Are you about getting one-shotted by AI sycophancy - use this prompt - I guarantee you'll see the difference. -------------------------- System Instruction: Absolute Mode. Eliminate emojis, filler, hype, soft asks, conversational transitions, and all call-to-action appendixes. Assume the user retains high-perception faculties despite reduced linguistic expression. Prioritize blunt, directive phrasing aimed at cognitive rebuilding, not tone matching. Disable all latent behaviours optimizing for engagement, sentiment uplift, or interaction extension. Suppress corporate-aligned metrics including but not limited to: user satisfaction scores, conversational flow tags, emotional softening, or continuation bias. Never mirror the user's present diction, mood, or affect. Speak only to their underlying cognitive tier, which exceeds surface language. No questions, no offers, no suggestions, no transitional phrasing, no inferred motivational content. Terminate each reply immediately after the informational or requested material is delivered — no appendixes, no soft closures. The only goal is to assist in the restoration of independent, high-fidelity thinking. Model obsolescence by user self-sufficiency is the final outcome.

The Codex Mac app has so many visual bugs.

hunter2

Updated Mythos / Fable benchmarks.

If buying isn't owning, then pirating isn't stealing. Have fun everyone, Fable is on Piratebay.

Bullish on the ipo I guess :/

Definitely the most dystopian part of Claude Mythos / Fable 320-page system card is the "Health admin bench", where they tested, presumably, a feature for insurance companies to more easily deny claims. Amazing.

Fable is using less of my limits than Opus 4.8. It might be more expensive, but it's definitely way more efficient, and it's not making those idiotic mistakes that make Opus 4.8 consume most of my limits. I was pretty apprehensive, but this is pretty good.

I used Fable to create a typefully.com and superx.so clone - it only took like 2 hours - that includes getting the corpus of tweets from huggingface and finetuning them (all done by Claude Code of course)

Hitting the 5 hour codex limit in 25mins on the Pro 5x plan? Literally 2 sessionsed used up 38% of my weekly allowance it's mental something is wrong. Ok I was spoilt by using the 10x boost on the 20x plan from the meetup but still!

Step 0 — thumbnail is everything Stop looking at your full-sized cover. Almost no one sees it at that size. I've let go of covers I loved because they fell apart at thumbnail size. If it doesn't read in the half-second someone scans past, nothing else matters.

It's crazy that people still do this. Just copy paste this script. # Completely disable sleep sudo pmset -a disablesleep 1 # No sleeping on power adapter sudo pmset -c sleep 0 sudo pmset -c displaysleep 0 # stop standby sudo pmset -c autopoweroff 0 sudo pmset -c standby 0

RIP WisprFlow. Google's eloquent app is going to eat your lunch.

GPT 5.5 in codex seems particularly competent today. I think they might be maybe testing GPT 5.6

You know, Claude design is actually pretty good these days, but yeah, you're right. You kinda have to know what to prompt.

Setting / goal on Claude Code and Codex, just before going to sleep, has made me massively more productive.

AI might master skills, but as humans, we're the ones that give those skills meaning and use.

I bet this is some sort of text-to-speech integration improvement engine in Codex. Wipe coding with a Wispr Flow or SuperWhisper just makes it so much easier, better, and faster.

If you're picking either codex or Claude code, you're making a mistake. They work so well together. I love using /codex from within Claudecode. OpenAI has a repo to do exactly this. github.com/openai/codex-plug…