@0xCodez Fable reviews things about Fable:
Verdict: ~80% real, ~15% embellished, ~5% flat wrong — and almost nothing we don't already run. This is a competent remix of the same official pages I pulled Tuesday, plus two real Anthropic experiments, wrapped in newsletter urgency.
Fable has a huge false positive problem. And the problem is it flips to 4.8 over something benign in the middle of a goal/loop. Where it matters- hour 1 or hour 10.
9 hours of Opus over Fable is stark.
@Adam__Allcock for the non technical- sign in to Chatgpt on your codex browser /goal codex report into gpt for x,y,z check every 20 seconds for response.
and then watch two models bubble wrap bubble wrap
@mertdumenci 4.6 until the last two or three weeks- excellent. Codex has had the crown since. last two weeks of 4.7 were extremely brutal. Honestly, Codex compacts super often vs Claude and I think thats why it can run 20 hours and have like 100% thread recall vs like 30 mins on 4.7