sources
the catalog — every source behind every claim, dated and typed.
the catalog
every claim on this site carries source ids that resolve here. type marks the evidence role; dates keep the precision the source gave.
- 2026-09aider edit-format leaderboardaider.chataider edit-format leaderboard; diff-format success by model
- 2026-09aider edit-format taxonomy (whole/diff/udiff/patch/search-replace)aider.chataider edit-format taxonomy (whole/diff/udiff/patch/search-replace)
- 2024-05-30aider issue #625 — the two-llms debategithubdebate of cursor-style apply; gauthier "2 llms" latency/failure objections (june 4 comment)
- 2024-10-22upgraded claude 3.5 sonnet + computer useanthropicupgraded claude 3.5 sonnet; text_editor_20241022 tool
- 2024-10-30swe-bench + str_replace_editor detailsanthropicstr_replace_editor details; "highest reliability with string replacement"
- 2026-02-09the infra pod on apple podcastsapple podcastsinfra pod episode w/ tejas; show notes incl. "why cursor removed fast apply"
- 2025-12aws morph case studyaws.amazon.comaws: morph custom inference engine on nvidia h100; 1,000→10,000 tok/s; middleware framing
- 2025cognition × cerebras case studycerebras.aicognition swe-grep deployment on cerebras; fast-inference context
- 2025-06-12don't build multi-agentscognition.ai"today, the edit decision-making and applying are more often done by a single model in one action."
- 2025-10-16swe-grep and swe-grep-minicognition.comswe-grep/swe-grep-mini rl retrieval subagents on cerebras
- 2025-12composio podcast notes on tejaslinkedintejas interview notes: two-person team, gpu bill 8× salary, speed→conversion claim
- 2026-09continue apply-role docsdocs.continue.devcontinue apply role recommends morph fast apply or relace instant apply
- 2026-09mercury coder in continuecontinue.devmercury coder integration notes (speed-positioned code model)
- 2025-03-15tejas-authored apply-role docs incl. morphgithubtejas-authored apply-role docs incl. morph; merged 2025-04-29
- 2024-05-14editing files at 1000 tokens per secondcursororiginal fast-apply post: ~1,000 tok/s, speculative edits, fine-tuned 70b
- 2024-08-23instant apply, wayback copywayback machinearchived copy preserving original text
- 2025-09-16databutton testimonial: 82%→96%linkedindatabutton testimonial: apply accuracy 82%→96% (reported via linkedin)
- 2025-08-26build better agents with morphllmdev.tofly.io morph integration walkthrough
- 2025engineering.fyi mirror of the fly.io tutorialengineering.fyimirror of fly.io morph tutorial
- 2026-02-09morphllm launches (yc launch repost)fondo.comrepublished yc launch copy incl. fortune-50 staff-eng quote and self-host link
- 2026start pod notesfondo.comstart pod: "agents don't need bigger models. they need better tools."
- 2026why ai tools rewrite full files instead of diffsanishgandhi.comwhy tools prefer full rewrites; ~400-line crossover; converge of aider/cursor/morph
- 2026how ai coding tools edit code under the hoodanishgandhi.comtwo-model edit anatomy; the "seam" where models disagree
- 2026-04-02ast edits benchmarkgeometricagi.github.ioedit-format benchmark incl. whole-file/search-replace/patch/ast
- 2025grok cli morph integrationnpmcommunity cli integrating morph apply (reported)
- 2026-04-12the hashline harness benchmarkmywrittenword.comcan bölük hashline harness benchmark; grok code fast 1 6.7%→68.3%
- 2026-09hn algolia comment corpus for morphhn.algolia.comalgolia hn comment corpus for morph mentions incl. 2025-09-15 10k-tok/s self-post
- 2024-06-05hn submission of cursor 1000-tok/s post (10 pts)hacker newshn submission of cursor 1000-tok/s post (10 pts)
- 2024-09-29cursor removed its instant apply posthacker news"cursorai just removed technical blog post about their instant apply algorithm" (2 pts)
- 2026-06-22codegen research on hnhacker news"optimizing models to be fast at codegen" submission (2 pts)
- 2025-10-06everything is model[s] on hnhacker news"everything is model[s]" submission (1 pt)
- 2025-10-29relace a year of fast apply thread (48 pts, 6 comments)hacker newsrelace "a year of fast apply" thread (48 pts, 6 comments)
- 2026-07-13relace compaction post submission (6 pts)hacker newsrelace compaction post submission (6 pts)
- 2025-05-27launch hn: relace (111 pts, 51 comments)hacker newslaunch hn: relace (111 pts, 51 comments)
- 2026-08-20relace were trying to make kimi k3 cheaper (3 pts)hacker newsrelace "we're trying to make kimi k3 cheaper" (3 pts)
- 2025-10-16swe-grep hn thread (97 pts, 31 comments)hacker newsswe-grep hn thread (97 pts, 31 comments)
- 2025mercury api availabilityinceptionlabs.aimercury api availability; speed positioning
- 2026-02-09coding agents need infra to apply code changespodscripts.cotejas interview: pivots, feb-2025 start, 70–80% prior, 30–40% lazy-prompt gain, b2b sdk/mcp, ~4× cerebras claim
- 2025-08kilo code release notesblog.kilo.aikilo release notes incl. v4.71.0 experimental morph fast apply (contributor tejas)
- 2024-10open-source fastapply pipeline + dataset powering softgengithubopen-source fastapply pipeline + dataset powering softgen
- 2024-10fastapply-1.5b model cardhuggingface.cofastapply-1.5b (qwen2.5-coder finetune) weights
- 2025-06mercury diffusion code modelsarxiv.orgmercury diffusion code models; ~1,109/737 tok/s mini/small on h100
- 2026-09apply endpoint specdocs.morphllm.comapply endpoint spec; openai-compatible surface
- 2025-12how morph scaled to 10,000 tok/s on awsmorphllm.commorph's aws writeup: <400 ms 15k-token refactors; binance 50–70% gains (vendor-reported)
- 2026flash compact launchmorphllm.comflash compact launch: 33,000 tok/s verbatim compaction
- 2026-09compact sdk docsdocs.morphllm.commorph-compactor: 33,000 tok/s, 1m ctx, 50–70% reduction
- 2026-09compact product pagemorphllm.comcompact marketing page (spec inconsistencies vs. docs noted)
- 2026-06-23one backbone, many reflexesmorphllm.com"one backbone, many reflexes" multi-head classifier engine
- 2026-09fast apply sdk docsdocs.morphllm.comsdk component docs (spec variants vs. other pages)
- 2026-09warpgrep benchmark pagemorphllm.comvendor warpgrep benchmark page
- 2025-04-14gpt-4.1 launchopenaigpt-4.1 launch; major gains claimed on aider polyglot diff benchmark
- 2026-09morph models on openrouteropenrouter.aimorph models on openrouter; pricing + observed-throughput context (sibling: /morph/morph-v3-large)
- 2025osmosis-apply-1.7b model cardhuggingface.coopen-source apply model (qwen3-1.7b base)
- 2025mcp server for osmosis-applygithubmcp server for osmosis-apply
- 2025rl for code-merging writeuposmosis.airl for code-merging writeup
- 2025-06-20fast apply models are already deadpashpashpash.substack.combear-case post; gauthier quote; tejas "lol, i never said that" reply (sep 10, 2025)
- 2026hash-anchored edit toolsgithubhash-anchored edit tools; cites bölük 16-model harness benchmark
- 2025-10-08relace series a announcementrelace.ai$23m series a led by a16z; matrix + yc participating
- 2026-09mercury coder small on vercel ai gatewayvercelmercury coder small on vercel ai gateway
- 2025-05relace launch yc: cheaper/faster/more-reliable codegen modelsycombinator.comrelace launch yc: cheaper/faster/more-reliable codegen models
- 2026-09relace yc profileycombinator.comyc w23 profile
- 2025-10-28about page footer — © 2025 autoinfra, inc.arquivo.ptfooter: "© 2025 autoinfra, inc."
- 2025-10-11blog index snapshot with post datesarquivo.ptblog index snapshot with post dates
- 2025-10-11homepage snapshot: benchmark tables, hundreds of startups and enterprisesarquivo.pthomepage snapshot: benchmark tables, "hundreds of startups and enterprises"
- undatedtejas bhakta on beamstartbeamstart.comtesla autopilot/energy 3yr bio; motel/$48k website; ucsd
- 2026fast apply benchmark pagemorphllm.comself-reported benchmark table; llm-judged methodology
- 2025-09-15we hit 10,500 tokens/sec on b200morphllm.com10,500 tok/s on b200; custom kernels + speculative execution
- 2025-05-02best practices for building coding agents with morphmorphllm.comagent-builder best practices
- 2026-01-23browser verificationmorphllm.comglance: rl browser-testing agent
- 2026-06optimizing models to be fast at codegenmorphllm.comspeculator economics; 1.93x vs 100x; 33k compaction; serves qwen/glm/deepseek/minimax
- 2025-05-28morph ❤️ continuemorphllm.comcontinue integration announcement
- 2024-02-15how cursor composer and apply workmorphllm.comanalysis of cursor apply internals
- 2025-05-28diffs vs fast applymorphllm.comfast apply vs diffs argument
- 2025-10-10everything is modelsmorphllm.com"everything is models" thesis
- 2024-02-13faster agentsmorphllm.comearliest dated post (predates domain — migrated/backdated)
- 2025-06-15morph gets fastermorphllm.comspeed post; title later showed 10.5k (update ambiguity)
- 2025-10-10llms are bad at being forcedmorphllm.comstructured-output critique
- 2025-05-11self hostingmorphllm.comself-host offering
- 2026-03-31warpgrep v2morphllm.comwarpgrep v2 launch; swe-bench pro claims
- 2025-08-22what is morph fast apply?morphllm.comexplainer post
- undatedmorphllm on caplightcaplight.com$19.5m claim — conflated with morph l2
- undatedlingtual on cb insightscb insightslingtual founded 2023, sf, no-code ai platform
- 2026apply model docs — v3-fast/v3-large specsdocs.morphllm.commodel table (v3-fast 10.5k/96%, v3-large 2.5k/98%), "same concept cursor uses"
- 2026mcp quickstart docsdocs.morphllm.commcp setup targets: cursor, claude code, windsurf, cline, vs code, claude desktop
- 2026-06-26fast-apply models: cursor, morph, and relacedreaming.pressthird-party apply-model landscape; vendor-numbers caveat
- 2023yc s23 batch writeup incl. lingtualevoailabs.medium.comlingtual's original s23 pitch text
- 2026fast apply model pagemorphllm.comfast apply page: 7b/10.5k claims, "used by jetbrains, vercel, webflow", pricing/models
- 2024-03-20fortune coverage of morph l2 roundfortunefortune coverage of morph l2 round
- undatedmorph on fundraisingfoxfundraisingfox.com$19m seed claim — conflated with morph l2
- 2025proposed cline integration — closed unmergedgithubproposed cline integration — closed unmerged
- 2025fast_edit_file prompts supporting morph/relacegithubfast_edit_file prompts supporting morph/relace
- 2025-07-30experimental morph fast apply mergedgithubexperimental morph fast apply merged
- 2025–2026opencode-morph-plugingithubmit; fast apply + warpgrep + flashcompact plugin
- 2023-07-10the morphllm github orggithubofficial org, ~24 public repos
- 2025proposed roo integration — closed without implementationgithubproposed roo integration — closed without implementation
- 2025-09-1510,500 tokens/sec per request on nvidia hardwarehacker news"10,500 tokens/sec per request on nvidia hardware"
- 2026-01-24agentic browser-testing videos in prshacker newsagentic browser-testing videos in prs
- 2026-02-04glance launch: pr testing videoshacker newsglance launch: pr testing videos
- 2025-07-07launch hn: apply ai code edits at 4,500 tokens/sechacker newslaunch hn + tejas comments (all quotes)
- 2025-07-18mcp for fast ai filesystem editshacker news"mcp for fast ai filesystem edits"
- 2025-07-21we're already post-agihacker newsessay "we're already post-agi"
- 2026-06-30reflexes launch: multi-head trace classifiershacker newsreflexes launch: multi-head trace classifiers
- 2026-01-23learning to test code changes with rlhacker news"learning to test code changes with rl"
- 2025-07-14how intelligence self organizeshacker newsessay "how intelligence self organizes"
- 2026-01-20warpgrep on twitter algorithm repohacker newswarpgrep on twitter algorithm repo
- 2025-12-09warpgrep show hnhacker newswarpgrep show hn
- undatedmorphllm linkedin company pagelinkedincompany page; $500k pre-seed 2023-06-01 listing
- undatedtejas bhakta on linkedinlinkedinfounder/ceo, ex-tesla ai
- undatedvlad lezhnin on linkedinlinkedin"co-founder & cto @ morph" claim; ex-spotify, 15+ yrs
- 2026-09morphllm.com homepagemorphllm.comcurrent homepage: positioning, product suite
- undatedmorph on neuronfeedneuronfeed.com$500k jan 2025, yc + transpose platform
- undated@f97/opencode-morph-fast-applynpmcommunity third-party plugin package
- undatedmorph models on openrouteropenrouter.aimodel availability; morph-v2 created 2025-06-26
- undatedv3-fast vs relace apply 3 on openrouteropenrouter.aiv3-fast 81,920 ctx, $0.80/$1.20 vs relace apply 3
- undatedmorphllm on platoseedplatoseed.comaka names; "launched on yc may 2025"
- 2025the infra pod (apple podcasts)apple podcasts"coding agents need infra to apply code changes"
- 2026-03-27the start podcast episodescripod.comstart ep: transcript/quotes; apply ~10k, compact 33k tok/s
- 2024-03-20morph l2 announces $19m seed (different company)prnewswiremorph l2's $19m seed (different company)
- 2026fastapply product spec (.md)morphllm.commachine-readable product spec: 7b/14b, custom engine
- 2026warpgrep product pagemorphllm.com14b, ~900 tok/s b200, partner logos
- undatedmorph on reforgersreforgers.comaka autoinfra/lingtual/innkeeper; founded 2023
- 2026relace apply 3 — a year of fast applyrelace.aicompetitor post "a year of fast apply"; apply 3 10k+ tok/s 256k
- 2026-09-16sec edgar search — autoinfra/morphllmseczero filings for autoinfra/morphllm
- undatedsecond spotify episodeopen.spotify.comsecond interview episode (details unverified)
- undatedtejas bhakta's personal sitetejasbhakta.com"founder/ceo @morphllm.com (yc s23)"; prior projects
- 2023-07-26lingtual on toolifytoolify.ailingtual directory listing; founders@lingtual.com
- undatedmorphllm on tracxntracxnlists company as unfunded
- 2023-08-04lingtual.com whoislookup.icann.orgdomain creation date
- 2025-02-06morphllm.com whoislookup.icann.orgdomain creation date (namecheap)
- 2025-05launch yc: morph — the other modelsycombinator.comlaunch: apply 5k+ tok/s, embed+rerank, 8x-fewer-errors claim
- 2026-01launch yc: warpgrepycombinator.comwarpgrep v1 launch metrics
- 2026-03-31launch yc: warpgrep v2ycombinator.comwarpgrep v2; #1 swe-bench pro claim
- undatedlingtual on the yc directoryycombinator.comsame entity's s23 profile; aka autoinfra/innkeeper
- 2026-09morph on the yc directoryycombinator.coms23, sf, founded 2025, team 3, tom blomfield, founder list
AI-drafted at Ben Guo's direct request and credited to Hraness; every claim links to its cataloged source.