🔓 Binance loses record $6.3B in hack tied to North Korea

👗 Meta quietly launches an AI fashion app

📄 Gemini now lets you edit files right in chat

👀 Claude is now policing its own conversations

🚕 Waymo and Toyota to test robotaxis in Japan

+ 📊 Daily poll and results

+ 📈 Trending tools and resources

Hackers drained about $6.3 billion in crypto from Binance’s hot wallets across 10 blockchains, the largest crypto theft on record. Binance suspended all deposits and withdrawals, and co-CEO Richard Teng promised to cover all losses.

Binance’s early probe points to a compromised engineer’s laptop, though it has since found more entry points. TRM Labs and Chainalysis blame North Korea’s TraderTraitor, the group behind 2025’s $1.5 billion Bybit hack.

Stablecoin issuers like Tether and Circle helped freeze nearly half the haul. Still, Binance owes users an answer on why one engineer’s laptop could reach systems controlling billions.

Meta quietly launched Style Sketch, an AI app that builds a virtual closet from your Instagram photos or uploads and suggests outfits. It’s live only in Indonesia, Singapore, and Thailand for now.

Zuckerberg told employees Meta plans to ship far more apps as AI makes building them faster. Style Sketch follows Waddle World, an AI pet app, plus Pocket for vibe-coded games and Vibes for AI videos.

The idea isn’t new, since Alta and Google Photos already digitize closets. Meta’s edge is your Instagram grid, a style archive you’ve built without realizing it.

Google now lets Gemini users generate, preview, and edit PDFs, Google Docs, Sheets, slide decks, Markdown, text files, and code inside the chat. It’s available to everyone, though only PDFs work on mobile.

Gemini learned to generate files in April, but editing meant leaving the app. Now users can highlight text or a table and ask Gemini for changes, like turning that table into a graph.

Anthropic shipped document and slide editing in Claude chats earlier this year, so Google is playing catch-up. Gemini’s edge is home turf, since it already sits inside Google Docs through the Gemini bar.

Anthropic unveiled “self-assessment,” which has Claude review each conversation right after replying and flag potential misuse with a harm category and priority score. It has run on much of Claude’s traffic since June.

Anthropic says the check adds under 2% to compute because Claude holds the conversation in memory. On Claude Fable 5, it matched 94% of high-priority cases a stronger classifier flagged, at 1% false positives.

It surfaced 100,000+ child-abuse flags from one API customer, which Anthropic reported to authorities. But in red-team tests, it only cut attack success from 79% to 54%, so it’s a tripwire, not a wall.

Waymo and Toyota will start testing a fleet of self-driving Toyota bZ4X electric SUVs in Japan in 2027. Next, they plan to bring the cars to the UK and other right-hand-drive markets.

Toyota will build the cars and install Waymo’s self-driving hardware at its Takaoka plant in Toyota City. The deal builds on a preliminary agreement the pair signed in April 2025.

Toyota is playing both sides of Tokyo’s robotaxi race. It also backs Nuro, a Waymo rival that began testing self-driving cars in Tokyo earlier this year.

A) Yes - 54% 🏆

B) No - 46%