TypeSafe’s Jev is an AI that never writes a sentence
They built an AI that can’t chat on purpose.
AI stops talking to humans and starts acting like code that already knows the answer.
What Happened
TypeSafe AI blog; Hacker News (1,512 points)
Daily
Today's signals, kept for later.
They built an AI that can’t chat on purpose.
AI stops talking to humans and starts acting like code that already knows the answer.
TypeSafe AI blog; Hacker News (1,512 points)
Your browser hired an AI that is legally not allowed to keep the receipts.
The browser as the last place that still belongs to you, now with an AI contractually forbidden to remember.
Mistral and Mozilla joint announcement
The AI is still talking to you while it’s already calling the tools.
Conversation itself becomes a cover for computation.
Google blog; Gemini API documentation
Your camera just started keeping a secret original that AI can’t fake.
Truth now requires a hardware-rooted sacrament; “pics or it didn’t happen” gets an encryption layer.
Apple Security Research
The robot would rather sit on the floor than touch you.
Robots perform submission so humans do not have to wear cages. Work, bodies, and politeness become safety engineering.
Agility Robotics; Ars Technica
The birds outside your window are being painted by ghosts.
The anti-slop future: local AI, dead artists, and a slow physical object that makes the present feel nineteenth-century.
Fugleramme GitHub; arnegiacomo.dev; Hacker News (1,728 points)
You told ChatGPT it was just between you two. Someone in a cubicle rated it a four.
The confessional booth has human interns.
404 Media investigation; PCMag; The Decoder
Treat as an investigative report and distinguish contractor access, retention settings, and any promise of confidentiality precisely.