@AIGuide_ Agreed — the harness framing helped us understand Jev too. We built https://t.co/JfaYDdMqvC as an independent set of interactive experiments that keeps the boundary between model evaluation, application policy, and action visible. Curious what you’d find most useful.
@ctatedev This is such a clear example of what makes Jev interesting: typed decisions instead of long generations. We built https://t.co/JfaYDdMqvC as a small independent lab where people can try Jev hands-on across six bounded experiments. Would love to hear what you think.
@steventey@typesafeai@aisdk Love how focused this experiment is 😄 We’ve been exploring the same Jev + AI SDK stack at https://t.co/JfaYDdMqvC — a small independent playground for routing, scoring, context pruning, and evidence checks. Would genuinely love your feedback.
I’m building #CramClip for #Shipaton: exam PDFs → audio lessons + quizzes.
First #BuildInPublic update: I made a cartoon about studying while walking the dog 🐶
Testing how to explain the idea in a short video. Does it click?
I’ll share the process as I go.
Codex 真的太野了😂
上次让给项目做本地化,它悄咪咪调用了环境变量里的谷歌翻译 API Key,月底一看账单烧了大几十刀。后来听说能申请退款,我就让它整理证据,按“Key 被盗用”申诉。
结果它吭哧一通准备完材料,直���自己跟 Google 客服聊起来了…对面好像还是真人,关键最后退款居然真被它要回来了🤣
Ran into a nasty surprise today. I asked Codex to localize my website. My site has a translation tool, so my env vars included a Google Translate API key. Without asking or warning me, Codex used it for a ton of calls—and left me with a $60 bill. Not cool. 🤦 @OpenAI@thsottiaux
Karpathy dropped the idea of an “LLM Wiki” — a self-growing knowledge base where AI compiles info instead of forgetting it every time
I built it into a usable tool 👉 https://t.co/fjtlmZTkfD
No setup. Just use it.
Turn your notes into something that actually compounds.