# cognition: the idea

> the ai-software-engineer bet — what cognition inherited, what it proved, what stayed unfinished, and who inherits it next.

## lineage

- 1950s– — automatic programming: the dream itself — automatic programming goes back to the 1950s; the wikipedia lineage runs through program synthesis, end-user programming, and every 'natural language to code' promise. devin's 'first ai software engineer' claim inherits seventy years of the same promise, at the moment the models could nearly keep it. (sources: [parnas: automatic programming as euphemism for higher-level language](https://en.wikipedia.org/wiki/Automatic_programming), [matt welsh: dinosaurs waiting for the meteor](https://cacm.acm.org/opinion/the-end-of-programming/))
- 2021 — codex and the copilot era: openai's original codex (2021) — the descendant of gpt-3 that powered github copilot and established the assistive paradigm: autocomplete, pair programming, a human still driving. the thing devin was positioned against from its first sentence. (sources: [codex model launch (copilots engine)](https://openai.com/index/openai-codex/), [your ai pair programmer — assistive-era framing](https://github.blog/news-insights/product-news/introducing-github-copilot-ai-pair-programmer/))
- 2023– — swe-bench and its audits: swe-bench (2023) — princeton's real-github-issue benchmark turned 'can an llm fix a bug' into a number. cognition exploited its asymmetries (the 25% subset, the unassisted framing); the audit literature (swe-bench+, verified) then eroded the ruler itself. (sources: [swe-bench benchmark (princeton et al.)](https://arxiv.org/abs/2310.06770), [swe-bench+ audit: 32.67% solution leakage](https://arxiv.org/abs/2410.06992), [human-validated 500-issue subset](https://openai.com/index/introducing-swe-bench-verified/), [methodology: 25% subset (570/2294), unassisted, github eval harness](https://cognition.com/blog/swe-bench-technical-report))
- 2023–2024 — the agent convergence: the agent scaffolding lineage — tool use, planning loops, and long-horizon task decomposition converging in 2023–24; the arxiv 'converged' argument is that agentic coding systems landed on roughly the same architecture at roughly the same time. (sources: [leaderboard-convergence audit](https://arxiv.org/pdf/2609.17394), [category taxonomy](https://earlyterms.com/term/coding-agents))
- 2000s–2023 — the ioi pipeline: the competitive-programming pipeline — ioi/icpc culture as recruiting and method: cognition's founders plus neal wu, korotkevich, and andrew he; peer companies scale, perplexity, and decagon came from the same circuit. the contest style — bounded problems, adversarial test cases, speed — shaped what 'a software engineer' meant to the people building one. (sources: [10 ioi gold medals](https://cognition.com/about), [long q&a](https://colossus.com/article/scott-wu-tapes-cognition/), [ashlee vance launch feature](https://www.bloomberg.com/news/articles/2024-03-12/cognition-ai-is-a-peter-thiel-backed-coding-assistant), [tourist](https://en.wikipedia.org/wiki/Gennady_Korotkevich))
- 2024– — from copilot to coding agent: the copilot→agent vocabulary shift — 'ai that helps you code' to 'ai that does the ticket.' devin's launch forced the reframe; by 2025–26 the whole field (codex, jules, copilot coding agent, claude code, swe-1) had adopted the agent frame devin introduced. (sources: [scott wu launch post](https://cognition.com/blog/introducing-devin), [codex cloud se agent, codex-1/o3-derived](https://openai.com/index/introducing-codex/), [jules public beta](https://blog.google/innovation-and-ai/models-and-research/google-labs/jules/), [claude code: preview 2025-02-24, ga 2025-05-22](https://aicost.tools/q/claude-code/when-was-claude-code-released/), [windsurf swe-1 family launch](https://techcrunch.com/2025/05/15/vibe-coding-startup-windsurf-launches-in-house-ai-models/))
- 2024– — the migration wedge: the enterprise-migration wedge — the boring work nobody wants: code modernization, legacy upgrades, large-repo context. microsoft's azure partnership aimed devin at it in 2024; headsup's migrations writeup and goldman's 'drudgery work' framing are the idea finding its actual market. (sources: [microsoft build partnership](https://www.techzine.eu/news/devops/120539/microsoft-brings-independent-application-builder-devin-within-reach/), [angellist 14k dashboards 5.2x](https://headsupai.io/updates/cognition-deploys-devin-enterprise-scale-database-monolith-migrations), [argenti: hundreds… might go into the thousands](https://www.cnbc.com/2025/07/11/goldman-sachs-autonomous-coder-pilot-marks-major-ai-milestone.html), [hybrid workforce](https://fortune.com/2025/07/14/goldman-sachs-ai-powered-software-engineer-devin-new-employee-increase-productivity-fears-of-job-replacement/))

## what cognition proved

- naming the category: the category got named — 'the first ai software engineer' was a claim, not a description, but it made 'coding agent' a category overnight. by 2025–26 every major lab ships one under devin's framing: codex, jules, copilot coding agent, claude code. the naming is the durable win. (sources: [scott wu launch post](https://cognition.com/blog/introducing-devin), [codex cloud se agent, codex-1/o3-derived](https://openai.com/index/introducing-codex/), [jules public beta](https://blog.google/innovation-and-ai/models-and-research/google-labs/jules/), [your ai pair programmer — assistive-era framing](https://github.blog/news-insights/product-news/introducing-github-copilot-ai-pair-programmer/), [claude code: preview 2025-02-24, ga 2025-05-22](https://aicost.tools/q/claude-code/when-was-claude-code-released/))
- agent work sells: agents can do real billable-shaped work — not reliably enough to justify 'software engineer,' but enough to run a business: goldman's hundreds-of-devins deployment, the '16x' timeline claims, a $20 entry plan that works because the failure modes are priced in rather than eliminated. (sources: [argenti: hundreds… might go into the thousands](https://www.cnbc.com/2025/07/11/goldman-sachs-autonomous-coder-pilot-marks-major-ai-milestone.html), [hybrid workforce](https://fortune.com/2025/07/14/goldman-sachs-ai-powered-software-engineer-devin-new-employee-increase-productivity-fears-of-job-replacement/), [ga at $500/mo, no seat limits, slack/ide/api](https://cognition.com/blog/devin-generally-available), [eye-watering $500→$20](https://techcrunch.com/2025/04/03/devin-the-viral-coding-ai-agent-gets-a-new-pay-as-you-go-plan/))
- the valuation arc: a 2024 launch company can ride the agent wave to frontier-adjacent scale — $350m → $2b → $4b → $10.2b → $26b → $48b in ~30 months, with in-house models (swe-1.x, swe-2, swe-check) served by cerebras. the valuation arc is the strongest evidence the bet worked — all figures reported, most self-reported at the company layer. (sources: [~$4b round led by 8vc](https://www.bloomberg.com/news/articles/2025-03-18/cognition-ai-hits-4-billion-valuation-in-deal-led-by-lonsdale-s-firm), [arr $1m(sep24)→$73m(jun25)](https://techcrunch.com/2025/09/08/cognition-ai-defies-turbulence-with-a-400m-raise-at-10-2b-valuation/), [$25b pre/$26b post](https://techcrunch.com/2026/05/27/ai-coding-startup-cognition-raises-1b-at-25b-pre-money-valuation/), [series e $2b at $48b](https://www.reuters.com/technology/cognition-ai-raises-2-billion-48-billion-valuation-2026-09-08/), [frontier-size model](https://cognition.com/blog/swe-1-5))
- the windsurf conversion: distribution beats capability at the margin — the windsurf deal converted a dying acquisition target into an ide channel, a customer list, and $82m of arr in a weekend; by april 2026 'devin in windsurf' and june 2026 devin desktop made the acquisition the product, not the press release. (sources: [definitive agreement](https://cognition.com/blog/windsurf), [250-person team left behind](https://techcrunch.com/2025/07/14/cognition-maker-of-the-ai-coding-agent-devin-acquires-windsurf/), [devin integration into windsurf ships (index listing)](https://cognition.com/blog/devin-in-windsurf), [devin desktop (index listing)](https://cognition.com/blog/introducing-devin-desktop))
- the labor frame held: the labor framing survived the debunk — 'hire devin,' 'like our new employee,' 'a full operating ai software engineer,' engineer-hour guarantees: cognition kept selling labor-shaped software while discourse argued about whether it could. the guarantee even prices the claim. (sources: [scott wu launch post](https://cognition.com/blog/introducing-devin), [argenti: hundreds… might go into the thousands](https://www.cnbc.com/2025/07/11/goldman-sachs-autonomous-coder-pilot-marks-major-ai-milestone.html), [verbatim origin-story post (ti-83 basic, age 9)](https://www.linkedin.com/posts/scott-wu-8b94ab96_i-learned-to-program-when-i-was-nine-years-activity-7480660748705157120-EBoW), [ai productivity guarantee](https://cognition.com/blog/ai-guarantee))

## what stayed unfinished

- the gap stayed a gap: the reliability gap never closed on the original terms — answer.ai's 3-of-20 stands as the independent benchmark; the 'unassisted 13.86%' never got a like-for-like reproduction; cognition's answer was to stop reporting swe-bench and to lower the price until 'sometimes wrong' was affordable, not to close the gap. (sources: [month-long eval: 3 successes/14 failures/3 inconclusive of 20](https://www.answer.ai/posts/2025-01-08-devin), [frontier-size model](https://cognition.com/blog/swe-1-5), [eye-watering $500→$20](https://techcrunch.com/2025/04/03/devin-the-viral-coding-ai-agent-gets-a-new-pay-as-you-go-plan/), [price-collapse-as-admission read](https://plainai.aguidetocloud.com/changes/cognition-devin-2/))
- autonomy receded into supervision: the 'fully autonomous' claim stalled into supervision — devin 2.x added interactive planning, confidence indicators, devin review, and 'devins managing devins' — the product itself migrating from 'replace the engineer' toward 'augment the reviewer.' the launch's strongest claim became the roadmap's quietest admission. (sources: [devin 2.0](https://cognition.com/blog/devin-2), [full dated post index used for timeline items](https://cognition.com/blog), [ga at $500/mo, no seat limits, slack/ide/api](https://cognition.com/blog/devin-generally-available))
- no trusted ruler: the benchmark problem went unsolved — swe-bench's authority decayed (leakage, subset shopping, vendor self-evals) without a replacement ruler; by 2026 every agent company grades its own homework and the 'how good is it really' question the launch raised is still unanswered in public. (sources: [swe-bench+ audit: 32.67% solution leakage](https://arxiv.org/abs/2410.06992), [human-validated 500-issue subset](https://openai.com/index/introducing-swe-bench-verified/), [openai stops self-reporting (feb 2026)](https://agentmarketcap.ai/blog/2026/04/12/openai-swe-bench-self-reporting-contamination-crisis), [99/100 entries self-reported](https://www.digitalapplied.com/blog/swe-bench-verified-june-2026-benchmark-vs-scaffolding-analysis))
- opaque agent economics: the unit economics stay opaque — acus, core plans, run-rate revenue, 'net burn under $20m,' 3–4x productivity claims, the guarantee's estimator: the metrics are real business data but all company-published; nobody outside cognition can compute the margin on an engineer-hour. (sources: [$500/mo as b2b junior-engineer pricing read](https://www.productgeeks.com/p/devin-ai-pricing), [scott wu: >$400m at $10.2b post](https://cognition.com/blog/funding-growth-and-the-next-frontier-of-ai-coding-agents), [ai productivity guarantee](https://cognition.com/blog/ai-guarantee), [run-rate $492m→~$900m](https://techcrunch.com/2026/09/08/cognition-hits-48b-valuation-signaling-investors-believe-ai-coding-is-far-from-a-winner-take-all-market/))

## the agent-era turn

- (for) devin won the frame — the whole industry now sells 'coding agents' and ships agent-native ides; cognition's sweep (models, agent, ide, enterprise contracts) positions it as the independent agent lab while the foundation-model companies treat coding as a feature. (sources: [$1b+ at $26b](https://cognition.com/blog/series-d), [codex cloud se agent, codex-1/o3-derived](https://openai.com/index/introducing-codex/), [jules public beta](https://blog.google/innovation-and-ai/models-and-research/google-labs/jules/), [devin desktop (index listing)](https://cognition.com/blog/introducing-devin-desktop))
- (for) the wedge is the mundane — migrations, legacy updates, security swarms, 'drudgery work': goldman's deployment and the enterprise partnerships describe a business automating the work engineers avoid, not the work they do. that is enough to be very large. (sources: [argenti: hundreds… might go into the thousands](https://www.cnbc.com/2025/07/11/goldman-sachs-autonomous-coder-pilot-marks-major-ai-milestone.html), [angellist 14k dashboards 5.2x](https://headsupai.io/updates/cognition-deploys-devin-enterprise-scale-database-monolith-migrations), [government offering (index listing)](https://cognition.com/blog/cognition-for-government), [microsoft build partnership](https://www.techzine.eu/news/devops/120539/microsoft-brings-independent-application-builder-devin-within-reach/))
- (against) the capability claims were always ahead of the product — the debunk was right about the demo, answer.ai was right about the reliability, and the 13.86% was a subset result marketed as a headline. the company's response — price cuts, supervision features, benchmark disclaimers — is an admission the launch oversold. (sources: [internet of bugs frame-by-frame debunk of upwork demo](https://www.youtube.com/watch?v=tNmgmwEtoWE), [month-long eval: 3 successes/14 failures/3 inconclusive of 20](https://www.answer.ai/posts/2025-01-08-devin), [first ai software engineer is bad at its job](https://www.theregister.com/software/2025/01/23/first-ai-software-engineer-is-bad-at-its-job/), [price-collapse-as-admission read](https://plainai.aguidetocloud.com/changes/cognition-devin-2/), [frontier-size model](https://cognition.com/blog/swe-1-5))
- (against) the moat is contested — 'everyone builds the coding agent': openai has codex, anthropic has claude code, google has jules plus windsurf's founders, cursor has the mindshare, and the openhands community replicates the surface area for free. independence is a strategy or a euphemism for 'no platform underneath.' (sources: [codex cloud se agent, codex-1/o3-derived](https://openai.com/index/introducing-codex/), [claude code: preview 2025-02-24, ga 2025-05-22](https://aicost.tools/q/claude-code/when-was-claude-code-released/), [jules public beta](https://blog.google/innovation-and-ai/models-and-research/google-labs/jules/), [google $2.4b nonexclusive license + hires](https://www.reuters.com/business/google-hires-windsurf-ceo-researchers-advance-ai-ambitions-2025-07-11/), [open-source replication created day after launch](https://github.com/opendevin/OpenDevin), [ramp data: cursor share 41%→26%](https://business-news-today.com/spacex-finalises-60bn-cursor-acquisition-targets-anthropic-in-ai-coding-fight/))
- (nuance) the labor math is real but earlier than advertised — amodei's '90%' and cognition's internal '89–95%' describe the same trend from opposite sides of the cap table; the honest version is that agents absorb tickets faster than headcount shrinks, and the surplus shows up as throughput before it shows up as layoffs. (sources: [amodei cfr prediction (spoken 2025-03-10)](https://www.businessinsider.com/anthropic-ceo-ai-90-percent-code-3-to-6-months-2025-3), [scorecard: ~50% merged-code share at anthropic](https://blog.redwoodresearch.org/p/is-90-of-code-at-anthropic-being), [bi launch coverage, job-loss framing](https://www.businessinsider.com/cognition-labs-devin-ai-software-engineer-humans-lose-jobs-2024-3), [$1b+ at $26b](https://cognition.com/blog/series-d), [hybrid workforce](https://fortune.com/2025/07/14/goldman-sachs-ai-powered-software-engineer-devin-new-employee-increase-productivity-fears-of-job-replacement/))
- (nuance) the talent-machine question — the ioi pipeline supplied the founders and the mythos; whether it supplies the moat depends on whether competitive-programming skill transfers to agent systems, or whether it mostly transferred to storytelling. the windsurf bench being 'five or six people' cuts both ways. (sources: [10 ioi gold medals](https://cognition.com/about), [long q&a](https://colossus.com/article/scott-wu-tapes-cognition/), [wang interview: reorgs every 2–3 months](https://www.youtube.com/watch?v=uNQLawiYhQE), [ashlee vance launch feature](https://www.bloomberg.com/news/articles/2024-03-12/cognition-ai-is-a-peter-thiel-backed-coding-assistant))

## successors

- [openai codex](https://openai.com/codex): the foundation-model answer — chatgpt's coding agent, launched into the post-devin category in 2025; the platform-underneath threat cognition's 'independence' is priced against. (sources: [codex cloud se agent, codex-1/o3-derived](https://openai.com/index/introducing-codex/), [codex launch coverage](https://techcrunch.com/2025/05/16/openai-launches-codex-an-ai-coding-agent-in-chatgpt/))
- [google jules + deepmind](https://jules.google): google's async coding agent — plus the windsurf founders inside deepmind working on agentic coding for gemini. the deal that made cognition also made its most dangerous competitor. (sources: [jules public beta](https://blog.google/innovation-and-ai/models-and-research/google-labs/jules/), [google $2.4b nonexclusive license + hires](https://www.reuters.com/business/google-hires-windsurf-ceo-researchers-advance-ai-ambitions-2025-07-11/))
- [github copilot coding agent](https://github.com/features/copilot): the incumbent's pivot — copilot grew an agent mode and an async coding agent; the 'copilot' brand now covers the thing devin defined against it. (sources: [your ai pair programmer — assistive-era framing](https://github.blog/news-insights/product-news/introducing-github-copilot-ai-pair-programmer/))
- [claude code](https://claude.com/product/claude-code): anthropic's agent — the terminal-native tool that repriced expectations, and the vendor whose claude access was the windsurf deal's hidden asset. cognition builds on it while competing with it. (sources: [claude code: preview 2025-02-24, ga 2025-05-22](https://aicost.tools/q/claude-code/when-was-claude-code-released/), [anthropic cuts claude 3.x first-party capacity](https://techcrunch.com/2025/06/03/windsurf-says-anthropic-is-limiting-its-direct-access-to-claude-ai-models/), [definitive agreement](https://cognition.com/blog/windsurf))
- [cursor](https://cursor.com): the ide-first rival — the mindshare leader in agentic editing, the customer-base overlap (spacex-scale logos on both sides), and walden yan's reported former employer. the fork in the road cognition's ide pivot was answering. (sources: [ramp data: cursor share 41%→26%](https://business-news-today.com/spacex-finalises-60bn-cursor-acquisition-targets-anthropic-in-ai-coding-fight/), [$60b all-stock anysphere deal](https://www.forbes.com/sites/sandycarter/2026/06/16/spacex-buys-cursor-in-largest-startup-acquisition-ever-at-60-billion/), [compiled leadership bios](https://mungomash.com/orgs/cognition/leadership/))
- [openhands](https://github.com/All-Hands-AI/OpenHands): the open-source replication — created the day after the launch, ~85k stars; the standing 'we can just build it' answer and the control group for every capability claim. (sources: [open-source replication created day after launch](https://github.com/opendevin/OpenDevin))

## open questions

- does 'independent agent lab' survive the foundation-model squeeze — can a company that rides claude, gpt, and its own swe models out-innovate the labs it depends on?
- what does an agent-hour actually cost — the acu pricing, the $20 plan, and the guarantee's estimator all describe the same opaque unit; when does someone outside cognition get to check the math?
- does the 13.86%→benchmark-decay arc repeat — will swe-2's self-reported numbers meet a swe-bench-pro audit, and does cognition keep a benchmark it can report?
- was windsurf a rescue or a raid — the 'one boat' memo promised 100% participation; does the retained team's output at devin desktop vindicate the six-day memo?
- can the ioi-pipeline moat compound — does competitive-programming skill keep transferring to agent systems, or does the model layer commoditize the advantage?
- what happens to the 'is devin real' question when the answer is a $48b company — does the discourse update, or does the demo-reality gap just get amortized?
- does outcome pricing hold — the productivity guarantee is the strongest version of 'agents are labor'; does the self-measured estimator survive a contested claim?
- who is customer no. 1 in the next category — does the government/genesis-mission work make devin critical infrastructure before the discourse notices?

AI-drafted at Ben Guo's direct request and credited to Hraness; every claim links to its cataloged source.

Canonical page: https://hraness.com/cognition/the-idea


Published by @ben in profile revision 1.
Original document: https://hraness.com/cognition/the-idea
Original credited author: Hraness
AI-drafted at Ben Guo's direct request and credited to Hraness; every claim links to its cataloged source.

## Structured source and record data

````json
{"anchors":[{"id":"lineage","originalUrl":"https://hraness.com/cognition/the-idea#lineage","title":"lineage"},{"id":"proved","originalUrl":"https://hraness.com/cognition/the-idea#proved","title":"what cognition proved"},{"id":"stalled","originalUrl":"https://hraness.com/cognition/the-idea#stalled","title":"what stayed unfinished"},{"id":"agent-era","originalUrl":"https://hraness.com/cognition/the-idea#agent-era","title":"the agent-era turn"},{"id":"successors","originalUrl":"https://hraness.com/cognition/the-idea#successors","title":"successors"},{"id":"open-questions","originalUrl":"https://hraness.com/cognition/the-idea#open-questions","title":"open questions"}],"provenance":{"draftingDisclosure":"AI-drafted at Ben Guo's direct request and credited to Hraness; every claim links to its cataloged source.","originalAuthor":"Hraness","originalRevision":"be164a1f22c8a4e49c835ce16046598d74925a34d2bd8180454a6b91df5b0c48","originalUrl":"https://hraness.com/cognition/the-idea","source":"hraness"},"records":[{"id":"record-e00dc8f3ca10cc6334cb","kind":"cognition-agent-era-arguments","label":"devin won the frame — the whole industry now sells 'coding agents' and ships agent-native ides; cognition's sweep (models, agent, ide, enterprise contracts) positions it as the independent agent lab while the foundation-model companies treat coding as a feature.","original":{"claim":"devin won the frame — the whole industry now sells 'coding agents' and ships agent-native ides; cognition's sweep (models, agent, ide, enterprise contracts) positions it as the independent agent lab while the foundation-model companies treat coding as a feature.","id":"arg-frame-won","side":"for","sourceIds":["src-ci-cognition-seriesd","src-ci-openai-codex","src-ci-google-jules","src-ci-cognition-devin-desktop"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-1bbad291f76f701d0539","source-ab68c048f622c5487cd3","source-64ac502a1459aa6bdd73","source-b2840ab1745c1bf7e839"]},{"id":"record-216a95631a4a48f6e5b9","kind":"cognition-agent-era-arguments","label":"the wedge is the mundane — migrations, legacy updates, security swarms, 'drudgery work': goldman's deployment and the enterprise partnerships describe a business automating the work engineers avoid, not the work they do. that is enough to be very large.","original":{"claim":"the wedge is the mundane — migrations, legacy updates, security swarms, 'drudgery work': goldman's deployment and the enterprise partnerships describe a business automating the work engineers avoid, not the work they do. that is enough to be very large.","id":"arg-mundane-wedge","side":"for","sourceIds":["src-ci-cnbc-goldman","src-ci-headsup-migrations","src-ci-cognition-gov","src-hp-techzine-msft"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-27be554657a19b7483e8","source-dd9b0f8f0929d484fb44","source-2e3148dc6befba8e5d15","source-52d6624601297ebd17a9"]},{"id":"record-3b62213cfcb0ec60d19d","kind":"cognition-agent-era-arguments","label":"the capability claims were always ahead of the product — the debunk was right about the demo, answer.ai was right about the reliability, and the 13.86% was a subset result marketed as a headline. the company's response — price cuts, supervision features, benchmark disclaimers — is an admission the launch oversold.","original":{"claim":"the capability claims were always ahead of the product — the debunk was right about the demo, answer.ai was right about the reliability, and the 13.86% was a subset result marketed as a headline. the company's response — price cuts, supervision features, benchmark disclaimers — is an admission the launch oversold.","id":"arg-oversold","side":"against","sourceIds":["src-ci-iob-debunk-video","src-ci-answerai-devin","src-ci-register-devin-bad","src-ci-plainai-devin2","src-ci-cognition-swe15"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-b0dafab463859ad5de28","source-3e2dace97d27f013de2c","source-bd86c98a6793c69ef8c4","source-034c7082737b3e1133af","source-0db751b75021fab92d04"]},{"id":"record-e6be1ff5cbd9c4ae1c20","kind":"cognition-agent-era-arguments","label":"the moat is contested — 'everyone builds the coding agent': openai has codex, anthropic has claude code, google has jules plus windsurf's founders, cursor has the mindshare, and the openhands community replicates the surface area for free. independence is a strategy or a euphemism for 'no platform underneath.'","original":{"claim":"the moat is contested — 'everyone builds the coding agent': openai has codex, anthropic has claude code, google has jules plus windsurf's founders, cursor has the mindshare, and the openhands community replicates the surface area for free. independence is a strategy or a euphemism for 'no platform underneath.'","id":"arg-moat-contested","side":"against","sourceIds":["src-ci-openai-codex","src-ci-aicost-claudecode","src-ci-google-jules","src-hp-reuters-google-windsurf","src-ci-opendevin-repo","src-ci-bnt-cursor-share"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-ab68c048f622c5487cd3","source-75343f5d5daaea74a9ea","source-64ac502a1459aa6bdd73","source-b87d95c58f8c4efe90a7","source-96df22d52de2b697c293","source-5d96dd0653e8ac0fc447"]},{"id":"record-27b135c615c9a5318318","kind":"cognition-agent-era-arguments","label":"the labor math is real but earlier than advertised — amodei's '90%' and cognition's internal '89–95%' describe the same trend from opposite sides of the cap table; the honest version is that agents absorb tickets faster than headcount shrinks, and the surplus shows up as throughput before it shows up as layoffs.","original":{"claim":"the labor math is real but earlier than advertised — amodei's '90%' and cognition's internal '89–95%' describe the same trend from opposite sides of the cap table; the honest version is that agents absorb tickets faster than headcount shrinks, and the surplus shows up as throughput before it shows up as layoffs.","id":"arg-labor-math","side":"nuance","sourceIds":["src-ci-amodei-90","src-ci-redwood-amodei","src-ci-bi-devin-jobs","src-ci-cognition-seriesd","src-ci-fortune-goldman"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-67737b0b0ce0e8a02d33","source-ebeb343faa4710da4e12","source-f23722382305071bb0f1","source-1bbad291f76f701d0539","source-9c8b8da0c4acd715460c"]},{"id":"record-068e62918c0a50b224f4","kind":"cognition-agent-era-arguments","label":"the talent-machine question — the ioi pipeline supplied the founders and the mythos; whether it supplies the moat depends on whether competitive-programming skill transfers to agent systems, or whether it mostly transferred to storytelling. the windsurf bench being 'five or six people' cuts both ways.","original":{"claim":"the talent-machine question — the ioi pipeline supplied the founders and the mythos; whether it supplies the moat depends on whether competitive-programming skill transfers to agent systems, or whether it mostly transferred to storytelling. the windsurf bench being 'five or six people' cuts both ways.","id":"arg-talent-machine","side":"nuance","sourceIds":["src-hp-cognition-about","src-hp-colossus-wu-tapes","src-ci-jeffwang-neoshow","src-hp-bloomberg-gold-medalist-coders"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-6d3eaa02c5959695b1f4","source-915cc5f7006791fc9a01","source-b174a05dfca3de702d20","source-9000b86d1e91e6cdee86"]},{"id":"record-2bcd2c81c4ece5b2da05","kind":"cognition-idea-lineage","label":"automatic programming","original":{"contribution":"the dream itself — automatic programming goes back to the 1950s; the wikipedia lineage runs through program synthesis, end-user programming, and every 'natural language to code' promise. devin's 'first ai software engineer' claim inherits seventy years of the same promise, at the moment the models could nearly keep it.","era":"1950s–","id":"lin-automatic-programming","name":"automatic programming","sourceIds":["src-ci-wikipedia-autoprog","src-ci-cacm-end-of-programming"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-866b7c02ca55d3a1d84c","source-ccae308cc63e68ee2466"]},{"id":"record-3779db7f3fc92b960dd1","kind":"cognition-idea-lineage","label":"codex and the copilot era","original":{"contribution":"openai's original codex (2021) — the descendant of gpt-3 that powered github copilot and established the assistive paradigm: autocomplete, pair programming, a human still driving. the thing devin was positioned against from its first sentence.","era":"2021","id":"lin-codex-copilot","name":"codex and the copilot era","sourceIds":["src-ci-openai-codex-2021","src-ci-copilot-launch"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-eef0e6c4854564bf3f68","source-60f035c084f29a27c33d"]},{"id":"record-f4c7f1f6e3967a7c6f68","kind":"cognition-idea-lineage","label":"swe-bench and its audits","original":{"contribution":"swe-bench (2023) — princeton's real-github-issue benchmark turned 'can an llm fix a bug' into a number. cognition exploited its asymmetries (the 25% subset, the unassisted framing); the audit literature (swe-bench+, verified) then eroded the ruler itself.","era":"2023–","id":"lin-swebench","name":"swe-bench and its audits","sourceIds":["src-ci-swebench-paper","src-ci-swebench-plus","src-ci-swebench-verified","src-hp-cognition-swebench-report"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-521b4186c88ec2132123","source-7fb1ba94cb65b0e117ca","source-d0e3c86f1a860f403712","source-1ecd5bcad1a9647b2911"]},{"id":"record-7f0a0c7bcfbe06dc89e8","kind":"cognition-idea-lineage","label":"the agent convergence","original":{"contribution":"the agent scaffolding lineage — tool use, planning loops, and long-horizon task decomposition converging in 2023–24; the arxiv 'converged' argument is that agentic coding systems landed on roughly the same architecture at roughly the same time.","era":"2023–2024","id":"lin-agent-convergence","name":"the agent convergence","sourceIds":["src-ci-arxiv-converged","src-ci-earlyterms-agents"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-e36f16e72ae999b85b10","source-c992c96cdd103473ee96"]},{"id":"record-1bfc5b42f08b0f9cf206","kind":"cognition-idea-lineage","label":"the ioi pipeline","original":{"contribution":"the competitive-programming pipeline — ioi/icpc culture as recruiting and method: cognition's founders plus neal wu, korotkevich, and andrew he; peer companies scale, perplexity, and decagon came from the same circuit. the contest style — bounded problems, adversarial test cases, speed — shaped what 'a software engineer' meant to the people building one.","era":"2000s–2023","id":"lin-ioi-pipeline","name":"the ioi pipeline","sourceIds":["src-hp-cognition-about","src-hp-colossus-wu-tapes","src-hp-bloomberg-gold-medalist-coders","src-hp-wiki-korotkevich"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-6d3eaa02c5959695b1f4","source-915cc5f7006791fc9a01","source-9000b86d1e91e6cdee86","source-c15321fb46b7ed3bf5b8"]},{"id":"record-931cf729bd59355e2ee6","kind":"cognition-idea-lineage","label":"from copilot to coding agent","original":{"contribution":"the copilot→agent vocabulary shift — 'ai that helps you code' to 'ai that does the ticket.' devin's launch forced the reframe; by 2025–26 the whole field (codex, jules, copilot coding agent, claude code, swe-1) had adopted the agent frame devin introduced.","era":"2024–","id":"lin-vocab-shift","name":"from copilot to coding agent","sourceIds":["src-hp-cognition-introducing-devin","src-ci-openai-codex","src-ci-google-jules","src-ci-aicost-claudecode","src-ci-tc-swe1"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-24b7f0d0ba5fb7d9ce77","source-ab68c048f622c5487cd3","source-64ac502a1459aa6bdd73","source-75343f5d5daaea74a9ea","source-1097c6dfa60fa364a20c"]},{"id":"record-1467761a7ea43ec19a82","kind":"cognition-idea-lineage","label":"the migration wedge","original":{"contribution":"the enterprise-migration wedge — the boring work nobody wants: code modernization, legacy upgrades, large-repo context. microsoft's azure partnership aimed devin at it in 2024; headsup's migrations writeup and goldman's 'drudgery work' framing are the idea finding its actual market.","era":"2024–","id":"lin-migration-wedge","name":"the migration wedge","sourceIds":["src-hp-techzine-msft","src-ci-headsup-migrations","src-ci-cnbc-goldman","src-ci-fortune-goldman"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-52d6624601297ebd17a9","source-dd9b0f8f0929d484fb44","source-27be554657a19b7483e8","source-9c8b8da0c4acd715460c"]},{"id":"record-fa3203b90f2e61e47426","kind":"cognition-idea-open-questions","label":"does 'independent agent lab' survive the foundation-model squeeze — can a company that rides claude, gpt, and its own swe models out-innovate the labs it depends on?","original":{"text":"does 'independent agent lab' survive the foundation-model squeeze — can a company that rides claude, gpt, and its own swe models out-innovate the labs it depends on?"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":[]},{"id":"record-930c30ac6adf25e38d47","kind":"cognition-idea-open-questions","label":"what does an agent-hour actually cost — the acu pricing, the $20 plan, and the guarantee's estimator all describe the same opaque unit; when does someone outside cognition get to check the math?","original":{"text":"what does an agent-hour actually cost — the acu pricing, the $20 plan, and the guarantee's estimator all describe the same opaque unit; when does someone outside cognition get to check the math?"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":[]},{"id":"record-a6ea2be56ce1ad521d96","kind":"cognition-idea-open-questions","label":"does the 13.86%→benchmark-decay arc repeat — will swe-2's self-reported numbers meet a swe-bench-pro audit, and does cognition keep a benchmark it can report?","original":{"text":"does the 13.86%→benchmark-decay arc repeat — will swe-2's self-reported numbers meet a swe-bench-pro audit, and does cognition keep a benchmark it can report?"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":[]},{"id":"record-3ec1e6e01bd706fa4b9f","kind":"cognition-idea-open-questions","label":"was windsurf a rescue or a raid — the 'one boat' memo promised 100% participation; does the retained team's output at devin desktop vindicate the six-day memo?","original":{"text":"was windsurf a rescue or a raid — the 'one boat' memo promised 100% participation; does the retained team's output at devin desktop vindicate the six-day memo?"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":[]},{"id":"record-ec5d54f2e7718b33c7ac","kind":"cognition-idea-open-questions","label":"can the ioi-pipeline moat compound — does competitive-programming skill keep transferring to agent systems, or does the model layer commoditize the advantage?","original":{"text":"can the ioi-pipeline moat compound — does competitive-programming skill keep transferring to agent systems, or does the model layer commoditize the advantage?"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":[]},{"id":"record-2365cfc8ad4a7eeff6a8","kind":"cognition-idea-open-questions","label":"what happens to the 'is devin real' question when the answer is a $48b company — does the discourse update, or does the demo-reality gap just get amortized?","original":{"text":"what happens to the 'is devin real' question when the answer is a $48b company — does the discourse update, or does the demo-reality gap just get amortized?"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":[]},{"id":"record-d2c2cfcdb00678d0b13b","kind":"cognition-idea-open-questions","label":"does outcome pricing hold — the productivity guarantee is the strongest version of 'agents are labor'; does the self-measured estimator survive a contested claim?","original":{"text":"does outcome pricing hold — the productivity guarantee is the strongest version of 'agents are labor'; does the self-measured estimator survive a contested claim?"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":[]},{"id":"record-fbd1b84072095e53ef64","kind":"cognition-idea-open-questions","label":"who is customer no. 1 in the next category — does the government/genesis-mission work make devin critical infrastructure before the discourse notices?","original":{"text":"who is customer no. 1 in the next category — does the government/genesis-mission work make devin critical infrastructure before the discourse notices?"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":[]},{"id":"record-e74c4e2274221eb17365","kind":"cognition-idea-proved","label":"naming the category","original":{"detail":"the category got named — 'the first ai software engineer' was a claim, not a description, but it made 'coding agent' a category overnight. by 2025–26 every major lab ships one under devin's framing: codex, jules, copilot coding agent, claude code. the naming is the durable win.","id":"clm-category-named","sourceIds":["src-hp-cognition-introducing-devin","src-ci-openai-codex","src-ci-google-jules","src-ci-copilot-launch","src-ci-aicost-claudecode"],"title":"naming the category"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-24b7f0d0ba5fb7d9ce77","source-ab68c048f622c5487cd3","source-64ac502a1459aa6bdd73","source-60f035c084f29a27c33d","source-75343f5d5daaea74a9ea"]},{"id":"record-28f063e02b47e7b61f10","kind":"cognition-idea-proved","label":"agent work sells","original":{"detail":"agents can do real billable-shaped work — not reliably enough to justify 'software engineer,' but enough to run a business: goldman's hundreds-of-devins deployment, the '16x' timeline claims, a $20 entry plan that works because the failure modes are priced in rather than eliminated.","id":"clm-agent-work-sells","sourceIds":["src-ci-cnbc-goldman","src-ci-fortune-goldman","src-hp-cognition-ga","src-ci-tc-devin2"],"title":"agent work sells"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-27be554657a19b7483e8","source-9c8b8da0c4acd715460c","source-6605de8983b4cf7a104f","source-f1e10ae59149bbf19428"]},{"id":"record-7d1c1220bd57685e718b","kind":"cognition-idea-proved","label":"the valuation arc","original":{"detail":"a 2024 launch company can ride the agent wave to frontier-adjacent scale — $350m → $2b → $4b → $10.2b → $26b → $48b in ~30 months, with in-house models (swe-1.x, swe-2, swe-check) served by cerebras. the valuation arc is the strongest evidence the bet worked — all figures reported, most self-reported at the company layer.","id":"clm-arc-worked","sourceIds":["src-hp-bloomberg-4b","src-ci-tc-10b","src-ci-tc-26b","src-hp-reuters-48b","src-ci-cognition-swe15"],"title":"the valuation arc"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-e89e502cd59154c94c6f","source-edd94c292c5fa0292bc5","source-2f9fa33e2b5ef408d29f","source-0760f59db9d9b7b278b7","source-0db751b75021fab92d04"]},{"id":"record-fa099d270c556a8cc0a7","kind":"cognition-idea-proved","label":"the windsurf conversion","original":{"detail":"distribution beats capability at the margin — the windsurf deal converted a dying acquisition target into an ide channel, a customer list, and $82m of arr in a weekend; by april 2026 'devin in windsurf' and june 2026 devin desktop made the acquisition the product, not the press release.","id":"clm-distribution","sourceIds":["src-ci-cognition-windsurf","src-ci-tc-cognition-windsurf","src-ci-cognition-devin-windsurf","src-ci-cognition-devin-desktop"],"title":"the windsurf conversion"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-b5c30eea0d72caf8fa23","source-6de360c4af0fe6cbb866","source-ce62ab0d51a90b48ed1f","source-b2840ab1745c1bf7e839"]},{"id":"record-0135c95d06c43315d473","kind":"cognition-idea-proved","label":"the labor frame held","original":{"detail":"the labor framing survived the debunk — 'hire devin,' 'like our new employee,' 'a full operating ai software engineer,' engineer-hour guarantees: cognition kept selling labor-shaped software while discourse argued about whether it could. the guarantee even prices the claim.","id":"clm-labor-frame","sourceIds":["src-hp-cognition-introducing-devin","src-ci-cnbc-goldman","src-hp-scottwu-linkedin-post","src-ci-cognition-guarantee"],"title":"the labor frame held"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-24b7f0d0ba5fb7d9ce77","source-27be554657a19b7483e8","source-8d6eed5ab8be721ab011","source-e6d3a1b178c2d535cf22"]},{"id":"record-03fcc4cfd8b700af362d","kind":"cognition-idea-stalled","label":"the gap stayed a gap","original":{"detail":"the reliability gap never closed on the original terms — answer.ai's 3-of-20 stands as the independent benchmark; the 'unassisted 13.86%' never got a like-for-like reproduction; cognition's answer was to stop reporting swe-bench and to lower the price until 'sometimes wrong' was affordable, not to close the gap.","id":"clm-reliability-gap","sourceIds":["src-ci-answerai-devin","src-ci-cognition-swe15","src-ci-tc-devin2","src-ci-plainai-devin2"],"title":"the gap stayed a gap"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-3e2dace97d27f013de2c","source-0db751b75021fab92d04","source-f1e10ae59149bbf19428","source-034c7082737b3e1133af"]},{"id":"record-e5e484c8b73573777057","kind":"cognition-idea-stalled","label":"autonomy receded into supervision","original":{"detail":"the 'fully autonomous' claim stalled into supervision — devin 2.x added interactive planning, confidence indicators, devin review, and 'devins managing devins' — the product itself migrating from 'replace the engineer' toward 'augment the reviewer.' the launch's strongest claim became the roadmap's quietest admission.","id":"clm-autonomy-receded","sourceIds":["src-ci-devin2","src-hp-cognition-blog-index","src-hp-cognition-ga"],"title":"autonomy receded into supervision"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-4f32044a32d07efe7de4","source-69d3c314b0e03d32799f","source-6605de8983b4cf7a104f"]},{"id":"record-bf367ebd965dfae5a56d","kind":"cognition-idea-stalled","label":"no trusted ruler","original":{"detail":"the benchmark problem went unsolved — swe-bench's authority decayed (leakage, subset shopping, vendor self-evals) without a replacement ruler; by 2026 every agent company grades its own homework and the 'how good is it really' question the launch raised is still unanswered in public.","id":"clm-benchmark-unsolved","sourceIds":["src-ci-swebench-plus","src-ci-swebench-verified","src-ci-agentmarketcap-swebench","src-ci-digitalapplied-swebench"],"title":"no trusted ruler"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-7fb1ba94cb65b0e117ca","source-d0e3c86f1a860f403712","source-e7e24fdf9169558ae8eb","source-3626bb632e52d9515da5"]},{"id":"record-c0e21aada276bc8be644","kind":"cognition-idea-stalled","label":"opaque agent economics","original":{"detail":"the unit economics stay opaque — acus, core plans, run-rate revenue, 'net burn under $20m,' 3–4x productivity claims, the guarantee's estimator: the metrics are real business data but all company-published; nobody outside cognition can compute the margin on an engineer-hour.","id":"clm-economics-opaque","sourceIds":["src-ci-productgeeks-pricing","src-hp-cognition-series-c","src-ci-cognition-guarantee","src-ci-tc-48b"],"title":"opaque agent economics"},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-9d13db24b6d98238da40","source-8aee4983e8bd1c160df7","source-e6d3a1b178c2d535cf22","source-5385b4f4e18becb33bbf"]},{"id":"record-4440691a383c0ac59e38","kind":"cognition-idea-successors","label":"openai codex","original":{"href":"https://openai.com/codex","id":"suc-codex","name":"openai codex","note":"the foundation-model answer — chatgpt's coding agent, launched into the post-devin category in 2025; the platform-underneath threat cognition's 'independence' is priced against.","sourceIds":["src-ci-openai-codex","src-ci-tc-codex"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-ab68c048f622c5487cd3","source-02bdccaebbf3168398e0"]},{"id":"record-2f39cc1b8d920ff99508","kind":"cognition-idea-successors","label":"google jules + deepmind","original":{"href":"https://jules.google","id":"suc-jules","name":"google jules + deepmind","note":"google's async coding agent — plus the windsurf founders inside deepmind working on agentic coding for gemini. the deal that made cognition also made its most dangerous competitor.","sourceIds":["src-ci-google-jules","src-hp-reuters-google-windsurf"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-64ac502a1459aa6bdd73","source-b87d95c58f8c4efe90a7"]},{"id":"record-14b061f14f74e4afff26","kind":"cognition-idea-successors","label":"github copilot coding agent","original":{"href":"https://github.com/features/copilot","id":"suc-copilot-agent","name":"github copilot coding agent","note":"the incumbent's pivot — copilot grew an agent mode and an async coding agent; the 'copilot' brand now covers the thing devin defined against it.","sourceIds":["src-ci-copilot-launch"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-60f035c084f29a27c33d"]},{"id":"record-4115fc96d1649be88326","kind":"cognition-idea-successors","label":"claude code","original":{"href":"https://claude.com/product/claude-code","id":"suc-claude-code","name":"claude code","note":"anthropic's agent — the terminal-native tool that repriced expectations, and the vendor whose claude access was the windsurf deal's hidden asset. cognition builds on it while competing with it.","sourceIds":["src-ci-aicost-claudecode","src-ci-tc-anthropic-windsurf","src-ci-cognition-windsurf"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-75343f5d5daaea74a9ea","source-8e9720f44994fc564ffa","source-b5c30eea0d72caf8fa23"]},{"id":"record-2a57bf345533aebd876a","kind":"cognition-idea-successors","label":"cursor","original":{"href":"https://cursor.com","id":"suc-cursor","name":"cursor","note":"the ide-first rival — the mindshare leader in agentic editing, the customer-base overlap (spacex-scale logos on both sides), and walden yan's reported former employer. the fork in the road cognition's ide pivot was answering.","sourceIds":["src-ci-bnt-cursor-share","src-ci-forbes-spacex-cursor","src-hp-mungomash-leadership"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-5d96dd0653e8ac0fc447","source-78cec34a9a8099f1b8ff","source-b7bf0d790557c9ef893c"]},{"id":"record-7abbca4792b1ec64071a","kind":"cognition-idea-successors","label":"openhands","original":{"href":"https://github.com/All-Hands-AI/OpenHands","id":"suc-openhands","name":"openhands","note":"the open-source replication — created the day after the launch, ~85k stars; the standing 'we can just build it' answer and the control group for every capability claim.","sourceIds":["src-ci-opendevin-repo"]},"originalUrl":"https://hraness.com/cognition/the-idea","sourceIds":["source-96df22d52de2b697c293"]}],"sources":[{"author":null,"id":"source-e7e24fdf9169558ae8eb","note":"openai stops self-reporting (feb 2026); verified→pro 80.9%→45.9%","originalId":"src-ci-agentmarketcap-swebench","published":{"precision":"day","text":"2026-04-12"},"publisher":"agentmarketcap.ai","title":"openai stops self-reporting (feb 2026)","type":"reference","url":"https://agentmarketcap.ai/blog/2026/04/12/openai-swe-bench-self-reporting-contamination-crisis"},{"author":null,"id":"source-75343f5d5daaea74a9ea","note":"claude code: preview 2025-02-24, ga 2025-05-22","originalId":"src-ci-aicost-claudecode","published":{"precision":"year","text":"2025"},"publisher":"aicost.tools","title":"claude code: preview 2025-02-24, ga 2025-05-22","type":"reference","url":"https://aicost.tools/q/claude-code/when-was-claude-code-released/"},{"author":null,"id":"source-67737b0b0ce0e8a02d33","note":"amodei cfr prediction (spoken 2025-03-10)","originalId":"src-ci-amodei-90","published":{"precision":"day","text":"2025-03-14"},"publisher":"business insider","title":"amodei cfr prediction (spoken 2025-03-10)","type":"reporting","url":"https://www.businessinsider.com/anthropic-ceo-ai-90-percent-code-3-to-6-months-2025-3"},{"author":null,"id":"source-3e2dace97d27f013de2c","note":"month-long eval: 3 successes/14 failures/3 inconclusive of 20","originalId":"src-ci-answerai-devin","published":{"precision":"day","text":"2025-01-08"},"publisher":"answer.ai","title":"month-long eval: 3 successes/14 failures/3 inconclusive of 20","type":"reference","url":"https://www.answer.ai/posts/2025-01-08-devin"},{"author":null,"id":"source-e36f16e72ae999b85b10","note":"leaderboard-convergence audit; scaffold swings to ~30 pts","originalId":"src-ci-arxiv-converged","published":{"precision":"month","text":"2026-09"},"publisher":"arxiv.org","title":"leaderboard-convergence audit","type":"reporting","url":"https://arxiv.org/pdf/2609.17394"},{"author":null,"id":"source-f23722382305071bb0f1","note":"bi launch coverage, job-loss framing","originalId":"src-ci-bi-devin-jobs","published":{"precision":"day","text":"2024-03-13"},"publisher":"business insider","title":"bi launch coverage, job-loss framing","type":"reporting","url":"https://www.businessinsider.com/cognition-labs-devin-ai-software-engineer-humans-lose-jobs-2024-3"},{"author":null,"id":"source-5d96dd0653e8ac0fc447","note":"ramp data: cursor share 41%→26%; anthropic ~half of category","originalId":"src-ci-bnt-cursor-share","published":{"precision":"day","text":"2026-08-17"},"publisher":"business-news-today.com","title":"ramp data: cursor share 41%→26%","type":"reporting","url":"https://business-news-today.com/spacex-finalises-60bn-cursor-acquisition-targets-anthropic-in-ai-coding-fight/"},{"author":null,"id":"source-ccae308cc63e68ee2466","note":"matt welsh: \"dinosaurs waiting for the meteor\"","originalId":"src-ci-cacm-end-of-programming","published":{"precision":"month","text":"2023-01"},"publisher":"cacm.acm.org","title":"matt welsh: dinosaurs waiting for the meteor","type":"primary","url":"https://cacm.acm.org/opinion/the-end-of-programming/"},{"author":null,"id":"source-27be554657a19b7483e8","note":"argenti: \"hundreds… might go into the thousands\"; supervised deployment","originalId":"src-ci-cnbc-goldman","published":{"precision":"day","text":"2025-07-11"},"publisher":"cnbc","title":"argenti: hundreds… might go into the thousands","type":"reporting","url":"https://www.cnbc.com/2025/07/11/goldman-sachs-autonomous-coder-pilot-marks-major-ai-milestone.html"},{"author":null,"id":"source-b2840ab1745c1bf7e839","note":"devin desktop (index listing)","originalId":"src-ci-cognition-devin-desktop","published":{"precision":"day","text":"2026-06-02"},"publisher":"cognition.com","title":"devin desktop (index listing)","type":"primary","url":"https://cognition.com/blog/introducing-devin-desktop"},{"author":null,"id":"source-ce62ab0d51a90b48ed1f","note":"devin integration into windsurf ships (index listing)","originalId":"src-ci-cognition-devin-windsurf","published":{"precision":"day","text":"2026-04-15"},"publisher":"cognition.com","title":"devin integration into windsurf ships (index listing)","type":"primary","url":"https://cognition.com/blog/devin-in-windsurf"},{"author":null,"id":"source-2e3148dc6befba8e5d15","note":"government offering (index listing)","originalId":"src-ci-cognition-gov","published":{"precision":"day","text":"2026-02-25"},"publisher":"cognition.com","title":"government offering (index listing)","type":"primary","url":"https://cognition.com/blog/cognition-for-government"},{"author":null,"id":"source-e6d3a1b178c2d535cf22","note":"ai productivity guarantee; credits up to $10m; engineer-hour estimator","originalId":"src-ci-cognition-guarantee","published":{"precision":"day","text":"2026-06-04"},"publisher":"cognition.com","title":"ai productivity guarantee","type":"primary","url":"https://cognition.com/blog/ai-guarantee"},{"author":null,"id":"source-1bbad291f76f701d0539","note":"$1b+ at $26b; $492m run-rate; customers; \"89% of code committed by our engineers is committed by devin\"; \"independent agent lab\"","originalId":"src-ci-cognition-seriesd","published":{"precision":"day","text":"2026-05-27"},"publisher":"cognition.com","title":"$1b+ at $26b","type":"primary","url":"https://cognition.com/blog/series-d"},{"author":null,"id":"source-0db751b75021fab92d04","note":"frontier-size model; 950 tok/s via cerebras; \"stopped reporting swe-bench numbers in 2024\"","originalId":"src-ci-cognition-swe15","published":{"precision":"day","text":"2025-10-29"},"publisher":"cognition.com","title":"frontier-size model","type":"primary","url":"https://cognition.com/blog/swe-1-5"},{"author":null,"id":"source-b5c30eea0d72caf8fa23","note":"definitive agreement; $82m arr; 350+ enterprise customers; 100% participation/vesting","originalId":"src-ci-cognition-windsurf","published":{"precision":"day","text":"2025-07-14"},"publisher":"cognition.com","title":"definitive agreement","type":"primary","url":"https://cognition.com/blog/windsurf"},{"author":null,"id":"source-60f035c084f29a27c33d","note":"\"your ai pair programmer\" — assistive-era framing","originalId":"src-ci-copilot-launch","published":{"precision":"day","text":"2021-06-29"},"publisher":"github.blog","title":"your ai pair programmer — assistive-era framing","type":"primary","url":"https://github.blog/news-insights/product-news/introducing-github-copilot-ai-pair-programmer/"},{"author":null,"id":"source-4f32044a32d07efe7de4","note":"devin 2.0: agent-native ide; $20 plan; search/wiki","originalId":"src-ci-devin2","published":{"precision":"day","text":"2025-04-03"},"publisher":"cognition.com","title":"devin 2.0","type":"primary","url":"https://cognition.com/blog/devin-2"},{"author":null,"id":"source-3626bb632e52d9515da5","note":"99/100 entries self-reported; scaffold moves scores 10–20 pts","originalId":"src-ci-digitalapplied-swebench","published":{"precision":"day","text":"2026-06-16"},"publisher":"digitalapplied.com","title":"99/100 entries self-reported","type":"reference","url":"https://www.digitalapplied.com/blog/swe-bench-verified-june-2026-benchmark-vs-scaffolding-analysis"},{"author":null,"id":"source-c992c96cdd103473ee96","note":"category taxonomy; 2025-05-19 triple launch; codex 3m+ wau feb 2026","originalId":"src-ci-earlyterms-agents","published":{"precision":"year","text":"2026"},"publisher":"earlyterms.com","title":"category taxonomy","type":"reference","url":"https://earlyterms.com/term/coding-agents"},{"author":null,"id":"source-78cec34a9a8099f1b8ff","note":"$60b all-stock anysphere deal; largest venture-backed acquisition","originalId":"src-ci-forbes-spacex-cursor","published":{"precision":"day","text":"2026-06-16"},"publisher":"forbes","title":"$60b all-stock anysphere deal","type":"reporting","url":"https://www.forbes.com/sites/sandycarter/2026/06/16/spacex-buys-cursor-in-largest-startup-acquisition-ever-at-60-billion/"},{"author":null,"id":"source-9c8b8da0c4acd715460c","note":"\"hybrid workforce\"; salary contrast $115–180k","originalId":"src-ci-fortune-goldman","published":{"precision":"day","text":"2025-07-14"},"publisher":"fortune","title":"hybrid workforce","type":"reporting","url":"https://fortune.com/2025/07/14/goldman-sachs-ai-powered-software-engineer-devin-new-employee-increase-productivity-fears-of-job-replacement/"},{"author":null,"id":"source-64ac502a1459aa6bdd73","note":"jules public beta; async cloud-vm agent","originalId":"src-ci-google-jules","published":{"precision":"day","text":"2025-05-20"},"publisher":"blog.google","title":"jules public beta","type":"primary","url":"https://blog.google/innovation-and-ai/models-and-research/google-labs/jules/"},{"author":null,"id":"source-dd9b0f8f0929d484fb44","note":"angellist 14k dashboards 5.2x; nubank details","originalId":"src-ci-headsup-migrations","published":{"precision":"month","text":"2024-12"},"publisher":"headsupai.io","title":"angellist 14k dashboards 5.2x","type":"reporting","url":"https://headsupai.io/updates/cognition-deploys-devin-enterprise-scale-database-monolith-migrations"},{"author":null,"id":"source-b0dafab463859ad5de28","note":"internet of bugs frame-by-frame debunk of upwork demo","originalId":"src-ci-iob-debunk-video","published":{"precision":"month","text":"2024-04"},"publisher":"youtube","title":"internet of bugs frame-by-frame debunk of upwork demo","type":"reporting","url":"https://www.youtube.com/watch?v=tNmgmwEtoWE"},{"author":null,"id":"source-b174a05dfca3de702d20","note":"wang interview: reorgs every 2–3 months; \"windsurf… maybe five or six people now\"","originalId":"src-ci-jeffwang-neoshow","published":{"precision":"year","text":"2026"},"publisher":"youtube","title":"wang interview: reorgs every 2–3 months","type":"reporting","url":"https://www.youtube.com/watch?v=uNQLawiYhQE"},{"author":null,"id":"source-ab68c048f622c5487cd3","note":"codex cloud se agent, codex-1/o3-derived","originalId":"src-ci-openai-codex","published":{"precision":"day","text":"2025-05-16"},"publisher":"openai","title":"codex cloud se agent, codex-1/o3-derived","type":"primary","url":"https://openai.com/index/introducing-codex/"},{"author":null,"id":"source-eef0e6c4854564bf3f68","note":"codex model launch (copilot's engine)","originalId":"src-ci-openai-codex-2021","published":{"precision":"day","text":"2021-08-10"},"publisher":"openai","title":"codex model launch (copilots engine)","type":"primary","url":"https://openai.com/index/openai-codex/"},{"author":null,"id":"source-96df22d52de2b697c293","note":"open-source replication created day after launch; now openhands; ~85k stars","originalId":"src-ci-opendevin-repo","published":{"precision":"day","text":"2024-03-13"},"publisher":"github","title":"open-source replication created day after launch","type":"reporting","url":"https://github.com/opendevin/OpenDevin"},{"author":null,"id":"source-034c7082737b3e1133af","note":"price-collapse-as-admission read","originalId":"src-ci-plainai-devin2","published":{"precision":"month","text":"2025-04"},"publisher":"plainai.aguidetocloud.com","title":"price-collapse-as-admission read","type":"reference","url":"https://plainai.aguidetocloud.com/changes/cognition-devin-2/"},{"author":null,"id":"source-9d13db24b6d98238da40","note":"$500/mo as b2b junior-engineer pricing read","originalId":"src-ci-productgeeks-pricing","published":{"precision":"day","text":"2024-12-13"},"publisher":"productgeeks.com","title":"$500/mo as b2b junior-engineer pricing read","type":"reference","url":"https://www.productgeeks.com/p/devin-ai-pricing"},{"author":null,"id":"source-ebeb343faa4710da4e12","note":"scorecard: ~50% merged-code share at anthropic","originalId":"src-ci-redwood-amodei","published":{"precision":"year","text":"2025"},"publisher":"blog.redwoodresearch.org","title":"scorecard: ~50% merged-code share at anthropic","type":"reference","url":"https://blog.redwoodresearch.org/p/is-90-of-code-at-anthropic-being"},{"author":null,"id":"source-bd86c98a6793c69ef8c4","note":"\"'first ai software engineer' is bad at its job\"","originalId":"src-ci-register-devin-bad","published":{"precision":"day","text":"2025-01-23"},"publisher":"theregister.com","title":"first ai software engineer is bad at its job","type":"reporting","url":"https://www.theregister.com/software/2025/01/23/first-ai-software-engineer-is-bad-at-its-job/"},{"author":null,"id":"source-521b4186c88ec2132123","note":"swe-bench benchmark (princeton et al.)","originalId":"src-ci-swebench-paper","published":{"precision":"month","text":"2023-10"},"publisher":"arxiv.org","title":"swe-bench benchmark (princeton et al.)","type":"reporting","url":"https://arxiv.org/abs/2310.06770"},{"author":null,"id":"source-7fb1ba94cb65b0e117ca","note":"swe-bench+ audit: 32.67% solution leakage; 31.08% weak tests","originalId":"src-ci-swebench-plus","published":{"precision":"month","text":"2024-10"},"publisher":"arxiv.org","title":"swe-bench+ audit: 32.67% solution leakage","type":"reporting","url":"https://arxiv.org/abs/2410.06992"},{"author":null,"id":"source-d0e3c86f1a860f403712","note":"human-validated 500-issue subset","originalId":"src-ci-swebench-verified","published":{"precision":"month","text":"2024-08"},"publisher":"openai","title":"human-validated 500-issue subset","type":"primary","url":"https://openai.com/index/introducing-swe-bench-verified/"},{"author":null,"id":"source-edd94c292c5fa0292bc5","note":"arr $1m(sep'24)→$73m(jun'25); net burn <$20m","originalId":"src-ci-tc-10b","published":{"precision":"day","text":"2025-09-08"},"publisher":"techcrunch","title":"arr $1m(sep24)→$73m(jun25)","type":"reporting","url":"https://techcrunch.com/2025/09/08/cognition-ai-defies-turbulence-with-a-400m-raise-at-10-2b-valuation/"},{"author":null,"id":"source-2f9fa33e2b5ef408d29f","note":"$25b pre/$26b post; revenue $37m→$492m","originalId":"src-ci-tc-26b","published":{"precision":"day","text":"2026-05-27"},"publisher":"techcrunch","title":"$25b pre/$26b post","type":"reporting","url":"https://techcrunch.com/2026/05/27/ai-coding-startup-cognition-raises-1b-at-25b-pre-money-valuation/"},{"author":null,"id":"source-5385b4f4e18becb33bbf","note":"run-rate $492m→~$900m; the information $4–5b eoy-2026 projection; cursor ~$6b pace","originalId":"src-ci-tc-48b","published":{"precision":"day","text":"2026-09-08"},"publisher":"techcrunch","title":"run-rate $492m→~$900m","type":"reporting","url":"https://techcrunch.com/2026/09/08/cognition-hits-48b-valuation-signaling-investors-believe-ai-coding-is-far-from-a-winner-take-all-market/"},{"author":null,"id":"source-8e9720f44994fc564ffa","note":"anthropic cuts claude 3.x first-party capacity","originalId":"src-ci-tc-anthropic-windsurf","published":{"precision":"day","text":"2025-06-03"},"publisher":"techcrunch","title":"anthropic cuts claude 3.x first-party capacity","type":"reporting","url":"https://techcrunch.com/2025/06/03/windsurf-says-anthropic-is-limiting-its-direct-access-to-claude-ai-models/"},{"author":null,"id":"source-02bdccaebbf3168398e0","note":"codex launch coverage","originalId":"src-ci-tc-codex","published":{"precision":"day","text":"2025-05-16"},"publisher":"techcrunch","title":"codex launch coverage","type":"reporting","url":"https://techcrunch.com/2025/05/16/openai-launches-codex-an-ai-coding-agent-in-chatgpt/"},{"author":null,"id":"source-6de360c4af0fe6cbb866","note":"250-person team left behind; friday-call→monday-signing; $100m arr april peak","originalId":"src-ci-tc-cognition-windsurf","published":{"precision":"day","text":"2025-07-14"},"publisher":"techcrunch","title":"250-person team left behind","type":"reporting","url":"https://techcrunch.com/2025/07/14/cognition-maker-of-the-ai-coding-agent-devin-acquires-windsurf/"},{"author":null,"id":"source-f1e10ae59149bbf19428","note":"\"eye-watering\" $500→$20; acu math","originalId":"src-ci-tc-devin2","published":{"precision":"day","text":"2025-04-03"},"publisher":"techcrunch","title":"eye-watering $500→$20","type":"reporting","url":"https://techcrunch.com/2025/04/03/devin-the-viral-coding-ai-agent-gets-a-new-pay-as-you-go-plan/"},{"author":null,"id":"source-1097c6dfa60fa364a20c","note":"windsurf swe-1 family launch","originalId":"src-ci-tc-swe1","published":{"precision":"day","text":"2025-05-15"},"publisher":"techcrunch","title":"windsurf swe-1 family launch","type":"reporting","url":"https://techcrunch.com/2025/05/15/vibe-coding-startup-windsurf-launches-in-house-ai-models/"},{"author":null,"id":"source-866b7c02ca55d3a1d84c","note":"parnas: automatic programming as \"euphemism\" for higher-level language","originalId":"src-ci-wikipedia-autoprog","published":{"precision":"unknown","text":"undated"},"publisher":"wikipedia","title":"parnas: automatic programming as euphemism for higher-level language","type":"reference","url":"https://en.wikipedia.org/wiki/Automatic_programming"},{"author":null,"id":"source-e89e502cd59154c94c6f","note":"~$4b round led by 8vc","originalId":"src-hp-bloomberg-4b","published":{"precision":"day","text":"2025-03-17"},"publisher":"bloomberg","title":"~$4b round led by 8vc","type":"reporting","url":"https://www.bloomberg.com/news/articles/2025-03-18/cognition-ai-hits-4-billion-valuation-in-deal-led-by-lonsdale-s-firm"},{"author":null,"id":"source-9000b86d1e91e6cdee86","note":"ashlee vance launch feature; corp ~2 months old; 10-person staff; $21m; founder roles; yan harvard color","originalId":"src-hp-bloomberg-gold-medalist-coders","published":{"precision":"day","text":"2024-03-12"},"publisher":"bloomberg","title":"ashlee vance launch feature","type":"reporting","url":"https://www.bloomberg.com/news/articles/2024-03-12/cognition-ai-is-a-peter-thiel-backed-coding-assistant"},{"author":null,"id":"source-6d3eaa02c5959695b1f4","note":"\"10 ioi gold medals\"; \"21 of initial 35 had founded a company\"; team employer list","originalId":"src-hp-cognition-about","published":{"precision":"day","text":"2026-09-16"},"publisher":"cognition.com","title":"10 ioi gold medals","type":"primary","url":"https://cognition.com/about"},{"author":null,"id":"source-69d3c314b0e03d32799f","note":"full dated post index used for timeline items (ga, 2.x, swe-x, partnerships, offices, acquisitions)","originalId":"src-hp-cognition-blog-index","published":{"precision":"day","text":"2026-09-16"},"publisher":"cognition.com","title":"full dated post index used for timeline items","type":"primary","url":"https://cognition.com/blog"},{"author":null,"id":"source-6605de8983b4cf7a104f","note":"ga at $500/mo, no seat limits, slack/ide/api","originalId":"src-hp-cognition-ga","published":{"precision":"day","text":"2024-12-10"},"publisher":"cognition.com","title":"ga at $500/mo, no seat limits, slack/ide/api","type":"primary","url":"https://cognition.com/blog/devin-generally-available"},{"author":null,"id":"source-24b7f0d0ba5fb7d9ce77","note":"scott wu launch post; swe-bench 13.86% claim; \"tireless, skilled teammate\"; early-access waitlist","originalId":"src-hp-cognition-introducing-devin","published":{"precision":"day","text":"2024-03-12"},"publisher":"cognition.com","title":"scott wu launch post","type":"primary","url":"https://cognition.com/blog/introducing-devin"},{"author":null,"id":"source-8aee4983e8bd1c160df7","note":"scott wu: >$400m at $10.2b post; arr $1m→$73m; <$20m lifetime burn; customer names; new hires","originalId":"src-hp-cognition-series-c","published":{"precision":"day","text":"2025-09-08"},"publisher":"cognition.com","title":"scott wu: >$400m at $10.2b post","type":"primary","url":"https://cognition.com/blog/funding-growth-and-the-next-frontier-of-ai-coding-agents"},{"author":null,"id":"source-1ecd5bcad1a9647b2911","note":"methodology: 25% subset (570/2294), unassisted, github eval harness","originalId":"src-hp-cognition-swebench-report","published":{"precision":"day","text":"2024-03-15"},"publisher":"cognition.com","title":"methodology: 25% subset (570/2294), unassisted, github eval harness","type":"primary","url":"https://cognition.com/blog/swe-bench-technical-report"},{"author":null,"id":"source-915cc5f7006791fc9a01","note":"long q&a; \"born in 1996\"; minecraft quote; ioi teammates→founders; $445m run rate claim","originalId":"src-hp-colossus-wu-tapes","published":{"precision":"year","text":"2026"},"publisher":"colossus.com","title":"long q&a","type":"interview","url":"https://colossus.com/article/scott-wu-tapes-cognition/"},{"author":null,"id":"source-b7bf0d790557c9ef893c","note":"compiled leadership bios; crypto-pivot claim; yan bio details (thiel, anysphere, deepreason)","originalId":"src-hp-mungomash-leadership","published":{"precision":"day","text":"2026-09-16"},"publisher":"mungomash.com","title":"compiled leadership bios","type":"community","url":"https://mungomash.com/orgs/cognition/leadership/"},{"author":null,"id":"source-0760f59db9d9b7b278b7","note":"series e $2b at $48b; a16z/accel/ff/gc/avenir","originalId":"src-hp-reuters-48b","published":{"precision":"day","text":"2026-09-08"},"publisher":"reuters","title":"series e $2b at $48b","type":"reporting","url":"https://www.reuters.com/technology/cognition-ai-raises-2-billion-48-billion-valuation-2026-09-08/"},{"author":null,"id":"source-b87d95c58f8c4efe90a7","note":"google $2.4b nonexclusive license + hires; no equity","originalId":"src-hp-reuters-google-windsurf","published":{"precision":"day","text":"2025-07-11"},"publisher":"reuters","title":"google $2.4b nonexclusive license + hires","type":"reporting","url":"https://www.reuters.com/business/google-hires-windsurf-ceo-researchers-advance-ai-ambitions-2025-07-11/"},{"author":null,"id":"source-8d6eed5ab8be721ab011","note":"verbatim origin-story post (ti-83 basic, age 9)","originalId":"src-hp-scottwu-linkedin-post","published":{"precision":"day","text":"2026-07-08"},"publisher":"linkedin","title":"verbatim origin-story post (ti-83 basic, age 9)","type":"primary","url":"https://www.linkedin.com/posts/scott-wu-8b94ab96_i-learned-to-program-when-i-was-nine-years-activity-7480660748705157120-EBoW"},{"author":null,"id":"source-52d6624601297ebd17a9","note":"microsoft build partnership; devin on azure","originalId":"src-hp-techzine-msft","published":{"precision":"month","text":"2024-05"},"publisher":"techzine.eu","title":"microsoft build partnership","type":"reporting","url":"https://www.techzine.eu/news/devops/120539/microsoft-brings-independent-application-builder-devin-within-reach/"},{"author":null,"id":"source-c15321fb46b7ed3bf5b8","note":"\"tourist\"; 4009 rating 2024-08-30; tier named for him","originalId":"src-hp-wiki-korotkevich","published":{"precision":"day","text":"2026-09-16"},"publisher":"wikipedia","title":"tourist","type":"reference","url":"https://en.wikipedia.org/wiki/Gennady_Korotkevich"}]}
````