morph is the apply-model company: a 7b model that merges a frontier llm's lazy edit sketch into the original file at a claimed 10,500 tokens per second — the trick cursor pioneered and kept internal, sold as an api. the same yc s23 entity was lingtual, then innkeeper, then autoinfra, then morph. this is the record of that arc — the pivot years, the 217-point hn launch, the 'already dead' counter-thesis, the funding fog, and a company that keeps shipping the other models coding agents need.
41 sourced events · 20 cataloged writings · 40 coverage items · 134 sources · researched by parallel agent lanes
history the company, sourced and dated — lingtual to innkeeper to autoinfra to morph, the hn launch, and the suite broadening. the pivot years yc s23 as lingtual, the innkeeper detour, autoinfra inc. — and the morph l2 collision that poisoned the funding record. the morph rebrand morphllm.com registers in february 2025 — the apply model emerges, the continue pr merges, and the cline pr doesn't. the launch july–september 2025 — the 217-point hn launch, the mcp server, kilo code, and the 10,500-tok/s claim. the model suite october 2025–february 2026 — 'everything is models,' the aws case study, warpgrep, and glance. the subagent era march 2026 on — warpgrep v2's swe-bench claims, the codegen research, and reflexes. people tejas bhakta, the reported co-founder who isn't on the yc page, and the three-person team — who they are, what they wrote, and what they claim. the founders tejas bhakta — tesla autopilot to the motel tech stack to the apply model — and vlad lezhnin, the linkedin co-founder the yc page doesn't list. the writing the morphllm.com blog corpus and the hn launch threads — a founder who argues the thesis in public. the team team size 3, a two-person-team podcast note, and the unverified github handles. interviews and talks the infra pod, start, and composio interviews — where the pivots and the 4×-cerebras claim got said out loud. the philosophy small models as tools, speculation as ip, edge-case tasks for overtrained models — plus the unresolved record. side quests the pursuits around fast apply — the prior pivots, the plugins and integrations, the other models, and the community packages. the founders' quests lingtual, innkeeper, and the founder's essay side-channel. the company's quests the mcp server, the plugin shelf, self-hosting, warpgrep, compact, glance, reflexes — and the integrations that didn't land. the ecosystem the community npm packages and the third-party plugins. coverage what the discourse made of morph — the launch thread, the shelf-life debate, the integration ledger, and the competitor coverage. the category before cursor's instant-apply post, aider's two-llms objection, and the open-source clones — the category before morph. the launches the yc launch, the 217-point hn thread, and the show-hn cadence that never matched it. the integrations continue's apply role, the kilo merge, the cline and roo rejections, databutton's testimonial, and the aws case study. the debate pash's 'already dead,' cognition's counter-thesis, the dreaming.press ledger, and the founder interviews. the competitors relace's $23m series a, cognition's swe-grep, mercury's diffusion models, and the open-weights line. controversies the running debates — vendor benchmarks, shelf life, bundling, the trust incident, the throughput gap, and the funding fog. sentiment eras sentiment phases: magic trick → commoditization → backlash → split verdict → validation → the compression race. the idea the fast-edit-model idea — the lineage morph inherited, what it demonstrated, and whether the frontier compresses the layer. lineage aider's edit-format taxonomy, cursor's speculative edits, the open clones, and the fast-inference era. what morph demonstrated demand, engineering depth, and the economics of a task that's mostly copying. what stayed unproven no shared benchmark, the openrouter gap, conversion friction, and the unverified ledger. the compression race the bet that frontier models stay lazy — for, against, and the nuance. successors and competitors relace, swe-grep, mercury, the open weights, and the harness-layer fixes. open questions the diligence questions the next attempt has to answer. sources the catalog — every source behind every claim, dated and typed. the catalog the full typed catalog — primary, interview, reporting, archive, community. method how this was researched — parallel agent lanes, the evidence model, and how to correct it. the lanes the two parallel research lanes and what the formula proved. the evidence rules the evidence model — confidence, precision, sentiment discipline. known gaps what the research could not verify — recorded, not hidden. corrections how to correct the record. AI-drafted at Ben Guo's direct request and credited to Hraness; every claim links to its cataloged source.
Ask AI about this ChatGPT Claude Perplexity Grok