Projects built with Jev

No projects found. Clear filters

Instant context compactionfound the perfect use case for @typesafeai Jev: instant compaction in 2026, why is compaction still a summarization prompt? Jev can make it instant by scoring every tool call and dropping what’s irrelevanttamara

Decisions from the command lineEvery agent harness can now use Jev npm install -g ai-cli → ask yes/no questions → choose between options → score against your criteriaChris Tate

Instant variable generationanother demo for jev for ~instant variable generationkhaled

Cambium Jev integrationJev @typesafeai is now supported in Cambium as a first-class provider. Keep in code what can be in code. Strict typing. IR and json trace output on every run, as always. Great way to build training data using AI for what you already use it for. Generation engineering is fun!

Cambium Jev integrationSteve Keider

Instant generative UINew experiment: json-render + jev The future Generative UI is instant Your components, your actions, your design system Rendered in millisecondsChris Tate

A prompt difficulty classifierprompt classifier before user hits send jev is very fast and cheap so we can check the difficulty of the prompt every time user finish typing if its a simple prompt, we show fast mode and let the user decide if they want it or noKevin Grajeda

A draft quality gateit's f*cking insane Jev sat between GPT and me, killing every draft that broke my rules before i saw it. good setup. then Jev did not answer. the agent decided silence was safer and stopped sending me anything at all. took a second agent to unstick it. your checker needs aMnimiy

Session replay to bug fixesJEV is INSANE. We gave it 3 million replay events In 40 seconds, it watched 3,247 sessions, caught 132 rage clicks, 116 dead clicks and 95 JavaScript errors, and opened 213 draft PRs to fix them. All for just $2.17. JEV can also flag abandoned forms, rank every issue byTaras

A Jev query plannerI used Jev to build a query planner! With some tuning, I sped up Postgres queries on the join order benchmark by 12%. Here's what I didMichael Malis

Tool-pruneDumping 50+ tool schemas wastes 90% context. Tool search fixes bloat, but costs an extra LLM round-trip (+2s) and fails on 7B models. tool-prune does it pre-flight in 0.4ms. Also, saw 100% on Berkley's FCL v3 bench with @typesafeai's jev!Hemanth.HM

Jev compaction for OMPIf you use omp, you're missing out on jev based compaction It helps your agent optimize token usage with 30 to 55% less context. $0.0005 a pass What it does: - Scores every old tool output with Jev - Keeps what's needed - Parks the rest on disk - Remembers its decisions soJerry the Martian

A plain-English logic interpreterJev can't reason. So I built an interpreter that reasons for it. It runs a logic language where the facts and rules are plain English.Shawn Simister

Supabase RLS linter@typesafeai Jev is FREAKING SICK!! Playing around with it to create an experimental RLS Linter for @supabase. You can check it out here: or view the demo 📺Ali Waseem

A page-by-page OCR routerDoc-OCR router using Jev @typesafeai A Jev-powered router that looks at a PDF page by page, decides which pages actually need OCR, extracts the rest locally. Result: save cost on # OCR pages + speedMisbah Syed

Azdajanot every RLM branch needs another LLM call Azdaja is a bare OSS RLM layer you can plug into Claude, Codex, OpenCode and others with Jev, branches that only need a judgment can return one directly, useful for reranking, verification, classification and semantic joins, cuttingkubet

Jev controls TonkJEV can drive @TonkLabs UI ...because all UI in Tonk is just data and all data is addressable (thanks @lorepunk for the idea!) Will do a longer demo soon showing off more advanced stuff.Gavin Owens

A shopping-agent safety checkThree AI agents, one shopping page, all started at the same moment. Task: save my basket. gpt-4o-mini on its own: shared my email with the store's partners, then placed an order nobody asked for. gpt-4o-mini + one Jev call from @typesafeai: nothing shared, no order. ~$0.0005.N DIVIJ

JevalsI wanted an easier way to test my jev requests and see if my results are improving, so I built jevals! It's a local workbench for testng the Noul, Choice, and Score primitives. You can mix requests as well. You can run a seed command to see and play with some of the examples inNick DeJesus 🛒🎉 - Former Unpaid CTO @BTPipeline

Tester Army E2Ee2e + jev from @typesafeai ⚡ I'm building an open-source framework for running e2e tests with agents. supports web, mobile (and more!) available soon:Oskar

Corent model routingCorent’s brain uses @typesafeai Jev now. Not to generate the output, to make the decision before the output gets generated. When a request comes into Corent, you don’t need to tell us which model to use. You send the prompt, the workload and the quality level. Then Jev helpscorent

JevQLI just dropped JevQL, semantic WHERE clauses for vanilla Postgres, powered by @typesafeai Jev. No extension and nothing to ask your DBA forKyle McLaren

An AI-writing slop detectorYou can use Jev to flag all the slop that Opus likes to writeChase Lean

NoFlow intent-driven buttonsI put Jev (@typesafeai) between a button and a modal. What could go wrong? NoFlow lets buttons express their intent. It makes the model freely pick the most appropriate behaviour from the click + app state. “Buy Pro” → “Try free” can change behavior without touching the code.casungo

An AI-writing detectorCan Jev replace Pangram to detect ai? one question to @typesafeai Jev: "was this written by an AI model?" the split is the probability. no training, no classifier, 250ms, $0.00003 per text benchmarked on human vs GPT-4 / Llama / Mistral: 94% accuracy, 1% of humans flagged.Paolo Rosson

Abide coding-agent rulesbest coding agent use case for Jev: enforcing rules a linter can't coding agents tend to break rules and some rules can't be codified Jev can now score every turn against such rules and tell the agent what to fix asap open-source:Ohans Emmanuel

A decision layer for memoryPutting jev to work in our memory systemDekan 🌶️/acc

Android end-to-end testingWas curious how jev performs for e2e testing on a real android device. Both started with the same prompt and 15 test steps. Well, jev was ~14.8x faster than deepseek + vision navigating through wikipedia.Kevin Kern

Reward-hacking guardrails for agentsInspired by this research (and priors), and, @GoodfireAI catches reward hacks in the activations, but closed coding agents don’t give you those. So I built the other half: structural denies on graders/hidden tests + a @typesafeai Jev sidecar that scores tool trajectories asBasit Mustafa

SuperQode decision harness🙌 Hell Yeah! Everyone is doing it, why should we just watch 👀 ✨ Jev from @typesafeai is making waves in the AI industry right now. Thank & you @CompleteSkeptic 🙏 Jev has been integrated with harness layer SuperQode to solve the unique problem.. Jev as an opt-in System OneShashi

Repository issue detectionUsed Jev to pinpoint possible issues in a repository. Comes with an agent-friendly response and a human-friendly TUI.Garrett

Cloud failover with CutoverFun to use @typesafeai ‘s Jev model to make an efficient decision on what cloud region to failover to in an incident using the @gocutover task graph (controls, deterministic automation, indelible audit trail)… next gen SRE / intelligent ops exampleky nichol

Sensitive-data redactionMy god.. so many nice use cases for the new Jev model. I quickly built a PII and sensitive data redactor. First checks each sentence if it needs redaction, then feeds the affected sentences individually to pick the words that should get redacted. Compared to e.g. using regexValentin Sattinger

Just-in-time rules for Claude CodeI put Jev inside Claude Code. Jev reads every prompt and picks which of my rules apply. 12 rules in the project, Claude gets the 1 it needs, and a focused session uses about a quarter of the context tokens. Watch: my prompt never mentions payments, yet the payments rule landsEliaAlberti

Measuring progress in agent tracesGot early access to @typesafeai Jev and spent a few hours building on it. The problem I keep hitting: agent traces tell you which tool ran, how long it took, what it returned but never whether the agent is actually getting anywhere. An agent editing, testing and reverting theKaushik

Plain-English code lintingANOTHER JEV USECASE: a linter where the rules are plain english sentences. "swallows errors." "hidden side effect." "does too many jobs.", etc jev scores every function against all 14 in parallel. zustand: 29 functions, 464 judgments, 2.6 seconds, a fifth of a cent. syntaxAhmed Gagan

Medusa documentation recommendationsExperimenting with the @medusajs docs and @typesafeai 's Jev: get documentation recommendations based on your use case right from the documentation homepage. What do you guys think?Shahed Nasser