coverage
what the discourse made of the lab — the hacker news threads, the community press, and the outside analysts.
the manifesto tests
- 2018-11-21slow software on hnthe latency essay lands — 438 points, 260 comments; the thread treats 'why does software feel slow' as a real research question.
- 2019-03-23end-user programming on hnthe end-user programming essay draws a mixed, qualified-positive 105-point thread — the same skepticism the essay itself catalogs arrives in the comments.
- 2019-05-02local-first software on hn (launch)566 points, 190 comments — the launch thread is enthusiastic but technically skeptical, and it surfaces the prior-art objection: offline-first predates the coinage by six years.
- 2019-05-03muse on hn137 points — a warm reception for the studio-for-ideas framing; 'the tl;dr is that muse is a studio for your ideas.'
- 2019-06-23capstone on hncapstone draws 254 points, 41 comments — a warm thread with curiosity about the lab model itself.
- 2019-11-20the morning paper on local-firstadrian colyer's summary — 'the seven ideals of local-first software' — drives the second-biggest thread in the dossier: 748 points, 239 comments.
- 2020-04-13malleable systems manifesto on hnthe malleable systems collective's manifesto — the predecessor community the lab's 2025 essay converged with — draws 336 points, 204 comments.
- 2022-11-01potluck on hnpotluck at 201 points — 'with web tech! this is revolutionary stuff' beside the open question of backwards compatibility on the user's own data shapes.
- 2022-11-30inkbase on hnthe lab's biggest-ever hn artifact: 675 points, 83 comments — 'this is the most exciting ink & switch prototype yet'; praise for the team 'that brought us muse.'
- 2023-03-09upwelling on hna quiet mixed 37-point thread — creative privacy praised, the demo's presentation critiqued.
- 2025-02-04ambsheets on hnambsheets — the scenario-exploration spreadsheet — draws 178 points; the track-era essays keep finding the front page.
- 2025-05-22sketchy calendar on hnsketchy calendar at 156 points — the programmable-ink line still drawing attention three years after inkbase.
- 2025-06-10malleable software on hnthe second flagship essay: 294 points, 114 comments (plus a second 113-point thread that august) — 'anyone can adapt their tools to their needs' praised as an answer to lock-in, with the hard-won skepticism about whether users want the job.
- 2025-07-05local-first re-posted (2025)the 2019 essay's biggest thread arrives six years late — 883 points, 297 comments on a re-post; the loyalist line: 'it brings back a much healthier incentive structure.'
the infrastructure argument
- 2020-10-06cambria on hncambria's lenses draw a mixed 101-point thread — schema evolution framed as 'one of the big problems that hasn't been solved' even by sympathetic readers.
- 2021-11-23peritext on hn76 points on the rich-text crdt — a smaller, more technical thread than the manifestos; the audience is now the builders.
- 2020-07-11crdts: the hard parts on hnkleppmann's own crdt hard-parts talk draws 404 points and a genuinely mixed thread — 'we're still in early days; don't use this in prod if you can help it' is advice from the community itself.
- 2020-09-28the ot veteran convertsjoseph gentle — sharejs and google wave veteran, ot-side authority — publishes 'i was wrong. crdts are the future': 1,181 points, 312 comments, the single largest hn artifact in this dossier.
- 2022-02-21automerge on hn (2022)the automerge repo thread — 274 points, 69 comments — the library already drawing a builder audience a year before the 2.0 rewrite.
- 2022-04merge what you can, fork what you can't'merge what you can, fork what you can't' — schiefer, litt (then an i&s researcher), and jackson argue auto-merge conflicts with data-integrity invariants and propose exposed forking histories. internal academic critique of forced convergence.
- 2023-01-30automerge 2.0 on hnautomerge 2.0: 717 points — celebration plus shipping notes: 'the api is much better now,' with the interoperability question attached ('what is its story with yjs?').
- 2023-09-26weidner's crdt benchmarksmatt weidner's benchmark series measures automerge's memory and time costs against yjs and collabs-list — the measurement that made the performance critique concrete. he later argues parts of collaborative text need no crdt machinery at all.
- 2023-11-08the benchmark spaton the automerge-repo thread (219 pts), heather miller — 'ouch peter. massive offense taken' — objects to van hardenberg calling the many-concurrent-editors benchmark 'artificial'; pvh apologizes unreservedly, defends the <10ms keypress-to-paint target, and concedes the systems optimize for different workloads.
- 2024-07-16automerge on hn (2024)the automerge site thread — 268 points, 58 comments — a builder audience asking production questions: storage, auth, sync servers.
- 2025-08-03automerge 3.0 on hnautomerge 3.0 at 353 points — the columnar-rewrite memory numbers land, and the thread now asks production questions instead of definitional ones.
- 2025-10-02keyhive on hnkeyhive — local-first access control — draws 167 points; the 2019 essay's hardest open problem getting its own thread.
the community coverage
- 2023-02-28the lofi community's own historyjames arthur's retrospective: the lofi community began as his localfirstweb.dev page, first meetup 2023-02-28 with van hardenberg billed as 'the father of local-first' — a community epithet, not a title. ~700 members by may 2023; lofi/40 scheduled for 2026-08.
- 2024-02-27local-first conf on hnthe local-first conf announcement draws 3 points — 'the world's first local-first conference' had real-world traction (berlin, may 2024, lab members organizing and attending) that the scoreboard doesn't show.
- 2024-06-25the prokopov debatenikita prokopov's 'local, first, forever' draws the canonical debate thread — 689 points, 280 comments: 'local first… almost always suck or isn't that important' against 'a much healthier incentive structure.' prokopov's own essay punctures purity: he still needs a sync server, so he builds on file sync.
- 2023-09-20the adjacent show-hnsthe label's pull measured in adjacent show-hns: electricsql 617pts, localgpt 331pts, anytype 184pts, livestore 156pts (timed to lfc 2025), colanode 147pts, fireproof 99pts, zero 99pts — and the lab's own darn at 3pts, a reminder the word travels further than any single artifact.
- 2026-04-18jazz credits the labjazz v2's public-alpha announcement credits ink & switch research 'especially… cambria, peritext, and keyhive' alongside automerge, yjs, loro, evolu, triplit — the ecosystem naming its lineage.
the mainstream edge
- 2019-12-06the wired profileklint finley's profile — 'you can have collaborative software that's wary of the cloud' — the only major mainstream-press piece on the lab found; sympathetic, hedged by stanford's matei zaharia questioning whether local-first apps can be compelling enough to pull users over.
the lab itself
- 2015the directory entriesthe directory layer: linkedin and cb insights list 'founded 2015'; cb insights adds a london hq and a blueyard capital 'unattributed vc' entry that no lab source confirms. the lab registers in the industry's catalogues without being part of its economy.
- 2025-09the notion pipelinethe alumni magnet: notion announces max schöning ('software should bend'), then geoffrey litt departs 'to work on an incredible team' — mary rose cook already there. coverage of the lab as a talent source.
- 2026endless on backstitchthe funder's own account of the backstitch partnership — 'real-time collaboration in godot is changing the game-design classroom' — the lab covered through its partners' lens.
the independent reads
- 2021-03-15matt webb on invention labsmatt webb's invention-labs survey — 'i saw a pitch deck for ink & switch in march 2015' — names the lab among a handful of 'indie labs,' with a clear-eyed read on what the model is and isn't. the founding-year gap between that deck and 'turns ten in 2026' is preserved, not reconciled.
- 2024joodaloop's reservations'reservations about local-first' — the seven ideals collide with storage limits (localstorage 5mb), schema migrations, partial sync, and permissions; conflict resolution 'gets a comically disproportionate amount of attention' as 'irreducible complexity' only app developers can resolve.
- 2024the crdt-scene self-critiquecrdt community session notes (aljoscha meyer): 'crdts can only express conflict-free settings, but many settings aren't conflict-free… we must not equate p2p or local-first with a crdt-only mentality'; crdts 'remove agency in how to handle conflicts.'
- 2025-09why local-first apps haven't become popular'why local-first apps haven't become popular' — sync complexity blamed for the gap between vocabulary victory and shipping software.
- 2026-01-01the definitional critiquesteven deobald's 'localfirst: you keep using that word' formalizes the definitional dispute — the term is 'a direct attack on vendor lock-in,' not a binary label, and kleppmann's own 2024 talk micro-redefined it as 'multi-device.'
- 2026the earned skepticthe earned-skeptic practitioner take: 'i spent the first year dismissing local-first as academic' — then built with it anyway; the movement's most persuasive coverage is written by its doubters.
- 2026the lineage essaya third-party lineage essay maps malleable software onto the kay canon — 'a modern-day counterpart to the software environments pioneered by alan kay… at xerox parc.' the outside analysts now write the lab's intellectual history for it.
controversies
- 2019–local-first vs offline-firstthe naming objection — 'offline-first' was coined by alex feyerke and the hoodie team in 2013, six years before the essay; on lobsters, a commenter reported the scuttlebutt community viewing the paper as 'an independent reinvention of offline-first philosophy by folks unaware of ssb.'the paper does distinguish its contribution — ownership plus collaboration across seven ideals — but the rename-for-old-idea critique recurs; deobald's 2026 essay is still litigating the definition.
- 2023-11the benchmarks issuethe benchmark spat — on the automerge-repo hn thread, heather miller ('ouch peter. massive offense taken') objected to van hardenberg calling weidner's many-concurrent-editors benchmark 'artificial'; pvh apologized unreservedly and conceded the systems optimize for different workloads. the most heated public exchange located.resolved in-thread — a disagreement about measurement, not a feud. no lasting rift appears in the record.
- 2020–the crdt performance critiquethe crdt performance critique — 'crdts are easy to implement badly… simple implementations often have terrible performance' is kleppmann's own formulation (2020); weidner's benchmarks made it measurable; the community version is 'still early days, don't use this in prod.'the strongest criticism the lab absorbed — and the one it spent three years rewriting a core to answer: automerge 3.0's moby-dick figure (700mb → 1.3mb) is engineering as rebuttal.
- 2019–the server-authoritative objectionthe server-authoritative objection — hn commenters repeatedly cite haverbeke's 'collaborative editing in prosemirror' (2015): 'you can save oh so much complexity by introducing a central point.' a 2020 commenter accused research groups of 'worship[ing] a particular technology for some imaginary practical applications.'a standing engineering objection, partially answered by the essay's own scope — 'we do not claim that crdts are the only possible solution.'
- 2022–the forced-convergence critiquethe forced-convergence critique — schiefer, litt, and jackson's papoc '22 paper ('merge what you can, fork what you can't') argues auto-merge conflicts with data-integrity invariants; the worm-blossom session notes add that crdts 'remove agency in how to handle conflicts.' a critique from inside the lab's own co-authors.litt co-authored the critique while working at the lab — the objections are internal, documented, and unresolved.
- 2024–the definition driftdefinition drift — deobald (2026): 'local-first' is 'a direct attack on vendor lock-in,' not a binary label, and kleppmann himself micro-redefined it as 'multi-device' in a 2024 talk; the nokia blog's earned skeptic 'spent the first year dismissing local-first as academic.' the word now covers sqlite sync, optimistic ui, and ordinary caching.the lab's own usage stays narrow; the drift is the price of the word winning.
sentiment eras
- 2018–2019the manifesto moment · mixedthe manifesto reception — slow software and the local-first essay are received as agenda-setting; wired's december 2019 profile is friendly but hedged with the zaharia question.slow software 438pts; the essay 566pts at launch and 748pts via the morning paper; the offline-first objection is born in the launch thread.
- 2020–2021the hard-parts scrutiny · mixedthe hard-parts scrutiny — kleppmann's own talk names the implementation traps, joseph gentle's conversion essay becomes the biggest artifact in the dossier, and the repost threads start carrying the crdt-worship objection.hard parts 404pts with 'don't use this in prod' as community advice; 'i was wrong. crdts are the future' at 1,181pts; the 2020 reposts carry the haverbeke central-server link.
- 2022–2023the ecosystem proof · positiveecosystem proof — attention moves from the idea to the tooling: automerge 2.0, the lofi community forming, the first serious performance measurement (weidner, then the miller exchange), and adjacent show-hns adopting the label.automerge 2.0 at 717pts; automerge-repo thread 219pts with the benchmark spat; electricsql's 617pt show hn; inkbase at 675pts.
- 2024the pragmatist turn · mixedthe pragmatist turn — lfc berlin convenes the movement in person while prokopov's diy-sync essay reframes the debate from local-first purity to 'sync engines and what actually ships.'lfc 2024 runs in may; 'local, first, forever' draws 689pts of skeptics vs loyalists; electric's july pivot and the december zero launch mark the sync-engine era.
- 2025the vocabulary victory · positiveconsolidation and vocabulary victory — automerge 3.0, the essay's biggest-ever thread six years on, the malleable essay, keyhive; 'local-first' is now ambient product vocabulary.automerge 3.0 at 353pts; the 2019 essay re-posts to 883pts — 'probably the most influential essay of the last ten years' per the thread; keyhive at 167pts.
- 2026the agent-era repositioning · favorablethe agent-era repositioning — tenfold's ten-year mark, lfc lab day, aria/gaios, subduction; the ecosystem builds agent-facing automerge tools (braid, mission-control, darn) and wiggins floats local models on infoq.localgpt's 331pt show hn vs the lab's own darn at 3pts; dispatch 018 logs the godotcon response; the coverage is now mostly the lab's own channels plus community media.
AI-drafted at Ben Guo's direct request and credited to Hraness; every claim links to its cataloged source.