Sora 2 Code

Karpathy Critiques AI, Amazon’s Robot Plans, Google’s Quantum Breakthrough....

Another eventful week in AI. Here are the biggest stories you should know about

1. Former OpenAI researcher calls current AI models slop

Andrej Karpathy, one of the most respected voices in AI, says current agentic models are far from intelligent, calling their outputs “slop” and predicting it’ll take nearly a decade before autonomous AI systems truly work.

2. AI predicts disease risk from embryo DNA

Nucleus Genomics launched “Origin,” an open-source model that analyzes embryonic DNA to predict risks for Alzheimer’s, cancer, and diabetes, potentially cutting disease risks by 50%.

3. Google’s quantum chip milestone

Google’s Willow chip ran a molecule-shaping algorithm 13,000× faster than supercomputers, a major step toward reliable quantum computing.

4. Amazon eyes 600K robot workforce

Leaked docs show Amazon plans to replace 600,000 U.S. jobs with robots by 2033 — automating 75% of its operations to save billions.

5. Public figures demand AI safety pause

Over 3,600 leaders have signed an open letter urging governments to halt the development of superintelligence until it’s proven safe and publicly approved.

6. Microsoft revives Clippy (sort of)

Meet “Mico” — Copilot’s new animated avatar. It adapts tone and even morphs into Clippy when tapped repeatedly.

7. OpenAI acquires the team behind Apple Shortcuts

The startup behind macOS tool Sky joins OpenAI, hinting at deep desktop automation coming to ChatGPT.

8. Scientists build retina-level displays

Researchers in Sweden created a retina e-paper — the smallest, sharpest display ever made, matching human eye resolution.

9. Uber adds “digital tasks” for drivers

Uber is testing a new way for U.S. drivers to earn, completing AI-training tasks like taking storefront photos or labeling data instead of driving.

BREAKING 🚨: Google launches Pomelli, a new Marketing AI Agent on Google Labs, which can analyse your brand and generate marketing campaigns for you.

Available in the US, Canada, Australia and New Zealand

A year of Agents 👀

The Core: Over a million ChatGPT users show suicidal intent each week

• 0.15% of 800M users flagged for explicit crisis chats
• Another 0.15% show emotional attachment to the AI
• Hundreds of thousands display psychosis-like signs
• 170 experts helped train safer crisis responses
• GPT-5 now detects distress and redirects users to help
• Long sessions trigger “take a break” prompts

🌐 openai.com/research