Jev for search — a daily feed

Cool little weekend OSS demo: Agent Seek You .com finds pages. @typesafeai Jev re-ranks the SERP 50 questions: 74% preferred the short list vs You .com order 86% had the answer in the top 3! Tune ranking -> less junk in context, less LLM $ https://t.co/S3nn2OrBKD

I made a tool with Jev that reads my whole codebase to answer one question: what should I be looking at? Every chunk of code in your project is scored for relevance. https://t.co/QFq2XBLatm

豆包工作的 /plan 和 /goal ，以及任务队列功能真的非常好用，真的做的越来越好了！ 使用豆包工作帮我完成 40 份关于 Jev 的研究文件，并整理成五主题知识库以及研究观察站，为字节团队点赞！ https://t.co/1fylU7WU6I

I made an Algolia alternative using Jev https://t.co/USs2C6kKAS

(2/5) hit some wild results. argument order and labels both move the score i gave jev 2 identical diffs (labelled A and B) and asked which is better. the results were shocking: fork A at 0.95 (five runs: 0.95, 0.96, 0.95, 0.96, 0.96) put B first with the same names and A is still 0.90. same diffs, different answer pretty outrageous we dont have label invariance for a model that outputs probabiliti

Built local search with Jev. No pre-indexing: just describe what you want in plain English and it ranks the best matches first. vs. keyword search: Hard set: 0.32 → 0.38 Science set: 0.69 → 0.75 <$0.001/query https://t.co/xocLuYjtOl https://t.co/mvJJOYASwq

Apparently “top-k and hope” isn’t a ranking strategy 😭 So I built jev-rankkit — reranking for RAG, agents, tools & Python objects with lexical + semantic + Jev/model ranking. GitHub: https://t.co/1dGA0jmzdF PyPI: https://t.co/luMgPMExOe #BuildInPublic #OpenSource #AI

The current incumbent to search #mtg cards is https://t.co/cZKy8bKIOi. It uses a complex query syntax that lets users find the type of card they want to find. With Jev, you can produce a better outcome using simple natural language. I made https://t.co/iSMZJ5faic to do just that.

This is one of the best use cases I have found for @typesafeai Jev so far, instantly find anything in a PDF using natural language. Try it out: https://t.co/mByvIRcoQp https://t.co/BB9AEerhJk

@iam_zachi Your pg-jev extension is included in Jev Gallery. We’ve documented the SQL semantic filtering and kept the benchmark’s dataset/cache conditions alongside the source: https://t.co/vJz1bK7Hkk

Jev Screen: second screener for systematic reviews. Abstracts → include/exclude/maybe, PRISMA + RIS. Uncertain cases stay human. Offline demo uses synthetic scores; 61 tests pass. No live accuracy claim. Where does screening slow you down? https://t.co/vcoeDSudI8

Jevで動く類語検索を開発したので、Chrome拡張機能として実装してみた。 素晴らしい。 これを個人レベルでちょちょいと作れて、しかも検索速くて、かつコストも激安。 なるほど、これは面白いな。 https://t.co/XrEbbf5Cvw

@typesafeai 를 이용해서 현재 서비스 되고 있는 유튜버 추천 카페, 맛집 검색 서비스, https://t.co/RftYyB6Jzy 에 적용한 사례에 대한 글입니다. 검색 리랭크시 10.3초에서 2.3초로 줄였습니다. 10.3초면 거의 서비스 불가능한 상태인데 jev 적용 후 서비스 가능한 수준으로 줄였습니다. https://t.co/zNILKX7Xsh

Launching JevFilter, a TypeScript library that uses Jev to turn natural-language search into filters for your existing APIs. Example: “urgent billing tickets from last week” will give you { priority: "high", category: "billing", createdAt: { ... } } You define the fields and allowed values. Dates and numbers are parsed in code, and the model only chooses from options generated from your schema. Yo

Pretty incredible. Just used Jev + @eltokh7's JSort to re-score the entire Fedlock corpus. 4005 FOMC public speeches and statements since the mid 90s. A glimpse of what "intelligence too cheap to meter" looks like. https://t.co/8yqR8DeLpH

I built a free internal linking tool using @typesafeai Jev for classifying and selecting the links. BYOK or pay $1 to use mine. It works for up to 500 pages and gives you a CSV or JSON to pass to an LLM to implement. https://t.co/i01qMCEXa6 https://t.co/7uhzyaiGQj

made a free tool to KILL @Octolens $1M (company btw) for the @convex hackathon. I'll show you a GTM strat no one is talking about... setup today for free. We use @treg_ai to find out our competitors @firecrawl to find good comms and @typesafeai for if content is relevant🤯🤯 https://t.co/z3JYP1fT3Q

Built matchcn: semantic component search for coding agents, powered by Jev. Describe the UI you need. It finds the component and returns the install command. No component name needed. No browsing 9 different libraries. 1,783 components across 9 @shadcn registries, classified across six dimensions: purpose, motion, density, interaction, data, and decoration. Jev returns calibrated probabilities ins

Smart web search CLI for agents, backed by Jev. Saves a lot of tokens. #golang https://t.co/OApFrXBoGS https://t.co/1jN4QrJmYw

In the first 24hrs of https://t.co/9tGq42lv3d Over 50,000 visitors Over 125,000 individual API requests (asks to Jev) Over 39 million tokens (majority being input) And it cost me... $0 https://t.co/n0I7gXItsn

I put 10,000 movie posters in one endless room with Jev. Tell it what you're in the mood for and the walls rearrange. https://t.co/1fPdHBSosj https://t.co/QGduf0QJfe

semantic search prototype w. Jev for Korale. https://t.co/rJpdwiT5w1

『Codex』×『Jev』の使い方を 取り入れるだけで月々の トークン節約をして、 月々の利益を最大化できる 👇 月々のコストは3000円に抑えて Codexに動画編集を自動化させて 自動収益構造を作る 【詳細】 使っているのは「JevGrep」。 質問に関係するコードを Jevで選別する検索ツール。(GitHub) 例えば、 「セッションの有効期限を 処理しているのはどこ？」 と聞くだけで、関連するコードを ファイル名・行番号付きで返してくれる。 この仕組みをCodexに接続して、 まずJevに読むべき場所を探させる。 Codexは、その候補を詳しく確認して 実装や修正に進むという分業。 狙いは、何度も検索したり、 ファイルを丸ごと読んだりする 探索の負担を減らすこと。 つまり、トークンを節約するために 「回答を短くする」のではなく、 「読ませる情報を先に絞る」。 Jevは「探して絞る」

Built on Upstash Redis Search + Jev! Repo: https://t.co/JhIs9lO5mb

So instead of chasing which posts I should pay attention to, I am using Jev (combine with my original Algo) to have this floating panel Now I can quickly find which posts are most relate to my interest, and probably save ton of time because of dump scrolling, but hey, dump scrolling is fun too,

Was inspired to combine my favorite search provider @p0 with the new JEV hotness from @typesafeai I bring you this fun demo https://t.co/hVJj7DIp1D Jev gives a yes/no with real probabilities and Parallel does the hard work on finding sources. https://t.co/fMb9rl2sd2

I benchmarked Jev vs. Qwen3-Reranker-0.6B on 623 BEIR queries, reranking 100 passages each. What I found: • Ranking quality was effectively tied on SciFact and NFCorpus datasets • Jev was ~2.4× faster at p95 • Qwen was 3.9× cheaper I open-sourced the harness and raw responses so you can reproduce the results. Blog: https://t.co/HMlf07Oa9l Github repo: https://t.co/53NrsMzXq6 @typesafeai @Alibaba_Q

I built a tool with Jev + Reddit that finds conversations where my product might actually be useful. The idea was very simple. I wanted to find someone who is already asking for a recommendation or trying to solve a problem, make sure the post itself is recent, and see if I could help. But here, searching for a topic gives me all sorts of things, like product launches, people promoting their tools

웹/토스앱으로 서비스 중인 '2026 아시안게임 한국 일정', 일단 웹쪽에만 Jev 기반의 자연어 검색을 적용했다. 잘 되네?ㅋ https://t.co/AD6zCIKOCn

模仿 hypit官方的视频结合 Jev ✖️ Hypit 做了一个 商品自动匹配AI数字达人的视频，效果绝了 项目地址 ⬇️ https://t.co/DRum3LX7No https://t.co/PNhsTsthSz

Jev is the FASTEST with huge amounts of data! So I decided to build Index, instantly searches over 100 posts per profile, identifies your niche and shows your rank among the top voices in the space. Just being able to achieve these results in a matter of seconds is next level. Really loved building this and you can try it yourself here: https://t.co/oZmtZM52Dw Feel free to tell me your thoughts ab

流行りのJevを使ってみた AI向け長期メモリ&意思決定基準検索RAG 過去の意思決定を、「判断構造」で探すためのGit-backed memory まだ全然試験段階 https://t.co/lA0w3lLElC

Thought it was clever how CJ used Jev to pick a specific sentence from Wikipedia's opening paragraph as an answer. That gave me the idea for a Chrome extension that uses Jev to pick the most relevant link from the first page of search results. It turns Google's "I'm Feeling Lucky" button into "I'm Feeling Jevvy". Click it and the extension fetches the results, lets Jev pick a link, and redirects y

ToneBird + Jev = find your reply, fast ⚡ 18 replies → 3 matches. 0.28–0.53s per Jev call in this demo. → type what you mean → change your mind → watch the replies reshuffle (real API calls. video at 1x speed.) Tiny experiment. Very everyday problem. #Tonebird https://t.co/wAfmPjtAwr

スプラシューターでガチヤグラ最強ギアはどれか！？ Jev（AI）を使って、まずはメインギア3つのみで全通り判定。100点満点で評価してもらい1番を決めました。 次にサブギアを0.1ずつギアを変更してAB比較。 膨大な検証の結果、Jevが考えるスシのヤグラ最強ギア決まりました。 ↓ https://t.co/0CtciNVnKJ

https://t.co/pbZHydFK9P Can we replace the black-box embeddings with content-relevant structural vectors for document retrieval and content similarity? The first experiment with @typesafeai's Jev shows a good starting point.