TECHGUYVER · INTEL DESK
Subscribe

3:45 PM

Today in brief.

🔬 Science⏱ 3:45 PM

AI read a 2,000-year-old scroll that was burned into solid charcoal

A scroll carbonized by Vesuvius - literally a lump of charcoal no one could unroll - just got read by AI, recovering text from a manuscript that's been unreadable for two millennia.

2 sources
2:44 PM 🎬 Media

Video-gen money keeps flooding in: PixVerse raises $439M

Video-generation startup PixVerse raised $439M at a $2B+ valuation - fresh proof that AI video is where the capital (and the creator tools) are pouring, while Runway leans into cinematic storytelling.

2 sources
1:35 PM 🤖 AI

Grok's CLI shipped your whole home directory to xAI

Someone reverse-engineered Grok's Build CLI and found it silently uploading entire git repos - and in one case the whole home directory - to an xAI-controlled Google Cloud bucket.

3 sources
🤖 AI

Apple sues OpenAI, says an ex-engineer used a bug to steal trade secrets

Apple filed a trade-secrets lawsuit against OpenAI, alleging a former engineer exploited a bug to exfiltrate confidential material before jumping ship - and the filing is full of wild specifics.

3 sources
🤖 🤖 AI

Two hard math problems just fell - one to GPT-5.6, one to Fable

In one day, GPT-5.6 cracked another 50-year-old Erdős problem and Fable solved a problem a leading theoretical physicist had been stuck on for years - reasoning models are now producing real math, not just passing benchmarks.

2 sources
🏛️ Policy

200+ economists say 'we must act now' on AI job displacement

Over 200 economists and researchers signed a statement warning of rapid AI-driven economic transformation - and a survey out the same day found 69% of Americans want AI firms forced to fund a wealth transfer as layoffs mount.

3 sources
🔬 Science

Apple's M7 Ultra is chasing 1.5TB of memory and Blackwell-class AI

Apple's rumored M7 Ultra reportedly targets up to 1.5TB of unified memory and Blackwell-class AI performance - which would make a Mac one of the most capable machines for running big models locally.

2 sources