336 entries · metrics captured 2026-09-20 · no media rehosted
Every entry links to its original post. We index and group; the posts and the media belong to their authors, and nothing here is republished.
ctrl-f has never understood a single query. It matches strings. This one scores meaning: every sentence on the page gets a probability with @typesafeai's Jev, and walks you through the hits.…
jev is insane 🤯 here is jev playing smash bros against itself he is controlling all 4 different characters. and literally deciding whats the best move to play against itself all within a fraction of…
btw this is what you get if you ask Jev to play Minecraft with GPT-6 Astra. I still can’t believe it
TypeSafe AI is challenging the industry’s reliance on large language models for every stage of the agentic stack with the launch of Jev, a specialized “System One Model” designed exclusively for…
Claude me integró Jev con Playwright para buscar usados. Lee unos 26 artículos por minuto y decide qué hacer con cada uno en 406 milisegundos Descartó los que no encajaban con lo que busco, ofertó…
#jev it turned out it's good in snake too! fast too! Survived 321 moves! #demo #day_2 #type_safe
pub-sub without sub @typesafeai jev does the routing almost instantly demo link below
Another crazy @typesafeai Jev example: Predictive spreadsheets Spreadsheets recalculate numbers, not meaning. Jev reads intent. Type "Urgency" at the top of a column and, as you type, it figures out…
Meet Jev (Typesafe's First "System One" model) who knew the next step in AI would be making LLMs obsolete... An AI fast enough for real time applications (including playing Doom in real time) where…
Just trying out Jev, I made a Chrome extension that: - Listens to your YouTube audio (optional) - Detects if it gets to a sponsor segment - Skips it ➡️➡️➡️ - All in real-time while costing ~$0.005…
I tried using Jev as an intelligent ordering tool for diffs
@typesafeai’s Jev playing Dragon Ball Sparking Zero. Astra and Fable each write an initial policy for Jev (system 1) to make the real-time decisions. Astra and Fable (system 2) updates the policy as…
First Jev use case published live on Bannerbear! Instant field mapping between template and source, when names are slightly different eg template - photo - name - company_name data source (eg…
small experiment where I used jev to power the global search in replicas. the possibilities of improving UX with models like jev are endless. we can now have things like: - search that understands…
I used Jev to flag AI replies on X. I defined common AI-writing signals as the positive criteria for a Noul question that returns the probability of whether a reply is machine-generated. Things like…
Okay soooo what if @typesafeai's Jev could 'hear'? here's what Toxicity sounds like according to Jev :
my mobile e2e testing library powered by @agent_device and AI sdk now supports Jev from @typesafeai Same test, same loop: Jev: 29s, $0.003 Haiku 4.5: 37s, $0.07
I built a voice controlled computer-use for my mac using @typesafeai's Jev and it's INSANE how fast it is! I can dictate "open the notes app and create..." and the app opens before I even finish my…
Alright. So we know that @typesafeai Jev can fight, but can Jev shoot and avoid getting shot? I had Jev playing one of my favorite arcade shooter, Time Crisis, to test it out. I would say it did…
i wanted to paint with Jev 🎨 watch this beautiful painting of the Golden Gate Bridge, powered by Codex(Astra) + Jev + @Stagehanddev + @browserbase (video at 1x speed)
I built Tetris where an AI model makes every move, but it's not an LLM. Jev (TypeSafe's new "decision model") gets a list of legal landings and picks one. ~400ms per move ~1 cent per 100 pieces Can't…
Jev by @typesafeai is amazing! i let it play pokemon firered, and it was able to beat elite 4 with a party of level 40 pokemons. all under 0.03$
Had to try Jev on Atari Pong. 🏓 I rewrote one structured question to this classic idea, and Jev went from missing the ball to this: 1 point scored, 0 lost in a 33-second run. Next experiment: can a…
Built a Pacman where every single move is decided by @typesafeai jev 👾 Every tile, Jev gets what is around Pacman and returns a direction, strategy, danger score, flags for trapped, committed,…
AIBackends web API now supports Jev by @typesafeai If you need some abstraction from your main app where you can customise the integration, you can use this API layer
Everyone picks Jev confidence thresholds by vibes. 0.95? 0.5? I built jevcal. Give it your data and say "I need 99% accuracy". You get the exact threshold, how much Jev can handle, and how much still…
How to read this index
Every entry links to its original post. We index and group; the posts, the media and the words belong to their authors, and nothing here is republished — excerpts are shortened and no media is rehosted.
The grouping is by the kind of decision being made, not by domain. “Someone sorted their files with it” and “someone drove a browser with it” are not the same kind of fact; what transfers between projects is the shape of the decision, so that is the axis this index is built on.
Entries are not ranked by popularity. A post with ninety views can be a better example than one with a million, because what makes an example useful is how clearly it states the state, the options and the result. Metrics are shown so you can see how much attention a claim got, not so you can sort by it.
What is in here, and what is not
We collected public posts about Jev, then kept only the ones that describe something actually built. Announcements, pricing commentary, comparisons between models and posts with no description were dropped — roughly two in five of everything we harvested. That is why the count on this page is smaller than the number of posts a search would find.
Each entry carries the metrics the author could see at the time we collected it, on 2026-09-20. They are a snapshot, not a live number, and they are the author-reported figure rather than anything we verified. Where a post includes a cost or latency claim, the claim stays with the post; our own same-task measurements live on the benchmark page instead, and they are deliberately not mixed with these.
What this index deliberately does not do
There is no page per post. Four sites in this category took the opposite approach — one has 371 pages, another more than 1,300 — and none of them rank for the queries their pages target. A page whose entire body is somebody else’s short post is not a page a search engine has a reason to serve, and shipping hundreds of them at once is the fastest way to have a whole site ignored.
So the index stays as one page, and the depth lives in the hubs: a handful of pages, each about one kind of decision, each carrying its own measured numbers and the code to run it. If you want to browse, stay here. If you want to build something, the hubs are the useful part.
Thumbnails are loaded directly from X's own CDN (pbs.twimg.com) — this site stores the URL only and never copies, downloads or rehosts the media. Where a thumbnail is unavailable we fall back to a graphic derived from the post's own numbers. Bars show magnitude only (views · likes · saves, log-scaled); they are not a quality score. Open a card at x.com to see the original post and its media.
