Jev Video

A TypeSafe media piece arguing that RLHF-trained language models please humans and assist rather than make reliable autonomous decisions, and asking what class of model comes next.

A TypeSafe media profile of co-founder and CEO Diogo Almeida, framed around why some AI applications over-promise while capable models can still under-deliver in practice.

Sean Brownell scores sample call transcripts in a harness, including a case where the agent reads out card details, and compares that pass with a frontier-model estimate.

AIEdTalks part 4 on the batching trick: a loop that resends a long document for every question, versus one call that asks all thirteen.

Dan Makes Things wires an ESP32 robot so Jev picks the next move, and asks whether it is the first robot controlled that way.

Max Jesch, on IBM Developer, walks Jev as a System One model that returns probabilities for classification instead of generating text.

Martin Keen, on IBM Technology, explains Jev as a System One model whose output is a decision rather than generated text.

LangChain maps Kahneman's split onto models: System 1 as the fast automatic decision, System 2 as the slow deliberate one, with Jev in the first role.

AiNow dates the Jev release to September 2026 and walks live demos of a model that returns a decision instead of the next token.

Aishwarya Srinivasan's 22-minute end-to-end on using System One models inside software, instead of sending every structured check to a text model.

Dots and Arrows on production checks that need one answer in about 200 milliseconds, and why that is a different job from streaming tokens.

Automate with Marc tests Laya, the open decision model people shorthand as Jev for free, on 100 sales leads beside Claude Code.

Ben Horowitz and Martin Casado interview Diogo Almeida on why capable models have not automated ordinary software, and what Jev is for instead.

Turing Post TV tests Jev from early access in the TypeSafe playground, starting from the constraint that it does not write essays or code.

India Jade explains RLCD as a response to using a frontier model for a small decision, which is the job TypeSafe trains Jev for.

ML Visualized retells Diogo Almeida's Latent Space point that a launch chart with the new model on top of every row is not something you can trust as a benchmark.

Neural Breakdown with AVB livestreams against the TypeSafe Jev API, walking parallel constrained decoding and System One call shapes in code.

Fireship on Diogo Almeida's System One launch claims: no chat output, and the headline 200× speed and 400× cost multiples.

Bloomberg News clip on TypeSafe's non-chat model and its promise of a cheaper, faster alternative to large language models.

Patrick Desjardins wires Jev into Cursor for deterministic decisions from the public docs, with blog posts on hundreds of text rules under two seconds.

Tim Lin's visual explainer on System 1 versus System 2, the Choice, Score, and Noul primitives, and how JevScout uses Jev inside a coding agent.

plain. on Jev as typed answers over state, the September 2026 launch, DCVC seed funding, and Diogo Almeida's InstructGPT background.

Hands-on QA automation examples with Jev: Playwright failure classification, guardrails, RAG, API calls from Postman, and decision-focused testing.

VikStack lists ten concepts for Jev: non-autoregressive parallel sampling, 70–500 ms claims, and boolean, enum, and rubric answer shapes.

Aditya Arya signs into early access and tours the UI, tying Jev to RLCD training and 70–500 ms structured decisions instead of chat.

Innovations for Education explains Jev for no-code and vibe coders when if-then rules need judgment calls like spam or toxicity without burning tokens.

Heroic steps on typed decisions with confidence scores, parallel output speed, the Doom demo, and citations to TypeSafe, The Register, and Every.to.

Riley Brown on a decision model that cannot write: a model router, Choice, Score, and Noul, then speed, cost, and where to get access.

Codevolution's walkthrough of state, questions, and answers, then a customer-message demo in the TypeSafe playground and the TypeScript SDK.

Yash Thakker's intro to System One, plus 50 open-source demos that compare Jev with a regular LLM, linked from the description.

Joe Maddalone on using Jev for routing and filtering before a generative model, with demo code in a transcript-analyzer repo.

KodeKloud on why a System One model returns calibrated probabilities, then a Papers, Please-style queue where Jev is compared with GPT-5.5 and DeepSeek.

CampusX on Jev as a classifier over schema-constrained choices, with a live benchmark and an e-commerce review demo.

Under the Hood AI compares TypeSafe's closed decision model with Laya, the open model Convai released that can run on a laptop.

Automate with Marc on using Jev to sort, tag, or score fixed answers, including a demo that tagged competitor ads.

ZAUEY on a model that picks from options you define and returns a probability, and why many questions on one state stay cheap.

TJ Robertson on the jobs that fit a classifier with a confidence score, and the speed and cost he trades against a frontier chat model.

Telusko Alien Coach on why generating text is overhead for a classifier, with Vercel's reported speedup and Armin Ronacher's objection.

Prism Labs checks the launch demos against TypeSafe's own footnotes: what they keep from the 70 ms claim, and what they drop from the speed multiples.

Jaco6 on game demos that feed structured state to Jev instead of screenshots, with TypeSafe's Doom rate and a Mario success note.

Grayson Ho walks colour sorting, an AI town, a Wikipedia race, model routing, and games, and separates narration from the measured parts.

AI with Rithesh rounds up early Jev loops: a drone, games, a browser, email sorting, and one labelled product benchmark.

Reindent separates a guaranteed output shape from a correct decision, and treats the launch speed figures as vendor claims.

Nimrod on what the 200× headline is comparing, where a decision model fits, and a DSPy result he treats as the practical number.

LangChain contrasts an LLM that writes a paragraph and then a boolean with a Noul probability you can threshold, via langchain-typesafe.

WitWireAI recreates a spreadsheet that asks Jev to score customer messages by meaning, and labels the clip as a demo rather than a benchmark.

AI Test Bench runs four tests on synthetic support tickets in the official TypeSafe playground and states what that small set does not prove.

A Japanese demo that asks Jev 0.5 seconds after speech ends whether the turn is really over, then extends the hold time based on the returned score.

A video walkthrough of the control-system pattern: a request becomes structured state, Jev returns typed decisions, and code keeps the expensive model out of every loop.

A video comparison of local Laya against cloud Jev on a Tetris agent, where the open-weights model made decisions 11 times faster on a 16 GB MacBook Air.

Cole Medin's week of hands-on testing, covering where Jev gates replace regex or slow LLM calls, plus demos that play his indie game and triage every open pull request.

A Krish Naik explainer on whether Jev replaces LLMs and where decision models fit next to generative models.

Scott Chacon runs hosted Jev against two local decision models, laya and kev, in a Tetris match and a GitHub Settings filter, with a public test repo.

A walkthrough of what Jev and Laya are and the prior-art controversy, including Laya's claim to have been built over a year earlier.

A skeptical computer-use test of Jev with a public test repository, from Builder.io's Steve.

codebasics explains why a non-generative model matters and walks through hands-on code, including the caveat that Jev cannot write text or code.

A chaptered walkthrough with a Python demo repo covering Noul, Choice, and Score, parallel multi-question calls, and an autonomous triage loop.

A local-inference walkthrough that compares Kev, SemIf, Von, and OpenJev on consumer hardware, including Brier scores and a fake-probabilities warning.

A two-hour recorded run of 50 financial workflows with saved outputs, baselines, and failures, from ledger integrity to AML triage and sanctions matching.

A head-to-head benchmark that prices 200 products from descriptions with Jev against GPT-5.6 Luna and GPT-4.1-Nano.

A video demo of Jev working through years of hoarded Twitter bookmarks in one pass instead of reading them by hand.

Alan's team built an intern-review console on Jev after 4,000+ applications: it checked 3,518 applications in 4 minutes 54 seconds and surfaced 20 for humans.

A ViZDoom agent with two Jev decision channels, navigation at 5 decisions per second and combat at 12, finishing a test run with 18 kills.

A Grammarly-like Mac app built on Jev to test how fast structured decisions feel in a real writing tool.

A small Kalshi trading bot across 15-minute and 1-hour BTC, ETH, and SOL markets, built with Effect and Cursor and powered by Jev.

Near-real-time A/B testing of TikTok and Instagram hooks against 100 rich personas, replacing a $1 to $5 per-run panel with a few cents of Jev.

A gomoku harness that does tactical work locally before asking Jev, shrinking 225 possible moves to about 40 candidates per turn.

Chrome and Firefox extension that reads a YouTube video's caption track and asks Jev for a sponsor probability per 30-second segment, painting the seek bar before the intro ends. It reports catching 77% of SponsorBlock's sponsor seconds across 23 videos at $0.0008 per video.

A Chrome extension and small web app that skips YouTube sponsor reads: Jev names transcript lines or answers yes/no about what it heard, and code owns every timestamp that gets cut.

Render a live Jev meter over any video: Whisper transcribes, every sentence is scored, and ffmpeg exports a 16:9 edit with flags and a scoreboard. Presets cover debates, calls, and pitches.

Diogo Almeida's launch thread: after co-inventing ChatGPT, he spent two years on RLCD and Jev, a System One model he pitches as 20-200x faster and 40-400x cheaper for typed decisions.

A video demo of a bot that asks Jev to decide buy or sell from a live price feed, then places real orders on Kuru's on-chain order book on Monad every 300ms block.

vogel ran Jev over 1,500 of his own emails to test classification quality and posted the results as a video, calling it the most impressive model he has tried for the task.

Kyle Jeong's demo of browser use with Jev and Stagehand: the accessibility tree is the state, actions are the questions, and one remote-browser task cost $0.001 at near-instant speed.

Tamara's demo of fast-jev-compaction for Claude Code: instead of summarizing context, Jev scores every tool call and drops the irrelevant ones, with a video of it running.

OpenRouter's beta announcement for Jev: state plus typed questions returns typed decisions with probabilities, so there is no JSON prompting, parsing layer, or output validation.

Gojiberry's Romàn reports scoring 700 high-intent leads and personalized outreach messages in 40 seconds for $0.09, with confidence scores and lead-message mismatch detection.

Browser Use's Ultrafast agent puts Jev in charge of picking each browser action from a dynamic DOM action space. The author reports a flight search in 7 seconds for about $0.0039.

A 33-second demo clip captioned "when a designer gets access to Jev", among the most-liked organic Jev posts, showing a polished interface built on top of typed decisions.

A 22-minute walkthrough covering what Jev is, API setup, and three demos: a voice-controlled browser, AI memory, and a YouTube predictor, with chapter timestamps in the post.

The pg-jev author demos jev(), a PostgreSQL extension that filters table rows with natural-language conditions and no index or embeddings, judging 129 rows in about a second.

A short demo of a model router built with Jev: each request is classified by Jev and forwarded to the model judged most suitable, shown as a screen recording.

The jevmeter author runs five yes/no questions per sentence across a debate: 1,191 Jev calls, 1.18M tokens, 415 ms median latency, and $0.0497 for the full run.

An independent Jev-compatible public API backed by Qwen3.6-35B-A3B and SGLang radix caching, reporting 64 parallel typed-decision tasks in under a second.