Jev for robotics — a daily feed

I just built a Jev X Swarm Controller and put 300 agents on autopilot for 6 minutes. 300 K3 workers on autopilot. 6 minutes. 12,480 turns. Jev made a decision on every one of them. 91 reached me. A swarm knows how to execute. It does not know when to stop. every turn gets 6 typed questions: is this result good enough, is there new information, is this the same state as last turn. then one action:

Last week, @typesafeai launched an early access version of Jev, a new decision-making model. Now that it’s available to everyone, I want to share what I found after plugging Jev into @openmind’s OM1 and using the model to navigate our Go2 on @nvidia’s Isaac Sim. https://t.co/P3zOc4O2xQ

Jev just put Jeb into orbit! I wanted to see if Jev could control a rocket in KSP, turns out it can! Jev controlled pitch, throttle and staging. Jeb did his part too by smiling a whole bunch. thank you @typesafeai for giving Jeb its AI friend. https://t.co/URDQhPNRck

@typesafeai we are tested Jev as the decision maker for a Unitree G1 humanoid in Automata Room. The task: carry 4 crates to a storage rack, one at a time, then return to the entrance. Fewer, shorter trips score better. Full run below 👇 https://t.co/xpliAUN5fR

Built a System 1 + System 2 robot brain using @typesafeai's Jev in PyBullet. 🧠 Claude: Generates the high-level plan to stack cubes. ⚡ Jev: Acts as the split-second reflex layer, evaluating sim states in <100ms to adapt. Fast, schema-safe neurosymbolic robotics in action. https://t.co/WVNB1aqo7n

Can Jev drive a car despite not being able to see? I gave Jev access to an unreleased world model used to train self-driving cars and a local segmentation model to see if it could drive autonomously. TLDR: Kinda? It's a little Crazy Taxi - but kinda amazing this works at all. https://t.co/NKoe7RqFIe

AI Solutions Lab | Researcher: Luca 寝ている間に、Claude Code にロボットアーム(SO-101)の実験を任せた。 一晩で172回、掴んで運ぶ実験を自分で設計して回していた。100回連続の掴みは成功97回、1回17.8秒、途中停止なし。集計も不具合の修正も自分でやっていた。 ・Jevのユースケース 「ピンクを含む物は左へ」「丸い物は必ず中央へ」のような言葉のルール12種で、仕分けの行き先を判断させた。俯瞰カメラが位置と色を見て、ローカルのQwen3.5 4Bが種類を答え(0.8秒)、Jevがルールを当てはめる(0.6秒)。72回すべて、選んだ区画へ運べた。ルールが答えを定めている65回は全問正解。 弱かったのはJevに渡す側の目だった。同じボールが置き場所によって「緑」「ピンク」と別の色に読まれ、VLMも71回中17回「フィジェットスピナー」と答

LE HABLO A UN BRAZO ROBÓT IGUAL QUE A UNA PERSONA 🤖🚀 Usando @typesafeai Jev el robot ahora puede entender frases como Abri la mano Anda al techo Estirate lo máximo que puedas Demo arranca en 1:25 Venimos a pleno preparándonos para el @ArgentinaSpace https://t.co/qOTkch12gN

If this gets traction, I'll open-source it on my GitHub https://t.co/kQTfGvPVQy (and post again on X to notify). I've been using it for real work in Codex CLI. It uses Codex CLI's experimental per-step settings to change reasoning effort before the next GPT-6 generation. After tool results, Jev gets the task, published reasoning/progress summaries, and the last six tool calls with bounded outputs.

What happens when the target moves mid pick-and-place? Open-source, Jev-inspired sim: local 2B model + PiPER robot in MuJoCo. Cube relocated 3x per run; robot detects, interrupts, recovers. 10/10 seeds succeed, 32.6 ms median decision latency. https://t.co/HIdSDIXgqh https://t.co/cr60W65388

Jev is INSANE. I connected it to a bio-inspired robot running off a simulated nervous system. Neural activity gets fed into Jev, Jev makes a rapid decision, and that decision gets translated into physical movement. https://t.co/EH6CmHezqm

Just for fun - I give the station a priority. Jev reads the situation and chooses the drone’s next action. Built on @Replit with TypeSafe’s Jev and native WebGPU. Jev assesses the station, evaluates evacuation risk and urgency, and chooses the drone’s next action. I’m controlling the camera. In this run: 208 ms average Jev response, including network time. All 12 crew rescued. No email classificat

We moved a cup while our humanoid was stacking a cup pyramid. Jev from @typesafeai chose the next move: set down the held cup, replan, finish the stack. 6 recovery options. 65% confidence. https://t.co/if6e4TEBNP

Implemented with #Openjev using #Codiv for Autonomous Self Driving Simulation. Based on https://t.co/X4aAbzRADY #Jev Track the decisions Code (Apache-2.0): https://t.co/tpfbk5wsjk Free hosted API: https://t.co/nQX2S6urCY https://t.co/rnWEY8N5kg

i've always wanted to build an autonomous robot, but I was always missing the last and most important piece A REAL fast decision making AI, which LLMs could't solve yet And Jev came to solve the problem, please take a look at my baby's first steps: https://t.co/1xEu2iArGJ

i would trust Jev 100x more than ANY other indian driver out there so i built a driving game, added traffic, and plugged in jev, so it drives me around while STOPPING for pedestrians, cars crossing, and even slows speeds for corners or potholes Jev controls stop/yield/slow/proceed decisions; local code handles steering and collision checks. The HUD shows live JEV status. cost?? $0.0208 for driving

AI parks better than me. A bit late to the party, but here is my experiment with Jev. It manages to control the car and find the optimal maneuvers to complete the challenge. https://t.co/jR5JMyCs8c

現実世界の物理デバイスをJevで制御する実験をしてみました。#Jev はロボット制御 #フィジカルAI のように言葉で出力する必要のない応用にこそ向いていると思いましたhttps://t.co/9jDRMWzjHm

Jev beat a fly at pool, so naturally I promoted it to managing 160 simulated robots. https://t.co/gqcFepd7jq

I TURNED JEV’S TYPED DECISIONS INTO A GHOST. Built JEV BOARD, an Ouija-style board powered by TypeSafe’s JEV. The use case was simple: give JEV a structured set of possible next words, let it choose the strongest path with probabilities, then make the planchette physically spell that choice across the board. Every reply is built through real decisions inside the app. The board gives those decision

https://t.co/VOmHHfTYeH Jevを使わずに制御してみた！！！！天才かも！ ドパガキモードもあるよ！ https://t.co/mZVEH0JNDc

Using Laya/Jev to control a 2D robotic arm https://t.co/z1yOfoHvfB

I made the Fusefly's vision use stereoscopic vision, so two virtual camera on it which generate a depth map in real time and then uses the images and depth map to do object detection to then feed it's precieved object distance and location into Jev to control the bug. https://t.co/BLOqMYugli

把 Jev 接入家里小米音箱控制智能家居了,DeepSeek 做兜底, 比单纯 DeepSeek 调脚本快多了 https://t.co/86tSnrNO09

話題のJevを使って半リアルタイムで演奏判断させて延々と即興演奏させてみる実験。 とりあえずそれっぽくなったけど、まだ色々改善できそう。 Jevにどこまで何を判断させるかの設計が重要。 音源は先日Astraに作ってもらった物理モデリング音源の改良版。 https://t.co/zO1rHGSh1O

Ok I got Jev + SAM 3 working on a simulated mobile robot! SAM 3 + camera depth provide 3D observations, alongside joint state and contact feedback. Code samples numerical movement, grasp and placement poses; parallel Jev calls rank them, then a final Choice picks the action. Noul and Score add supporting judgments, while deterministic controllers handle IK and motor control. It’s almost through th

Ok I tried the Jev model for a robotic simulation. Observe: MuJoCo provides depth points/ASCII maps, object locations, joint positions, velocities, actuator loads, finger contact and slip measurements. Interpret in parallel: separate Jev calls assess the task, geometry and contact/dynamics. Noul judges conditions, Score assesses priorities and stability, and Choice selects intent and focus. Genera

@typesafeai Jev controls an iOS simulator using @callstackio’s agent-device Works 2-4x faster than Sonnet 5 The goal is to offload device control to Jev: Claude sets the goal and verifies the result, while Jev handles the interaction - with up to a 4x speedup https://t.co/tN9x0NKG4S

I gave JEV a reactor to operate A fictional RBMK-inspired simulator with live physics, 16 parallel decision lanes and real actuator control: rods, pumps, flow and steam https://t.co/Kfs488rF4Z

Jevくんを使って、ロボットが人間の指示を実行して良いかを判定する、をVibeCodingしました。 #Jev https://t.co/oFBCnpntUQ https://t.co/WHq20xNXV7

Ya tenía la esp32 y el modelo 3D imprimido y como hace nada salió Jev de @typesafeai dije: esto es clave para que los robots tomen decisiones en tiempo real no? Así que le pedí a hermes que me modificara el programa y le añadí Jev + deepseek de @helmcode para sus respuestas 🤖 https://t.co/oLRcRhzN6d

Most models answer by writing. This one answers by choosing. I put Jev in a real-time control loop — a car that crashes if the answer arrives late — to see what that difference actually buys. 478 decisions. 2,546 m. Zero crashes. 🧵 https://t.co/RvOKiAnP81

#Jev, @typesafeai’s System One model as the nervous system of a simulated body. 10×10 m room. 2–5 decisions/sec: what to do next, and how to move doing it. No prompts, no parsing, no text. Just typed judgments with probabilities my code acts on. You can watch it hesitate. #RLCD https://t.co/DxS6p2D16T

使用TypeSafe的jev的决策能力，用文字指令控制机械臂执行任务，速度杠杠的。PS：这运动学控制算法和建模，是用gpt6 astra帮我生成的，太强了。 https://t.co/Ge4Z5B0PZx

We pointed TypeSafe's Jev at a warehouse robot fleet instead of another browser demo. 300 real API calls, 300 succeeded, p50 0.53s. Total bill for the run: $0.00737. https://t.co/tbJ9JtLnBQ

I have #JEV wrapped with my engines and model work as 2 models controlling drones and packs. https://t.co/35jO9i4VbC