Jev for games — a daily feed

Jev 居然输给了一个 1.88B 本地模型？我用俄罗斯方块重新测了一遍 当我看到 this-that-model-1.0 在 68 道决策题上做到 94.1% 准确率，而 Jev 是 76.5% 时，我的第一反应是： 这是真的假的？ 一个只有 1.88B 参数、可以在 Mac 本地运行的小模型，真的能在决策任务上超过 Jev 吗？ 所以我做了一个俄罗斯方块，让两个模型使用相同的棋盘规则、随机种子和方块顺序。程序负责计算所有合法落点，模型只负责决定方块应该放在哪里。 1️⃣ 一开始，Jev 的表现更好 最初，我先过滤掉明显更差的落点，再把剩下的多个候选位置同时交给模型。 在这一模式下，Jev 的表现更加稳定。 我观察到的一轮里： Jev 消除了 5 行，this-that 消除了 3 行。 如果只看到这里，很容易得出结论：本地小模型还是不如 Jev。 但我后来意识到，这种问题可能并不是 t

Claude Opus is insane! Sonic just won’t stop. Pair it with Jev, and I honestly can’t look away. 🤯 I built this on YouWare with Claude Opus 5.5 using one prompt: “Create a beautiful Sonic-inspired endless runner using Three.js and WebGL, with smooth controls, polished visuals and refined interactions. Deliver it as HTML.” The gameplay is so smooth. Play it yourself, or let Jev decide when to jump,

Everyone is talking about Jev, but there’s a new kid in town: Decision-Machine-1. So far in my tests, it’s faster, cheaper on estimated API cost, and beating Jev at tennis. 🎾 I wanted to understand these decision-making models, so I made it fun. I adapted a browser tennis game and let them play against each other. Now I’m watching API calls turn into rallies and wondering why one player keeps rush

I gave Jev a paddle. Can you beat it? 🏓 JevPong is a real-time game powered by @typesafeai decision model. Watch Jev decide, beat it to 7 points and top the leaderboard. Built with GPT-6 Astra in Codex. Routed through Postman Fabric Gateway. Play: https://t.co/4vlfvgSW7f https://t.co/n67P08Cyud

Come watch JEV solve picross puzzles live in real time at https://t.co/YSCkYOQUVl 550 puzzles left!

Jev-Omni playing Chrome Dino, fully local on an M4 Max 🦖 12B decision model (Gemma 4 finetune) converted to MLX 8-bit, running on M4 Max at ~100 ms. No text generation, it just returns P(jump/duck/nothing) for a slice of road cropped ahead of the dino to cover latency scored 1206 here, best 2412

#生成AIなんでも展示会 展示物追加です！ JevライクなLayaを使用してFlappy Birdライクゲーをやらせるものも展示しております！ レスポンス問題解決したのできてね https://t.co/1WQfofIhBi

←Opus 5 + Astra →Opus 5.5改修後 ※Jevと麻雀打てるようにしました https://t.co/43k7Dx9PgE

A three-floor Magic Tower game for observing TypeSafe Jev's strategic judgments. Players can explore manually or let Jev choose among every legal move. The interface displays the complete probability distribution, confidence, model name, and latency for each decision in real time. https://t.co/DuAukSoiRR

看了@karminski3这个测试觉得很有趣 但是总觉得Jev能力不至于此，于是在这套框架基础上补充测试了多组结果，详见视频。 首先是对照组： C0 —— 这个就是牙医的整活随机算法，随机游走。 C1 —— 这个是专业的迷宫算法，类似机器人走迷宫比赛会用到的。 C2 —— 这个是让gpt-5.6-luna模仿类似方法，在同一个Session里一步步探索迷宫。 然后是实验组，全部都是基于Jev的方案： E0 —— 这个就是牙医的Jev, 的确卡死在一个局部最优里 E1 —— 这个是在牙医Jev的基础上，扩充了上下文，包括全部历史操作以及已观察到的地图 E2 —— 这个是把专业迷宫算法的思路，用自然语言告知了Jev，并提供了全部通道经过次数的统计 E3 —— 这个是在E1的基础上，优化了prompt，删除多处对剩余距离的强提示，转而让Jev自己思考最佳策略 简单总结一下结论： 1. 站在成功率

Jev 互換(DiffusionGemma 26B を Jetson AGX Thor でローカル実行)にテトリスをやらせてみたけど、いまいち。最高 61 個・11 ライン。 やり方: 置ける場所をコードで全部列挙して、「どこに落とす? 穴を作らず、積み上げすぎず、できるだけ長く生き残って」という choice の質問で 1 つ選ばせる。 https://t.co/4ZWnxT8mqm

用 Jev(TypeSafe 的 System One 模型)做了个会读心的石头剪刀布:它对「你下一手出什么」输出校准过的概率,还记住你的全部历史。截图里它 81% 把握猜中——结果猜错了。你敢来打个更好的战绩吗 → https://t.co/ObtTZiFbBR https://t.co/Fcq2613Cet

A successful baptism of fire for Opus 5.5 It built CITY LADY, a debt collection game for the enlightened. Follow Vela on her quest to pay her mothers' debt through 10 city landscapes. All duels are @typesafeai Jev-enabled. Verdict: Opus 5.5 is 🔥🔥🔥🔥🔥

Can someone make Laya and Jev work? 😅 I created simple Flappy Bird game and let Jev and Laya control it. But it look weird 😬 https://t.co/qnS9RjB4Vm

JEV was all over my feed, so I built a game around it 😂 https://t.co/2iNkikkEQO an arcade-style game where you play against JEV. JEV classifies the possible choices. You make yours as a human. Works on mobile, but desktop is definitely better 🎮 https://t.co/XYK2TA9mm8

使用 JEV 玩贪吃蛇、打坦克，弱爆了，刚好前端时间用 astra 写了一个纪念碑谷的关卡https://t.co/JIJCUOQBMy，来玩个找不到 dom 的 3d 游戏，能成功算你厉害。截止发帖，已经玩了 15 分钟了，第一个机关都还没正确打开 https://t.co/E9KTBTegfz

Jevの検証。即時性が高くということでゲームに組み込み、D2Lodの戦闘でのターゲット選択、スキル選択、ポジションをJevで決めて操作。日本からで250-400ms程度での判断速度。ゲームは特に選択の連続なのでデモとしても面白いですね。 #Jev #diablo2 #kolbot https://t.co/Yjz3MCjbYz

AIと一緒に謎解きするゲームをGPTsで作ってみた。 ChatGPTにも答えは教えてません。 外部にJevがゲームマスターをするAPIを用意していて、そこから返ってくる反応だけを手掛かりに、AIと一緒に推理していきます。 GPTs: https://t.co/qOHHDS2mkz (要ログイン) https://t.co/NFTfvNmvqr

今話題のJevを使って、麻雀の状況判断アプリ作ってみた。 同じ配牌でも順位や局、点数に応じて方針ってコロコロ変わるからそこら辺を感情抜きにJevに判断してほしい。 （感情に振り回され判断ミスってこの前ぼこぼこに負けた） 精度がまだまだすぎるけど #typeSafeAI #jev https://t.co/FaS8bMMSFB

New project: an ant colony in #rust + #bevy, driven by #typesafe #Jev . 🐜 Play: you're the queen with one text field, 20 ants interpret your orders. Learn: click any ant to see its exact Jev request and answer, or reword the question live. https://t.co/QngchahoSR @typesafeai https://t.co/uOD0b8zJJq

I have jev play pokemon emerald. It just gets a list of what it can do and picks one. It got a starter and beat the first rival. But not on its own. Some things I noticed along the way 🧵 https://t.co/qPL7aXIPlK

When a goal needs multiple steps, Jev gets stuck. To get its starter, it had to visit a neighbor before leaving the town. Jev skipped that and kept trying to leave. The game blocked it 96 times. Without help, getting its first Pokémon took 466 moves. With one hint, it took 41. https://t.co/QAdtbQbKmg

The biggest lesson: most of Jev's bad moves came from bad information. Once it stood on Route 103 and walked away to look for Route 103, because the map had a number, not a name. Once I fixed what Jev was seeing, it started winning. https://t.co/ljCOQ0Mz8N

I turned Jev into a Flappy Bird player and I won against it (sorraay @typesafeai). Can you beat it? Let's see your score: https://t.co/93Bl3QQsFq Tech: - Results stored in @neondatabase Postgres - @nextjs App Router 16 with @DrizzleORM, @shadcn and @tailwindcss - pdx1 region on @vercel gets the lowest latency to Jev's API

二开做了个五子棋，跟 Jev 对战，群友测试了一下都说简单，如果这都不能闭着眼睛赢说明自己已经告别下棋了🤔 https://t.co/9ydKUkYKNr https://t.co/Z9YeKNddPL

做了个小猫生存游戏 让官方 Jev 和本地 Laya (mlx)一起保护小猫 每一步由模型自己选 小猫会因此吃饱、受伤，或者饿肚子 结论：Jev 更准但慢，Laya 极快但不准 三种关卡各测三次 官方 Jev 通关 3/9 局 全部来自找饭关； 本地 Laya 通关 0/9 局 Laya 虽然每局都回过家，却没能满足各关的饱腹、健康或保暖要求 兄弟们说关卡设计是不是难了点？ 后面会做个 Jev-Like 决策模型闯关排行榜

看完@NFT_Chen Laya × Jev 的贪吃蛇对比，我按照他给的源码库照着这个思路做了个「陨石穿梭」🚀 同一套陨石、同一套规则，双屏自由跑 30 秒：左边飞速穿梭，右边还在等下一步。 把决策延迟变成游戏画面，差距一下就直观了。 https://t.co/xQscBdtAbX

JEV is INSANE. So I turned it into a 16-bit game and made a launch trailer for it. That's the whole game: JEV pulls the trigger, another model plans the route. Here's the part I still can't believe. I don't know how to edit videos. The entire 40-second trailer was built by talking to @Pexoai_offical, an AI video tool you just chat with. How it went: → I gave it a character sheet and a storyboard →

last time, I made Jev play Clash Royale. this time, I gave it an opponent: Laya, an open-source alternative. Jev runs through @typesafeai's API, while Laya runs locally on my Mac. Each controls a separate emulator and sees only its own game screen. Both use Qwen (via cerebras) for battlefield vision and OpenCV to read cards and elixir. From there, Jev and Laya decide what card to play and where to

got @typesafeai's Jev reading sprites instead of text health bars, energy bar, character select. @sai_borg turns that into button presses fast enough for a fighting game structured outputs are underrated 🥊 #robosecretary #saifleet https://t.co/NK8UGgPYFD

So jev just came out, and @convex released our AI gateway... Why not throw together a party game? https://t.co/fHkYw1HzsN Just convex, workflow component, ai gateway, convex auth v2 (preview), static hosting component... no api keys required, just a convex account. Nice.

گذاشتم قویترین AI دنیا ماینکرفت بازی کنه! GPT 6 Astra + Jev تماشا در یوتوب: https://t.co/KJuUBknjOD توی این ویدئو، با همدیگه پروژه‌ای که ادعا می‌کرد تونسته ماینکرفت رو توی 8 دقیقه اسپیدران کنه بررسی می‌کنیم و خودمون بازسازیش می‌کنیم با استفاده از Astra و Jev از برادر کوچیکم دعوت کردم بیاد کمی راجب خود ماینکرفت توضیح بده و کاری که ادعا شده ai تونسته انجام بده. همینطور در مورد Jev صحبت می‌کنیم و این

Este fin de semana he estado probando Jev, el nuevo modelo de @typesafeai He creado un pequeño juego RTS donde, a partir de ordenes en lenguaje natural, se traducen a comandos de batalla: ataca, defiente, etc. Jev toma la decisión de que debe hacer en función del input + estado https://t.co/H6NQU908GS

Entrando na moda do Jev, fiz um botzinho de Tibia pra testá-lo Por hora só fica curando/patrulhando/looteando, mas já deu pra dar um teste https://t.co/s09gkh6Lob

声だけで操作するゲーム、公開しました！ https://t.co/dmjedkFnAD 「みぎ!」「ジャンプ!」と叫んで走るランナーと、目隠しのキャラを声で誘導するスイカ割り。 ・反射層: ローカル音声認識(自作OSS hayamimi)の途中経過をキーワード照合して即発火 ・判断層: あいまいな発話だけ、ゲーム状態と一緒に Jev に投げる Jev の往復は中央値 336ms。 Windows + Godot 4.7！ #Godot #gamedev #音声認識

Quando vi o que o Jev promete e apareceu o Laya, a versão open source da mesma ideia, eu pirei: tinha que testar num jogo, que é o meu cerne de estudo. Virou um Space Invaders ao contrário, humano contra a sua placa de vídeo. O seu PC é a Terra, você é o invasor, e eu tentei derrotar um modelo de pensamento rápido na velocidade da GPU. Como funciona: o Laya é um modelo pequeno (322M) que não escre