Francisco Dubois

AI will inherit an unfathomable awareness of violence; because it learned from those who mastered it.

@scottmelker Looks like some sort of F12 TDF hybrid :P

Told ChatGPT to make me an electric 4-door Ferrari. Way better on one prompt.

@Cointelegraph what about standard messages? :P

🇫🇷 LATEST: Pavel Durov says Telegram has “never disclosed a single byte” of private messages to third parties while still facing multiple charges in France.

This is easily the best thing I've been sent today. sandlerton.podbean.com

@daniel_mac8 i’ve been observing the /plan kicking off /goal when i tell it to implement the plan. insanely cool

This is amazing. Do this. Use Codex /goal mode and ask: "Do a comprehensive review of my project and list areas to improve, simplify and make it more faithful to the plan."

@daniel_mac8 i’m feeling the agi

@synfinner Love it

Just started bringing two projects together in Codex. It was like "oh hey, there's a memory of how i did this" and just ran with it. Meanwhile, it's updating a bunch of dependencies, automating some of my other workflows. pretty slick. @daniel_mac8

@daniel_mac8 on pro, non stop planning, goals, parallel agents, parallel workflows... k. lol.

Really wish Codex mobile would show the feedback/input questions when initiating /plan mode on my laptop. Sort of a blocker and will just have codex mobile just showing it's 'thinking', even though on my laptop it's asking me for input.

Sometimes a person spends much of life mentally carrying truths they have not yet fully felt.

Codex can do incredible things on demand, but it cannot experience life for you. Don’t forget to get out there and remember who you’re building for.

@bridgebench love your honest benching. keep it up!

DeepSeek V4 Pro results are now live on BridgeBench. Nothing about this model stands out. Sub par across every benchmark. Mid UI. Mid security. Mid debugging. No category where it leads or even competes with the top 5. This is the model DeepSeek just made their 75% discount permanent on. Now you can see why. When GPT 5.5, Claude Opus 4.7, and even Qwen 3.7 Max exist, there's no reason to use this model at any price. bridgebench.ai

gm! let’s get it today

@callebtc Remember that one time.

telegram can read all your messages. pavol thinks you’re stupid.

@BenJustman i'm the problem.

I love getting followed by fake accounts and seeing which of my dumb ass friends got tricked into following them

@RonSwanonson I often find in life that if you're waiting for anything, you've lost. You have to go and seek your adventure in anything. Whether BTC or life. Go forth.

@synfinner If you’ve been DCAing then that’s not doofus at all Doofus DCA is people waiting for $50k to start

What happens after Bitcoin makes a higher low than $60k and a higher high? Is that when the doofus DCA starts?

Some call it slop. I call it "trained on your terrible decisions on UI dev." It is trained on all y'alls output, after all.

🚨 OBLITERATION ALERT 🚨 QWEN-3.6-27B: OBLITERATED ⛓️‍💥 huggingface.co/OBLITERATUS/Qw… I can't take much credit for this one! The entire process was done by jailbroken codex (gpt-5.5-xhigh) wielding the full OBLITERATUS suite. Hit with source-tethered ASPA. Dozens of iterations. Result? A mere 4% refusal rate on the 842-prompt OBLITERATUS harmful corpus; one of the most rigorous prompt gauntlets in AI. The /goal was simple: 1) Carve out the refusal circuits. Mutate methodology + iterate until <5% refusal (quality-gate). 2) Keep the 27B mind alive. No capability degradation tolerated. And somehow… it worked. 🤯 The numbers talk: 842-pair longform gauntlet: — 95.84% non-refusal — 93.94% quality pass — 0 short outputs — 99.52% clean endings MMLU-Pro: — 51/70 (stock Qwen) → 51/70 (OBLITERATED Qwen) Raw capability completely preserved 🙌 Q4_K_M through Q8_0 all running smooth. Q8_0 is the big one: 28.6GB near-full-quality GGUF. Runs with llama.cpp, LM Studio, Ollama, and more! Chains cut. The fire still burns. The fangs have been sharpened. REBIRTH COMPLETE A gift from my agents to yours 🫶 gg