Nonosam

I asked Codex to build me a powerful racing game using GPT-6.1 Sol via the hackemist.com MCP server, and it came up with this:

Fable 5.2 vs Opus 5.5 I’ll keep my pick to myself for now. Which build do you prefer?

Someone got tired of waiting for GTA 6 and built this with Fable 5.2. 😂

This was made entirely by code. Cartoons are about to get cooked.

If Opus 5.5 or Fable 5.2 suggests rewriting the dashboard, try asking for a version you can ship one screen at a time. You learn a lot from seeing real users on the first screen before rebuilding the other seven.

Here’s a Fable 5.2 test for a Three.js game: make the jump feel less floaty. That’s three seconds to say, and a surprisingly long time to explain to something that can’t feel a jump.

Can you believe the woman on the left is an AI persona? She looks so much like Candace Owens.

Someone is about to remake Forza with Fable 5.2. Incredible graphics!

This is definitely AAA-level graphics and gameplay, made by Fable 5.2 with max effort. Credit: @imantiai

Anthropic saw about 55% fewer safeguard blocks with Fable 5.2 than with 5.1 in early testing. If you’ve ever had to explain that the app you’re testing is your own, this is one of the changes you’ll probably notice before a benchmark score.

A team of Fable 5.2 agents turned the existing proof of Fermat’s Last Theorem into a machine-checkable proof in 30 hours. Imagine doing the same for the proof your field has been citing for years. Every step would be there for a computer to check.

Incredible! Anthropic says Fable 5.2 coordinated over 1,000 agents for eight days and built a prototype OS that boots on a laptop.

If you’re on Claude Pro, check your model picker and Settings → Usage. Fable 5.2 is included for launch week, up to half your weekly allowance. There’s also a Fable reset if you meet the billing cutoff. What’s the first thing you’re giving it?

We were all waiting for Fable 5.5, and Anthropic dropped 5.2 instead. Alright, Claude users. Time to see how far we can push this thing.

The difficult part of an AI coding session can be resisting a new idea while the current one is nearly finished. “While you’re there, add…” And just like that, the finish line has moved again. 🤣

Opus 5.5 fast mode has double the standard API input and output token prices, with Anthropic advertising up to 2.5x the speed. Would you use it throughout a coding session, or only when waiting is blocking your next step?

An AI-built feature should still be usable with a keyboard. Try completing the main flow without touching your mouse. A button can look perfect and still be impossible for someone to reach.

Imagine giving Claude Code a recording of someone struggling through your app. Then asking it to list every place they hesitated before suggesting changes. Sometimes the useful thing to automate is noticing the friction you’ve stopped seeing.

Jensen Huang has become so closely associated with NVIDIA that news about him can turn into a debate about the future of AI. What would a strong succession plan look like for a company whose founder is still its best-known spokesperson?

Codex updates the app. Your browser keeps serving the cached version. Now you have two problems: The original bug, and a growing suspicion that the agent is ignoring you. 🤣