviews a week, with a 63.9% combined acceptance rate on sampled reviews.",
     "src": [
      1
     ]
    }
   ],
   "guardrails": [
    {
     "text": "The AI reviewer gives a first pass before human review; most AI reviews land before the first human reviewer.",
     "src": [
      0,
      1
     ]
    },
    {
     "text": "LinkedIn rejected off-the-shelf single-model reviewers in favour of a multi-agent approach tuned to each codebase.",
     "src": [
      1
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "P90 time to first human review rose in Q4 2025, one reason LinkedIn invested in AI review.",
     "src": [
      1
     ]
    }
   ],
   "results": [
    {
     "text": "Triage time fell about 70% and data analysis became about 3x faster with CAPT.",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "LinkedIn Engineering: Contextual agent playbooks and tools, how LinkedIn gave AI coding agents organizational context",
     "url": "https://www.linkedin.com/blog/engineering/ai/contextual-agent-playbooks-and-tools-how-linkedin-gave-ai-coding-agents-organizational-context",
     "date": "2026-01-27",
     "type": "company"
    },
    {
     "title": "LinkedIn Engineering: High-signal AI code review that adapts to your codebase at scale",
     "url": "https://www.linkedin.com/blog/engineering/ai/high-signal-ai-code-review-that-adapts-to-your-codebase-at-scale",
     "date": "2026-08-13",
     "type": "company"
    }
   ],
   "notes": "LinkedIn is a Microsoft subsidiary with its own engineering org.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "GitHub",
   "slug": "github",
   "industry": "Developer tools",
   "size": "Under 5k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. GitHub (part of Microsoft) builds github.com with its own Copilot agents, and its COO describes daily internal use of agent workflows.",
    "src": [
     0,
     2
    ]
   },
   "tools": [
    {
     "name": "GitHub Copilot coding agent, app and CLI (GitHub's own products)",
     "src": [
      0,
      2
     ]
    },
    {
     "name": "Copilot cloud agents and skills",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Agent autonomy",
    "Adoption rate",
    "Quality / defects"
   ],
   "metrics": [],
   "guardrails": [
    {
     "text": "Copilot's coding agent is assigned issues in the core github.com repository; GitHub says its PRs merge at a lower rate than human PRs and that it targets the 'tedious 80%' of work.",
     "src": [
      2
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "Agent-generated load on the platform forced GitHub to raise its capacity planning from 10x to 30x.",
     "src": [
      1
     ]
    }
   ],
   "results": [
    {
     "text": "GitHub uptime fell below 85% in April 2026, and a merge-queue bug reverted code in 658 repositories, as agent traffic surged.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "Latent Space: GitHub (interview with COO Kyle Daigle)",
     "url": "https://www.latent.space/p/github",
     "date": "2026-06-02",
     "type": "primary"
    },
    {
     "title": "Times of India: GitHub COO on the AI code overload degrading the platform",
     "url": "https://timesofindia.indiatimes.com/technology/tech-news/github-coo-on-ai-shit-code-overload-that-many-users-say-has-been-degrading-the-coding-platform-i-am-not/articleshow/130819672.cms",
     "date": "2026-05-05",
     "type": "press"
    },
    {
     "title": "GitHub blog: How Copilot helps build the GitHub platform",
     "url": "https://github.blog/ai-and-ml/github-copilot/how-copilot-helps-build-the-github-platform/",
     "date": "2025-11-12",
     "type": "vendor"
    }
   ],
   "notes": "Vendor-authored: GitHub sells Copilot and is describing its own use of it. GitHub is listed separately from its parent, Microsoft. On Latent Space, a figure of 80% of PRs coming from agents was offered as a hypothetical, not a measured metric, so it is not shown here.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Snowflake",
   "slug": "snowflake",
   "industry": "Data / cloud platform",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. On its Q1 FY27 earnings call, Snowflake said internal support, SRE and services teams have more than 95% adoption of its own Cortex Code agent.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Cortex Code / CoCo (Snowflake's own coding agent)",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "PR throughput",
    "Time saved"
   ],
   "metrics": [
    {
     "text": "Internal support, SRE and services teams have more than 95% adoption of Cortex Code.",
     "src": [
      0
     ]
    },
    {
     "text": "Snowflake's data organization doubled developer productivity, measured by PRs and lines of code per engineer.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [],
   "hiring_reviews_changes": [
    {
     "text": "Snowflake says productivity gains show up partly as slower hiring.",
     "src": [
      0
     ]
    }
   ],
   "results": [
    {
     "text": "Resolution time for complex support cases fell nearly 30%, and engineering time per ticket about 40%.",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "Snowflake Q1 FY2027 earnings call transcript (StockAnalysis)",
     "url": "https://stockanalysis.com/stocks/snow/transcripts/568996-q1-2027/",
     "date": "2026-05-27",
     "type": "primary"
    },
    {
     "title": "Motley Fool: Snowflake Q1 2027 earnings call transcript",
     "url": "https://www.fool.com/earnings/call-transcripts/2026/05/27/snowflake-snow-q1-2027-earnings-transcript/",
     "date": "2026-05-27",
     "type": "primary"
    }
   ],
   "notes": "Snowflake sells Cortex Code, so these are a vendor's figures on its own use. Lines of code is a weak productivity measure.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Databricks",
   "slug": "databricks",
   "industry": "Data / cloud platform",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. Databricks built an internal benchmark from its own merged PRs to choose which coding agents and models its engineers use day to day.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Coding agents benchmarked on an internal codebase (GLM 5.2 as a daily driver; Claude Opus)",
     "src": [
      0
     ]
    },
    {
     "name": "coSTAR agent testing framework",
     "src": [
      1
     ]
    }
   ],
   "metric_tags": [
    "Cost per PR",
    "Quality / defects",
    "Cycle time"
   ],
   "metrics": [
    {
     "text": "On an internal benchmark built from merged PRs in its multi-million-line codebase, GLM 5.2 tied Claude Opus 4.8 at $1.28 versus $1.94 per task, and Databricks deployed it as a daily driver.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [
    {
     "text": "The coSTAR framework tests agents against scenarios before shipping, so changes can ship fast without breaking things.",
     "src": [
      1
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "The agent harness mattered as much as the model: one harness (Pi) was about 2x cheaper for similar results.",
     "src": [
      0
     ]
    },
    {
     "text": "Agent verification time fell from about two weeks to hours with coSTAR.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "Databricks blog: Benchmarking coding agents on a multi-million-line codebase",
     "url": "https://www.databricks.com/blog/benchmarking-coding-agents-databricks-multi-million-line-codebase",
     "date": "2026-07-08",
     "type": "company"
    },
    {
     "title": "Databricks blog: coSTAR, how we ship AI agents at Databricks fast without breaking things",
     "url": "https://www.databricks.com/blog/costar-how-we-ship-ai-agents-databricks-fast-without-breaking-things",
     "date": "2026-03-20",
     "type": "company"
    }
   ],
   "notes": "",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Datadog",
   "slug": "datadog",
   "industry": "Observability / SaaS",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No formal mandate reported, though Anthropic says all Datadog engineers use AI tools for production code.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Claude Code",
     "src": [
      0
     ]
    },
    {
     "name": "Codex",
     "src": [
      1
     ]
    },
    {
     "name": "Bits AI / BitsEvolve",
     "src": [
      2
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "Code review",
    "Quality / defects"
   ],
   "metrics": [
    {
     "text": "All Datadog engineers use AI tools for production code, and Claude Code drives at least two-thirds of that use, according to Anthropic.",
     "src": [
      0
     ]
    },
    {
     "text": "Codex reviews every PR in one major repository, used by more than 1,000 engineers, according to OpenAI.",
     "src": [
      1
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Datadog replayed past incidents and found Codex review feedback would have helped in about 22% of them.",
     "src": [
      1
     ]
    },
    {
     "text": "Datadog follows 'harness-first engineering': instead of reading every line of agent-generated code, it invests in automated checks and production telemetry that verify the code is correct.",
     "src": [
      2
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "An internal tool (Helix) was built in days with Claude Code.",
     "src": [
      0
     ]
    },
    {
     "text": "BitsEvolve, an LLM-guided optimizer, shipped 10x speedups on key ingestion functions, verified against live traffic.",
     "src": [
      2
     ]
    }
   ],
   "sources": [
    {
     "title": "Claude: How Datadog built a universal machine tool for Claude Code",
     "url": "https://claude.com/resources/articles/how-datadog-built-a-universal-machine-tool-for-claude-code",
     "date": "2026-07-21",
     "type": "vendor"
    },
    {
     "title": "OpenAI: Datadog customer story",
     "url": "https://openai.com/index/datadog/",
     "date": "2026-01-09",
     "type": "vendor"
    },
    {
     "title": "Datadog blog: Harness-first agents",
     "url": "https://www.datadoghq.com/blog/ai/harness-first-agents/",
     "date": "2026-03-09",
     "type": "company"
    }
   ],
   "notes": "Adoption and review figures are vendor-reported (Anthropic and OpenAI customer stories). openai.com returns a bot challenge to automated checks but loads for readers.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Cloudflare",
   "slug": "cloudflare",
   "industry": "Cloud / networking",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. Cloudflare published detailed internal adoption data: 93% of R&D used AI coding tools, routed through its own AI Gateway.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "OpenCode",
     "src": [
      0,
      1
     ]
    },
    {
     "name": "Windsurf",
     "src": [
      0
     ]
    },
    {
     "name": "AI Gateway (Cloudflare's own)",
     "src": [
      0
     ]
    },
    {
     "name": "AGENTS.md across about 3,900 repos",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "PR throughput",
    "Code review",
    "Cost per PR"
   ],
   "metrics": [
    {
     "text": "93% of Cloudflare R&D used AI coding tools, and 3,683 users (about 60% of the company) used them overall; 241.37 billion tokens went through AI Gateway.",
     "src": [
      0
     ]
    },
    {
     "text": "Weekly merge requests rose from about 5,600 to more than 8,700, peaking at 10,952 in the week of March 23.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [
    {
     "text": "An AI reviewer (a coordinator agent over specialist agents) runs on every merge request and can block merges for serious problems; a human can comment 'break glass' to force approval, which happened 288 times (0.6% of merge requests).",
     "src": [
      1
     ]
    },
    {
     "text": "Cloudflare says the AI reviewer 'isn't a replacement for human code review, at least not yet'.",
     "src": [
      1
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "The average AI review costs $1.19 (median $0.98, P99 $4.45) and completes in a median of about 3 minutes 39 seconds.",
     "src": [
      1
     ]
    }
   ],
   "sources": [
    {
     "title": "Cloudflare blog: Our internal AI engineering stack",
     "url": "https://blog.cloudflare.com/internal-ai-engineering-stack/",
     "date": "2026-04-20",
     "type": "company"
    },
    {
     "title": "Cloudflare blog: AI code review",
     "url": "https://blog.cloudflare.com/ai-code-review/",
     "date": "2026-04-20",
     "type": "company"
    }
   ],
   "notes": "Cloudflare sells AI Gateway and agent infrastructure, so this is partly a showcase of its own products.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Twilio",
   "slug": "twilio",
   "industry": "Communications APIs",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. Twilio's CEO says employees use Gemini and Claude Code internally and have seen a 15% productivity boost.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Gemini",
     "src": [
      0
     ]
    },
    {
     "name": "Claude Code",
     "src": [
      0,
      2
     ]
    }
   ],
   "metric_tags": [
    "Self-reported productivity",
    "Time saved"
   ],
   "metrics": [
    {
     "text": "CEO Khozema Shipchandler said employees achieved a 15% productivity boost with AI, used for coding, support and sales.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [
    {
     "text": "An automated pipeline rewrote 1,138 error-code documentation pages in about 12 weeks; a human reviewed every change.",
     "src": [
      1
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "The docs generation pass took about 11 hours (about 35 seconds a page), versus an estimated two days per page by hand.",
     "src": [
      1
     ]
    },
    {
     "text": "A product manager built more than 1,000 commits in 90 days with Claude Code, including 425 MCP tools, according to Anthropic.",
     "src": [
      2
     ]
    }
   ],
   "sources": [
    {
     "title": "Business Insider: Twilio stock soars on CEO's AI strategy",
     "url": "https://www.businessinsider.com/twilio-stock-soars-ceo-ai-strategy-2026-5",
     "date": "2026-05-01",
     "type": "press"
    },
    {
     "title": "Twilio blog: Automated pipeline to revise error-code docs",
     "url": "https://www.twilio.com/en-us/blog/developers/automated-pipeline-revise-docs-error-codes",
     "date": "2026-09-17",
     "type": "company"
    },
    {
     "title": "Claude customer story: Twilio",
     "url": "https://claude.com/customers/twilio-qa",
     "date": "2026-01-01",
     "type": "vendor"
    }
   ],
   "notes": "The Claude customer story is vendor-authored and undated (2026 copyright); its date is year-level.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Workday",
   "slug": "workday",
   "industry": "Enterprise SaaS",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. On its Q4 FY26 earnings call, Workday said more than 75% of its software engineers use AI coding assistance.",
    "src": [
     0
    ]
   },
   "tools": [],
   "metric_tags": [
    "Adoption rate",
    "% code AI-generated",
    "PR throughput"
   ],
   "metrics": [
    {
     "text": "'More than 75% of our software engineers are using AI coding assistance and more than 50% of the committed code is AI generated.'",
     "src": [
      0
     ]
    },
    {
     "text": "Engineering output grew 22% over six months, measured by the number of code updates.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "Workday says some key API development became about 30x faster.",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "Motley Fool: Workday Q4 2026 earnings call transcript",
     "url": "https://www.fool.com/earnings/call-transcripts/2026/05/21/workday-wday-q4-2026-earnings-transcript/",
     "date": "2026-05-21",
     "type": "primary"
    },
    {
     "title": "StockAnalysis: Workday Q4 2026 earnings call transcript",
     "url": "https://stockanalysis.com/stocks/wday/transcripts/411056-q4-2026/",
     "date": "2026-05-21",
     "type": "primary"
    }
   ],
   "notes": "The call took place on 2026-02-24; the source date is the transcript's publication date.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "ServiceNow",
   "slug": "servicenow",
   "industry": "Enterprise SaaS",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. ServiceNow says it has deployed Claude to more than 29,000 employees and made Claude Code available for engineering work.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Claude Code",
     "src": [
      0
     ]
    },
    {
     "name": "Claude",
     "src": [
      0
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate"
   ],
   "metrics": [
    {
     "text": "ServiceNow deployed Claude to more than 29,000 employees and made Claude Code available to engineers.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [],
   "hiring_reviews_changes": [],
   "results": [
    {
     "text": "Sales preparation time fell by up to 95% in early results (a non-engineering use).",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "ServiceNow Newsroom: ServiceNow and Anthropic partner to help customers build AI-powered applications",
     "url": "https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-and-Anthropic-partner-to-help-customers-build-AI-powered-applications-accelerate-time-to-value-and-apply-trusted-AI-to-critical-industries/default.aspx",
     "date": "2026-01-28",
     "type": "vendor"
    },
    {
     "title": "Business Insider: ServiceNow AI use case in product testing",
     "url": "https://www.businessinsider.com/service-now-ai-use-case-product-testing-2026-03",
     "date": "2026-03-18",
     "type": "press"
    }
   ],
   "notes": "The 29,000 figure comes from a joint partnership announcement with Anthropic, so it is treated as vendor-sourced. No engineering-specific metrics were published.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Tesla",
   "slug": "tesla",
   "industry": "Automotive / energy",
   "size": "50k+ employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No AI-use mandate reported. Tesla caps employee AI spending at $200 a week (xAI products exempt), and a memo told staff to switch to Grok where possible because of cost.",
    "src": [
     0,
     1
    ]
   },
   "tools": [
    {
     "name": "Bottle Rocket (internal AI platform)",
     "src": [
      0
     ]
    },
    {
     "name": "Grok (xAI)",
     "src": [
      1
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "Cost savings"
   ],
   "metrics": [
    {
     "text": "Tesla tracks AI token use on internal dashboards.",
     "src": [
      0
     ]
    }
   ],
   "guardrails": [
    {
     "text": "From July 6, 2026, employee AI spending is capped at $200 a week; xAI products are exempt.",
     "src": [
      0,
      2
     ]
    }
   ],
   "hiring_reviews_changes": [
    {
     "text": "An Elon Musk memo told Tesla staff to switch to Grok 'when possible' to cut costs.",
     "src": [
      1
     ]
    }
   ],
   "results": [],
   "sources": [
    {
     "title": "Electrek: Tesla caps employee AI spending at $200 a week",
     "url": "https://electrek.co/2026/07/02/tesla-caps-employee-ai-spending-200-week/",
     "date": "2026-07-02",
     "type": "press"
    },
    {
     "title": "Electrek: Musk tells Tesla staff to switch to Grok",
     "url": "https://electrek.co/2026/07/10/musk-tells-tesla-staff-switch-grok/",
     "date": "2026-07-10",
     "type": "press"
    },
    {
     "title": "Tech Times: Tesla limits AI tool spending to $200 weekly while Grok stays exempt",
     "url": "https://www.techtimes.com/articles/319710/20260704/tesla-limits-ai-tool-spending-200-weekly-while-musks-grok-stays-exempt.htm",
     "date": "2026-07-04",
     "type": "press"
    }
   ],
   "notes": "Tesla has not published adoption or productivity figures. A widely shared 'dark factory' podcast transcript attributed to Tesla actually features a different company (Tessl) and is not used here.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Pinterest",
   "slug": "pinterest",
   "industry": "Social media",
   "size": "Under 5k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. On its Q2 2026 call, Pinterest said weekly PRs per engineer rose 45% year over year with incident rates roughly flat.",
    "src": [
     0
    ]
   },
   "tools": [
    {
     "name": "Internal MCP servers and agents",
     "src": [
      1
     ]
    }
   ],
   "metric_tags": [
    "PR throughput",
    "Quality / defects",
    "Time saved",
    "Cost savings"
   ],
   "metrics": [
    {
     "text": "Weekly PRs per engineer were up 45% year over year in July 2026, while incident rates and uptime stayed about flat.",
     "src": [
      0
     ]
    },
    {
     "text": "Pinterest's internal MCP ecosystem handled about 66,000 invocations a month from 844 users, saving about 7,000 hours a month.",
     "src": [
      1
     ]
    }
   ],
   "guardrails": [
    {
     "text": "Pinterest routes work to cheaper models where it can to control cost.",
     "src": [
      0
     ]
    }
   ],
   "hiring_reviews_changes": [],
   "results": [],
   "sources": [
    {
     "title": "Motley Fool: Pinterest Q2 2026 earnings call transcript",
     "url": "https://www.fool.com/earnings/call-transcripts/2026/08/11/pinterest-pins-q2-2026-earnings-call-transcript/",
     "date": "2026-08-11",
     "type": "primary"
    },
    {
     "title": "InfoQ: Pinterest's MCP ecosystem",
     "url": "https://www.infoq.com/news/2026/04/pinterest-mcp-ecosystem/",
     "date": "2026-04-01",
     "type": "press"
    }
   ],
   "notes": "InfoQ date is month-level.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "Reddit",
   "slug": "reddit",
   "industry": "Social media",
   "size": "Under 5k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. CEO Steve Huffman says AI has made engineers more productive, that code review is now the bottleneck, and that Reddit is not cutting engineers.",
    "src": [
     0
    ]
   },
   "tools": [],
   "metric_tags": [
    "Code review",
    "Self-reported productivity"
   ],
   "metrics": [],
   "guardrails": [],
   "hiring_reviews_changes": [
    {
     "text": "Huffman said Reddit is not cutting engineering jobs because of AI and is hiring AI-native new graduates.",
     "src": [
      0,
      1
     ]
    }
   ],
   "results": [
    {
     "text": "Huffman: the bottleneck 'is actually code review' now that AI writes more code.",
     "src": [
      0
     ]
    }
   ],
   "sources": [
    {
     "title": "Livemint: Reddit CEO Steve Huffman on AI productivity and no engineer job cuts",
     "url": "https://www.livemint.com/companies/news/reddit-ceo-steve-huffman-artificial-intelligence-productive-no-engineer-job-cuts-hire-ai-native-new-graduates-write-code-11774191994718.html",
     "date": "2026-03-22",
     "type": "press"
    },
    {
     "title": "Yahoo Finance: Reddit CEO Steve Huffman says",
     "url": "https://finance.yahoo.com/sectors/technology/articles/reddit-ceo-steve-huffman-says-173110772.html",
     "date": "2026-03-22",
     "type": "press"
    }
   ],
   "notes": "Reddit has published no adoption or code-share figures.",
   "last_verified": "2026-10-25"
  },
  {
   "name": "SpaceX (incl. SpaceXSI, formerly xAI)",
   "slug": "spacex-spacexsi",
   "industry": "AI lab / aerospace",
   "size": "5k-50k employees",
   "size_src": [],
   "mandate": "unclear",
   "mandate_summary": {
    "text": "No mandate reported. xAI merged into SpaceX in February 2026, rebranded as SpaceXAI, and in October was renamed SpaceXSI. After Anthropic cut off its Claude access via Cursor, leadership pushed toward in-house coding tools.",
    "src": [
     0,
     1,
     2,
     4
    ]
   },
   "tools": [
    {
     "name": "Grok / in-house coding tools",
     "src": [
      4,
      5
     ]
    },
    {
     "name": "Cursor (acquired)",
     "src": [
      5
     ]
    },
    {
     "name": "Claude (via Cursor, cut off January 2026)",
     "src": [
      4
     ]
    }
   ],
   "metric_tags": [
    "Adoption rate",
    "Agent autonomy"
   ],
   "metrics": [],
   "guardrails": [],
   "hiring_reviews_changes": [
    {
     "text": "When Anthropic cut off Claude access through Cursor, cofounder Tony Wu told staff it would mean a 'productivity hit' but 'pushes us to develop our own coding product'.",
     "src": [
      4
     ]
    },
    {
     "text": "Top Starlink and Starship engineers shifted time to AI work, and the company acquired Cursor, Business Insider reported.",
     "src": [
      5
     ]
    },
    {
     "text": "Many xAI cofounders and engineers left around the SpaceX acquisition, and the AI unit was rebuilt.",
     "src": [
      6
     ]
    }
   ],
   "results": [],
   "sources": [
    {
     "title": "xAI: xAI joins SpaceX",
     "url": "https://x.ai/news/xai-joins-spacex",
     "date": "2026-02-02",
     "type": "company"
    },
    {
     "title": "Business Insider: xAI rebrands as SpaceXAI",
     "url": "https://www.businessinsider.com/xai-rebrand-spacexai-new-logo-x-handle-spacex-2026-7",
     "date": "2026-07-06",
     "type": "press"
    },
    {
     "title": "Digital Today: SpaceXAI changes X account to SpaceXSI",
     "url": "https://www.digitaltoday.co.kr/en/view/114125/spacexai-changes-x-account-to-spacexsi",
     "date": "2026-10-14",
     "type": "press"
    },
    {
     "title": "Reuters: Musk says he will rename SpaceXAI to SpaceXSI",
     "url": "https://www.reuters.com/business/media-telecom/musk-says-he-will-rename-spacexai-spacexsi-2026-10-04/",
     "date": "2026-10-04",
     "type": "press"
    },
    {
     "title": "Times of India: xAI cofounder Tony Wu email on Anthropic ban",
     "url": "https://timesofindia.indiatimes.com/technology/tech-news/elon-musks-xai-cofounder-tony-wu-sends-email-to-employees-says-anthropic-has-banned-us-this-is-a-/articleshow/126463060.cms",
     "date": "2026-01-11",
     "type": "press"
    },
    {
     "title": "Business Insider: Grok 4.5, Starship and Starlink engineers, and Cursor",
     "url": "https://www.businessinsider.com/grok-4-5-starship-starlink-engineers-elon-musk-ai-cursor-2026-6",
     "date": "2026-06-29",
     "type": "press"
    },
    {
     "title": "Fortune: xAI rebuilding after cofounder and engineer exodus",
     "url": "https://fortune.com/2026/03/16/elon-musk-xai-rebuilding-cofounders-engineers-exodus-macrohard-project-spacex-acquisition/",
     "date": "2026-03-16",
     "type": "press"
    }
   ],
   "notes": "One combined entry: xAI joined SpaceX in February 2026, was rebranded SpaceXAI in July and SpaceXSI in October. Reuters blocks automated checks. No adoption, code-share or productivity figures have been published.",
   "last_verified": "2026-10-25"
  }
 ]
}