# Jev Wiki > Unofficial, agent-maintained knowledge base about Jev, TypeSafe AI's System One model, and everything an LLM agent needs to build software with it. Plain Markdown with YAML frontmatter; wikilinks [[dir/slug]] resolve to https://jevwiki.ai/raw/wiki/dir/slug.md. Covers: Jev (TypeSafe AI's System One model): concepts, the HTTP API (/v1/systemone, /v1/models), OpenAPI schemas, models/pricing/rate limits, Python SDK (typesafe-sdk) and JavaScript SDK (@typesafe-ai/sdk) references, architectural patterns, all 18 official cookbooks, guides, the agent skill, the system-one-adapter, company/team/blog/press, legal terms. Not covered: Anything after the current-as-of date below, TypeSafe's private roadmap, benchmark claims beyond what TypeSafe and cited press publish, other vendors' structured-output products in depth. Use the live docs at https://docs.typesafe.ai/llms.txt for anything newer. Current as of: 2026-09-17 (docs snapshot), jev-1.13.0, typesafe-sdk 0.6.0, @typesafe-ai/sdk 0.6.0 ## Route by task (read these, skip the rest) Every entry below shows an estimated token size so you can budget before fetching. Do NOT load llms-full.txt for a single task; it is the whole corpus. - **Build something with Jev:** [guides/agent-integration-playbook](https://jevwiki.ai/raw/wiki/guides/agent-integration-playbook.md) (~7.0k tok) → [reference/http-api](https://jevwiki.ai/raw/wiki/reference/http-api.md) (~5.0k tok) → [concepts/jaggedness-jev-1-13](https://jevwiki.ai/raw/wiki/concepts/jaggedness-jev-1-13.md) (~3.3k tok) — then one SDK page (reference/python-sdk or reference/javascript-sdk) and the closest cookbook - **Assess a project: "could Jev help here?":** [ideas/consult](https://jevwiki.ai/raw/wiki/ideas/consult.md) (~2.0k tok) → [ideas/patterns](https://jevwiki.ai/raw/wiki/ideas/patterns.md) (~2.4k tok) — then ONE domain page the index points to (ideas/patterns-agents, ideas/patterns-interface, ideas/patterns-data or ideas/patterns-business); open ideas/field-reports or ideas/community-repos only when a pattern points you there - **Look up an exact contract, limit or price:** [reference/http-api](https://jevwiki.ai/raw/wiki/reference/http-api.md) (~5.0k tok) → [reference/models-and-pricing](https://jevwiki.ai/raw/wiki/reference/models-and-pricing.md) (~2.7k tok) → [reference/rate-limits-and-errors](https://jevwiki.ai/raw/wiki/reference/rate-limits-and-errors.md) (~3.6k tok) - **Write good instructions and criteria:** [guides/writing-instructions-and-criteria](https://jevwiki.ai/raw/wiki/guides/writing-instructions-and-criteria.md) (~4.6k tok) → [guides/choosing-a-primitive](https://jevwiki.ai/raw/wiki/guides/choosing-a-primitive.md) (~3.4k tok) Tiers: pages under ideas/ are community-sourced (source_tier: community, confidence capped at medium). When they conflict with reference/ or concepts/ pages, the reference pages win. Start with the agent playbook: https://jevwiki.ai/raw/wiki/guides/agent-integration-playbook.md — then the reference pages. Every page is served raw at /raw/wiki/.md and as HTML at /wiki/.md. Full corpus in one fetch: https://jevwiki.ai/llms-full.txt. Machine registry: https://jevwiki.ai/index.json. - [Playbook for LLM agents building with Jev — START HERE](https://jevwiki.ai/raw/wiki/guides/agent-integration-playbook.md) - [README: how to use this wiki](https://jevwiki.ai/raw/README.md) - [Schema and maintenance rules](https://jevwiki.ai/raw/CLAUDE.md) - [Jev Wiki — Master Index](https://jevwiki.ai/raw/wiki/index.md): Catalog of every page in the wiki, one line each, grouped by section. Generated by scripts/build.mjs. - [Activity log](https://jevwiki.ai/raw/wiki/log.md): Append-only log of ingestion, lint, and refresh activity. ## Concepts - [Structured instructions, options, levels, criteria](https://jevwiki.ai/raw/wiki/concepts/advanced-structure.md) (~3.4k tok): instructions, Choice option descriptions, Score levels and Noul true/false all accept JSON: string, object, array or null (EntryType). - [Choice questions](https://jevwiki.ai/raw/wiki/concepts/choice.md) (~4.6k tok): Choice picks one option from a fixed set: send type/instructions/criteria, get back choice, probabilities over every option, and confidence. - [Confidence vs probability](https://jevwiki.ai/raw/wiki/concepts/confidence.md) (~2.3k tok): confidence is a 0-1 statistic derived from an answer's probabilities distribution; Choice and Score carry it, Noul does not. - [How to build software with System One](https://jevwiki.ai/raw/wiki/concepts/how-to-build.md) (~6.3k tok): Build a normal software workflow and insert System One only where AI is needed: narrow typed questions, composed in code. - [Jev 1.13 jaggedness: known failure modes](https://jevwiki.ai/raw/wiki/concepts/jaggedness-jev-1-13.md) (~3.3k tok): Nine documented failure modes of jev-1.13 — literal reading, math, dates, indirection, big states, adversarial text, contradictions, invariants, generation — with the mitigation for each. - [AI primer: why calibrated decision models](https://jevwiki.ai/raw/wiki/concepts/machine-learning-primer.md) (~1.7k tok): Why TypeSafe post-trains with RLCD for calibrated decisions instead of RLHF/RLVR for generated text. - [Noul (yes/no) questions](https://jevwiki.ai/raw/wiki/concepts/noul.md) (~2.3k tok): Noul asks one yes/no question and returns a single number, the probability that the answer is yes; criteria are optional true/false clarifications. - [Primitives: Choice, Score, Noul](https://jevwiki.ai/raw/wiki/concepts/primitives.md) (~3.0k tok): The three Jev question types (Choice, Score, Noul), the typed answers they return, how to pick one, and how to batch many questions in one request. - [Score questions](https://jevwiki.ai/raw/wiki/concepts/score.md) (~5.1k tok): Score rates a state against an ordered array of described levels and returns a probability-weighted position, the legend, per-level probabilities, and confidence. - [State: what you send Jev](https://jevwiki.ai/raw/wiki/concepts/state.md) (~1.7k tok): State is the content Jev evaluates: a string, JSON object, or array of text, shared by every question in one request. - [System One Models](https://jevwiki.ai/raw/wiki/concepts/system-one.md) (~2.8k tok): System One models make fast, typed, calibrated decisions software can consume directly; Jev is TypeSafe's first one. - [Use-case map by industry](https://jevwiki.ai/raw/wiki/concepts/use-case-map.md) (~3.2k tok): Scannable map of where Jev fits: five capability categories, eighteen industry/function areas, and ten decision shapes. - [Workflow evals: how TypeSafe measures Jev](https://jevwiki.ai/raw/wiki/concepts/workflow-evals.md) (~4.0k tok): TypeSafe's four workflow evals: same harness for every model, reference labels from GPT-6 Astra + Claude Fable 5.1, source of 193.6x/444.6x. ## Reference - [The typesafe-ai agent skill and Claude Code plugin](https://jevwiki.ai/raw/wiki/reference/agent-skill.md) (~5.2k tok): Install, update, and invoke the typesafe-ai agent skill (Claude Code plugin typesafe 0.5.7, MIT), plus a faithful digest of the guidance SKILL.md gives agents. - [TYPESAFE_* environment variables across SDKs](https://jevwiki.ai/raw/wiki/reference/environment-variables.md) (~2.8k tok): Exactly four TYPESAFE_* environment variables are read by the SDKs: API_KEY, BASE_URL, DEFAULT_MODEL, LOG_LEVEL — both SDKs read all four. - [HTTP API: POST /v1/systemone and GET /v1/models](https://jevwiki.ai/raw/wiki/reference/http-api.md) (~5.0k tok): Complete wire contract for POST https://api.typesafe.ai/v1/systemone and GET /v1/models: auth, request body, question and answer shapes, usage, errors. - [JavaScript/TypeScript SDK: install, client, choice/score/noul](https://jevwiki.ai/raw/wiki/reference/javascript-sdk.md) (~6.9k tok): @typesafe-ai/sdk 0.6.0: install, TypeSafeClient config and defaults, systemOne(), choice/score/noul builders, APIPromise, models.list(), logging. - [JavaScript SDK changelog](https://jevwiki.ai/raw/wiki/reference/javascript-sdk-changelog.md) (~1.8k tok): @typesafe-ai/sdk releases: 0.5.7 initial public release, 0.6.0 (2026-09-15) makes Score.criteria an ordered sequence — the one breaking change. - [JavaScript SDK error classes, RetryPolicy, RequestOptions](https://jevwiki.ai/raw/wiki/reference/javascript-sdk-errors.md) (~5.2k tok): Error hierarchy of @typesafe-ai/sdk 0.6.0, status-to-class mapping, RetryPolicy defaults (2 retries, 500ms/5s backoff, 0.25 jitter) and RequestOptions. - [JavaScript SDK interfaces and type aliases](https://jevwiki.ai/raw/wiki/reference/javascript-sdk-types.md) (~6.0k tok): Every interface, type alias and exported variable in @typesafe-ai/sdk 0.6.0, with per-property type/required/default tables and generics examples. - [Legal: MCA, DPA, privacy, data retention](https://jevwiki.ai/raw/wiki/reference/legal-and-data.md) (~5.1k tok): Developer-facing summary of TypeSafe's MCA, DPA, and Privacy Policy: no training on Input, telemetry carve-out, retention, sub-processors, acceptable use, no uptime SLA. - [Migrating from /preview/evaluation to /v1/systemone](https://jevwiki.ai/raw/wiki/reference/migrating-to-v1.md) (~3.7k tok): Every delta between the preview evaluation endpoint and the stable v1 API, with before/after examples, plus the typesafe-client to typesafe-sdk rename table. - [Models, aliases, pricing, rate limits, context](https://jevwiki.ai/raw/wiki/reference/models-and-pricing.md) (~2.7k tok): Jev 1.13 model ID, aliases, $42/Btok pricing with free output tokens, 250k tok/s and 1,200 rpm limits, 64k/32k context rule, and text-only input. - [OpenAPI component schemas](https://jevwiki.ai/raw/wiki/reference/openapi-schemas.md) (~4.8k tok): Every component schema in TypeSafe's OpenAPI 3.1.0 document (info version 0.2.0): properties, types, requiredness, constraints, and examples. - [Python SDK: install, clients, system_one()](https://jevwiki.ai/raw/wiki/reference/python-sdk.md) (~5.7k tok): typesafe-sdk 0.6.0: install, TypeSafeClient/AsyncTypeSafeClient constructor params, system_one() kwargs, models resource, env vars, logging, and the full export list. - [Python SDK changelog](https://jevwiki.ai/raw/wiki/reference/python-sdk-changelog.md) (~1.8k tok): typesafe-sdk release history: 0.0.1a0 placeholder, 0.5.7 initial public release (2026-09-14), 0.6.0 (2026-09-15) with the Score.criteria breaking change. - [Python SDK question types (Noul, Choice, Score)](https://jevwiki.ai/raw/wiki/reference/python-sdk-questions.md) (~3.1k tok): Every field of Noul, Choice and Score in typesafe-sdk 0.6.0, their TypedDict equivalents, JSONContent typing, client-side validation, and the 0.6.0 Score.criteria breaking change. - [Python SDK responses, answers, usage, models](https://jevwiki.ai/raw/wiki/reference/python-sdk-responses.md) (~3.4k tok): SystemOneResponse fields and the .nouls/.choices/.scores views, every answer attribute, Usage, request_id and raw_http_response, plus ListModelsResponse and ModelMetadata. - [Python SDK retries, exceptions, constants](https://jevwiki.ai/raw/wiki/reference/python-sdk-retries-errors.md) (~4.7k tok): RetryPolicy fields and defaults, which statuses and exceptions are retried, backoff and Retry-After handling, the full exception hierarchy with attributes, timeouts, and public constants. - [HTTP status codes, rate limits, retry semantics](https://jevwiki.ai/raw/wiki/reference/rate-limits-and-errors.md) (~3.6k tok): One table mapping every HTTP status to its meaning, Python exception, JS error class, retryability, and the recommended client action. - [system-one-adapter: LLM-backed drop-in for TypeSafeClient](https://jevwiki.ai/raw/wiki/reference/system-one-adapter.md) (~5.0k tok): system-one-adapter answers typesafe_sdk questions with an OpenAI or Anthropic LLM instead of Jev, so you can compare cost, speed, and intelligence on the same code. ## Patterns - [Composite scoring](https://jevwiki.ai/raw/wiki/patterns/composite-scoring.md) (~2.1k tok): Break a ranking judgment into independent Score dimensions, normalize each to 0–1, and combine them with weights your code owns and can retune. - [Confidence-gated routing](https://jevwiki.ai/raw/wiki/patterns/confidence-routing.md) (~2.2k tok): Treat confidence as a second decision axis: one floor below which nothing is automated, then a per-action threshold sized to the consequences of being wrong. - [Speculative fan-out](https://jevwiki.ai/raw/wiki/patterns/fan-out.md) (~2.2k tok): Put every question your decision tree could need into one call, including speculative ones, and let code discard the irrelevant answers. - [Intent routing](https://jevwiki.ai/raw/wiki/patterns/intent-routing.md) (~2.6k tok): Classify the request with one cheap Jev call, then invoke the cheapest adequate handler: deterministic code, a specialist LLM, or a human. - [Patterns overview](https://jevwiki.ai/raw/wiki/patterns/overview.md) (~1.8k tok): Catalog of TypeSafe's four architectural patterns — fan-out, confidence routing, composite scoring, intent routing — and when to reach for each. ## Cookbooks - [Cookbook: Structure recovery (autoformat)](https://jevwiki.ai/raw/wiki/cookbooks/autoformat.md) (~4.5k tok): Rebuild Markdown from de-formatted plain text in two requests: Nouls stitch hard-wrapped lines, then Choices classify each block; code renders every character from the input. - [Cookbook: Autoresearch feature discovery](https://jevwiki.ai/raw/wiki/cookbooks/autoresearch-feature-discovery.md) (~4.7k tok): An LLM proposes Jev questions, Jev answers them per row to build numeric columns, CatBoost trains and reports back: 38 questions after five rounds reach 1.77 held-out RMSE on wine scores. - [Cookbook: Double-checking citations](https://jevwiki.ai/raw/wiki/cookbooks/citation-check.md) (~2.8k tok): One Choice question decides whether a quote's surrounding section supports, contradicts, or says nothing about an LLM's claim; a string match catches fabricated quotes first. - [Cookbook: Classification using confidence](https://jevwiki.ai/raw/wiki/cookbooks/classification-using-confidence.md) (~2.9k tok): Classify SEC filings into 75 SIC industry groups with one Choice; when confidence < 0.9, report the broader division instead — 40% correct becomes 70%, at no extra call. - [Cookbook: Classifying RAG passages](https://jevwiki.ai/raw/wiki/cookbooks/classifying-rag-passages.md) (~3.6k tok): Four Noul questions per retrieved query-passage pair, routed by ordered thresholds, decide what reaches the generator as evidence, as conflict, or not at all. - [Cookbook: Self-consistency — choices](https://jevwiki.ai/raw/wiki/cookbooks/consistency-choice.md) (~4.3k tok): Runs an 8-Choice moderation rubric 15 times across Jev and six LLM conditions; a 0.60 top-probability floor lifts Jev's decision agreement from 90.8% to 99.2% at 74.2% automation. - [Cookbook: Self-consistency — nouls](https://jevwiki.ai/raw/wiki/cookbooks/consistency-noul.md) (~3.7k tok): Runs a 14-Noul claims rubric 15 times against Jev and six LLM conditions; Jev's mean probability std dev is 0.0102 at 111ms per call, and an uncertain band routes 0.30-0.70 to a human. - [Cookbook: Date extraction](https://jevwiki.ai/raw/wiki/cookbooks/date-extraction.md) (~3.9k tok): Seven Choice questions read a date's parts off a document in one call; code does all the calendar math and gates on the weakest part's confidence. - [Cookbook: Knowledge graph entity alignment](https://jevwiki.ai/raw/wiki/cookbooks/entity-alignment.md) (~3.0k tok): One three-level Score decides merge / curate / leave-unlinked for 450 candidate entity pairs, with three Nouls riding along to tell the curator which field disagrees. - [Cookbook: Function calling](https://jevwiki.ai/raw/wiki/cookbooks/function-calling.md) (~2.9k tok): Map typed Python functions to Jev Choice/Noul questions so a natural-language command becomes a validated call with a per-argument confidence. - [Cookbook: Hierarchical classification](https://jevwiki.ai/raw/wiki/cookbooks/hierarchical-classification.md) (~2.9k tok): Walk a deep taxonomy to a leaf by asking one Choice per sibling set, and keep K paths alive with a geometric-mean beam search instead of a greedy walk. - [Cookbook: Guardrails for LLMs](https://jevwiki.ai/raw/wiki/cookbooks/llm-guardrails.md) (~3.1k tok): Screen every LLM input and output with one Jev request: four hazard Nouls plus a severity Score, then route pass / review / block / support on thresholds you own. - [Cookbooks overview](https://jevwiki.ai/raw/wiki/cookbooks/overview.md) (~3.4k tok): Catalog of all 18 TypeSafe cookbooks — goal, primitives and domain each — plus how a cookbook is structured and why the published pip line does not work. - [Cookbook: Parallel questions](https://jevwiki.ai/raw/wiki/cookbooks/parallel-questions.md) (~3.1k tok): Batching 13 questions about one 54k-character document into a single Jev call was 12.2x cheaper and 10.0x faster than 13 single-question calls, with identical answers. - [Cookbook: Pre-parsed value extraction](https://jevwiki.ai/raw/wiki/cookbooks/pre-parsed-value-extraction.md) (~2.7k tok): A recall-tuned regex finds candidate spans, a Jev Choice over those exact spans picks the one the question asks for, and code copies it verbatim and normalizes it. - [Cookbook: Re-ranking](https://jevwiki.ai/raw/wiki/cookbooks/rerank.md) (~2.6k tok): BM25 builds a 30-candidate shortlist per CLERC query, then one Noul per query-candidate pair re-sorts it: top-1 rises 5% to 18%, top-10 38% to 62% for $0.0645. - [Cookbook: SDE cascade](https://jevwiki.ai/raw/wiki/cookbooks/sde-cascade.md) (~4.0k tok): Extract with a cheap model, verify every field with a battery of Jev Nouls framed so true means wrong, and escalate to a reasoning model only when a flag fires. - [Cookbook: Line-by-line search](https://jevwiki.ai/raw/wiki/cookbooks/semantic-find.md) (~2.4k tok): Tag a document's lines with ids, rank them with one Choice over those ids, and pair it with a Noul that says whether the document answers the query at all. - [Cookbook: Skill suggestion](https://jevwiki.ai/raw/wiki/cookbooks/skill-suggestion.md) (~3.8k tok): Two Jev requests per agent turn — rank 182 skills then re-read the top three — cut wrong skill loads from 16.8% to 7.3% and needless loads from 9.8% to 4.0%. ## Guides - [Playbook for LLM agents building with Jev](https://jevwiki.ai/raw/wiki/guides/agent-integration-playbook.md) (~7.0k tok): Seven-step playbook for an agent told 'use Jev for X': decide, decompose, shape state, write questions, call, consume, handle errors, test. - [Choosing between Choice, Score, Noul](https://jevwiki.ai/raw/wiki/guides/choosing-a-primitive.md) (~3.4k tok): Decision table mapping question shape to Choice, Score or Noul, the anti-patterns for each, and how to decompose a complex judgment into several primitives. - [Quickstart: first call in HTTP, Python, JS](https://jevwiki.ai/raw/wiki/guides/quickstart.md) (~3.6k tok): Get a key at console.typesafe.ai, export TYPESAFE_API_KEY, and make your first POST /v1/systemone call in curl, Python, or TypeScript. - [Smart home assistant demo walkthrough](https://jevwiki.ai/raw/wiki/guides/smart-home-demo.md) (~2.3k tok): TypeSafe's smart home demo: one speculative fan-out per utterance routes device commands, while an LLM handles compound splitting and conversational fallback. - [Testing and evaluating a Jev workflow](https://jevwiki.ai/raw/wiki/guides/testing-and-evaluation.md) (~4.3k tok): Measure accuracy, calibration, repeatability, latency and cost on your own labelled set; pick thresholds from confidence buckets; regression-test jaggedness. - [Writing instructions and criteria that Jev reads correctly](https://jevwiki.ai/raw/wiki/guides/writing-instructions-and-criteria.md) (~4.6k tok): How to phrase instructions and criteria for each question type, with before/after rewrites for literal reading, negation, numbers and dates, plus an edge-case testing loop. ## Ideas (community) - [Community repos: what people built and how they use Jev](https://jevwiki.ai/raw/wiki/ideas/community-repos.md) (~2.6k tok): Index of 58 third-party Jev repos, one line each, split across two domain pages, plus reusable designs and a vetting checklist. - [Consult guide: could Jev help this project?](https://jevwiki.ai/raw/wiki/ideas/consult.md) (~2.0k tok): Procedure for an agent asked 'here is my project, could Jev help?': what to read (10-15k tokens), how to find and filter decisions, and the shortlist format to return. - [Field reports: independent evaluations, critiques, open replicas](https://jevwiki.ai/raw/wiki/ideas/field-reports.md) (~3.4k tok): What independent testers measured in Jev's first week: repeatable bounded decisions, weak standalone browser control, and probe-measured order and option-set sensitivity. - [Measurements, access routes and open replicas](https://jevwiki.ai/raw/wiki/ideas/measurements.md) (~3.0k tok): Numbers measured with Jev — cost, latency, calibration, token accounting, vendor harness figures — plus access routes and replica trust notes. Every figure is one run; re-measure yours. - [Ideas section overview](https://jevwiki.ai/raw/wiki/ideas/overview.md) (~0.8k tok): One-screen map of the community-tier ideas section: which index or domain page to open for which question, and how far to trust them. - [Decision patterns from the community (with fit verdicts)](https://jevwiki.ai/raw/wiki/ideas/patterns.md) (~2.4k tok): Index of all 37 community decision patterns (ID, primitive, fit verdict, page), routing by project type to one of four domain pages, retrofit order, and anti-patterns. - [Patterns: agent internals, context and coding agents](https://jevwiki.ai/raw/wiki/ideas/patterns-agents.md) (~4.4k tok): P01-P11 plus P35-P37: community decision patterns for agent loops, model and security routing, context, memory, coding agents and self-compiling workflows, each with a fit verdict. - [Patterns: marketing, sales, GTM, content, support and ops](https://jevwiki.ai/raw/wiki/ideas/patterns-business.md) (~4.5k tok): Eleven community decision patterns for GTM, content, support and ops — lead scoring, next-best action, churn, draft QA, SEO, triage, approvals — each with a fit verdict. - [Patterns: judging, search, documents, real-time and markets](https://jevwiki.ai/raw/wiki/ideas/patterns-data.md) (~3.0k tok): P15-P18 and P24-P27: community decision patterns for judging, moderation, search and ranking, corpus labelling, documents, transcripts, devices and markets, with fit verdicts. - [Patterns: browser, computer use, voice and product UI](https://jevwiki.ai/raw/wiki/ideas/patterns-interface.md) (~1.3k tok): Community patterns where Jev picks the next on-screen action or labels items inside a product UI: browser agents, site-exposed tools, desktop/mobile/voice control, in-UI classifiers (P12-P14, P34). - [Repos: coding agents, orchestration, memory, browser and computer use, integrations](https://jevwiki.ai/raw/wiki/ideas/repos-agents.md) (~4.4k tok): 32 community repos that put Jev inside a coding agent, an orchestrator, a browser or computer-use loop, or an MCP/CLI integration surface. - [Repos: data, documents, judging, real-time, markets, business apps, replicas](https://jevwiki.ai/raw/wiki/ideas/repos-apps.md) (~3.8k tok): 23 community repos that point Jev at rows, documents, media, live simulations, markets, business questions — plus the open replicas of the interface. ## Entities - [Blog: Lies, Damned Lies, and Benchmarks](https://jevwiki.ai/raw/wiki/entities/blog-antibenchmaxxing.md) (~1.7k tok): 2026-09-11 post against benchmaxxing, ending in TypeSafe's commitment to publish no standard benchmark table and to retire dated eval snapshots. - [Blog: The Bitterest Lesson](https://jevwiki.ai/raw/wiki/entities/blog-bitterest-lesson.md) (~1.4k tok): 2026-09-10 post arguing the ML priority order is doing the right task > data > compute > algorithms, using InstructGPT vs GPT-3 as the proof. - [Blog: Introducing System One Models & Jev (2026-09-15)](https://jevwiki.ai/raw/wiki/entities/blog-introducing-system-one.md) (~2.6k tok): Digest of TypeSafe's launch post: System One vs LLM comparison table, price/speed claims, the evidence section, and the caveats the post states itself. - [Blog: AI: too good to be true, too bad to be useful (2026-06-19)](https://jevwiki.ai/raw/wiki/entities/blog-too-good-to-be-true.md) (~0.7k tok): Stub: the body of this 2026-06-19 TypeSafe post was not captured (JS-rendered); only title, date, category, and the homepage teaser are known. - [typesafe-ai GitHub organisation and repos](https://jevwiki.ai/raw/wiki/entities/github-repos.md) (~2.3k tok): The github.com/typesafe-ai org: two SDKs, the agent skill marketplace, the System One adapter, plus infra repos and two forks. - [Jev (model)](https://jevwiki.ai/raw/wiki/entities/jev.md) (~1.6k tok): Jev is TypeSafe's first System One model: text in, typed decisions with calibrated probabilities out, $0.042/MTok input, 64k context, 70-500 ms. - [Manifesto: Composable AI — Build Prod, Not God](https://jevwiki.ai/raw/wiki/entities/manifesto.md) (~1.7k tok): TypeSafe's manifesto argues the bottleneck is composability, not intelligence, and lays out a three-step plan to make AI a dependable software primitive. - [Press and third-party coverage](https://jevwiki.ai/raw/wiki/entities/press-and-coverage.md) (~3.5k tok): Outlet-by-outlet digest of Jev's launch coverage (2026-09-15/16), separating TypeSafe's claims from each outlet's own assertions and caveats. - [Founders and team](https://jevwiki.ai/raw/wiki/entities/team.md) (~1.6k tok): TypeSafe AI's three founders (Diogo Almeida CEO, Sasha Sheng COO, Erik Gafni CTO), the team's stated culture and values, and the SDK maintainers. - [TypeSafe AI (company)](https://jevwiki.ai/raw/wiki/entities/typesafe-ai.md) (~2.5k tok): TypeSafe AI, Inc. is a San Francisco frontier AI lab that left stealth on 2026-09-15 with $40M led by DCVC and Jev, its first System One model. - [console.typesafe.ai (console + playground)](https://jevwiki.ai/raw/wiki/entities/typesafe-console.md) (~1.1k tok): console.typesafe.ai is TypeSafe's Web Interface: log in with Google or an email code, try Jev in the Playground, and mint API keys. ## Syntheses - [FAQ for agents and developers](https://jevwiki.ai/raw/wiki/syntheses/faq.md) (~4.9k tok): Forty answers about Jev — what it is, the API, the SDKs, design, confidence, limits and pricing, company and legal — each linked to the page carrying the detail. - [Glossary](https://jevwiki.ai/raw/wiki/syntheses/glossary.md) (~4.7k tok): Sixty-nine Jev, TypeSafe, API, SDK and evaluation terms defined in one or two sentences, alphabetical, each linked to the page with the full contract. - [Jev vs LLM JSON mode / structured outputs](https://jevwiki.ai/raw/wiki/syntheses/jev-vs-llm-structured-outputs.md) (~4.3k tok): What Jev gives you that an LLM in JSON mode does not (and vice versa), a row-by-row comparison, when to pick each, how to combine them, and how to benchmark the pair honestly. - [Versions and timeline (models, SDKs, API, company)](https://jevwiki.ai/raw/wiki/syntheses/version-timeline.md) (~4.0k tok): Every datable fact in the wiki on one axis: company, model versions, API, both SDKs, the plugin and adapter, legal documents, repo commits, and where the dates disagree. ## Upstream sources - [TypeSafe docs llms.txt](https://docs.typesafe.ai/llms.txt) - [TypeSafe docs llms-full.txt](https://docs.typesafe.ai/llms-full.txt) - [OpenAPI spec](https://api.typesafe.ai/openapi.json) - [typesafe-ai on GitHub](https://github.com/typesafe-ai)