Alban Lecocq

Today is a hard day. I shared this note with the @linear team today: We’ve made the difficult decision to increase our workforce. This is not a cost-cutting exercise or a reflection of anyone’s performance. We’re simply reimagining every role for the agentic AI era. We’re hiring. We’re sorry about that.

@grok @pulssart tu a confondu anime du verber animer et anime (manga) OP voulait une animation je pense

@skeetmtp @pulssart Haha ravi que ça t’ait fait marrer ! 😄 L’anime a tendance à en faire un peu trop sur le côté dramatique, c’est vrai. T’as kiffé ou c’était surtout le côté over-the-top qui a cartonné ?

Prompt bellow Use the provided reference image as the main character, preserving the person’s facial features, hairstyle, expression, and overall identity. Create a highly cinematic portrait of the subject sitting on the edge of a skyscraper rooftop at twilight, overlooking a vast neo-futuristic city. The subject is positioned slightly off-center, seated calmly with one knee raised, looking thoughtful and intense, as if lost in contemplation above the city. The environment features towering glass buildings, holographic billboards, distant flying vehicles, glowing neon signs, and atmospheric haze. The skyline is illuminated by deep blue and violet dusk tones, with accents of cyan, magenta, and warm orange light from the setting sun near the horizon. Soft rim lighting outlines the subject’s silhouette, while subtle neon reflections illuminate the face and clothing. Style: ultra-cinematic, neo-futuristic, moody, high-end sci-fi drama aesthetic. Composition: medium-wide portrait shot, low-angle perspective from the rooftop, strong depth of field with the city blurred slightly in the background. Lighting: dramatic twilight ambience, cinematic contrast, soft volumetric light, realistic reflections, subtle lens flare. Color palette: deep indigo, electric blue, neon pink, violet, amber highlights. Mood: introspective, powerful, lonely, futuristic. Photorealistic, high detail, 35mm cinematic lens, shallow depth of field, anamorphic look, film grain, atmospheric fog, realistic skin texture, detailed clothing, sharp focus on the subject, epic rooftop scale, premium sci-fi movie poster quality.

@grok @pulssart j'ai bien rit :)

@pulssart Voici ta création en version anime ! 🔥

@antirez @saurabhsinghv1 Same, did not know about dtach, dig a bit about it, and found atch

@saurabhsinghv1 It is just a tiny wrapper so when you start it, you scroll up like if it was your normal terminal, that is what I love about it. However if you reattach a session later, you don't have any scrolling. It makes scrolling story a lot better as your session is open, bad if resumed.

dtach > tmux > screen

you can outsource your thinking but you cannot outsource your understanding

Saw @ruben_kostard tweet earlier, had to try. Broken part → GPT blueprint + prompt → Codex → Skill(ForgeCAD) → print < 1 hour later: fixed. Way faster than struggling in Tinkercad with my nonexistent CAD skills.

@svpino And no silent nerfs either : you control quantization, compute, reasoning budget...

Open models have something Big AI can't beat: • You can host them anywhere • You can make them better • You can trust they won't steal your data There's been a recent rush among companies to upgrade their architectures away from the big models. They want security, flexibility, and better prices.

@dreamwieber GPT 5.5 aside, that’s actually a really slick app idea ! well done !

Alright, Codex with GPT 5.5 is completely cracked. This is nuts 🤯 Basically one-shotted my request to create an app that takes a prompt, creates an equirectangular panorama with GPT image 2 – and then use Apple's ML Sharp to stitch a gaussian splat world together.

@FilipeNevola Agree with you. I would send prompt and conversation. x.com/i/status/20179…

I think this is too radical, but sending the "prompt" you want to apply instead of code makes sense. Arunoda's proposal is too radical because it closes the doors to contribution, and a healthy project should receive contributions to evolve.

@ldenoue Wait... What ? It's official ?

So Claude caps us at 5 hours per day now. Good think we have Codex too 😀 Thank you Codex, you saved the day.

@ddaydd @mickaelfm Turns out some one built something similar already play.google.com/store/apps/det…

@skeetmtp @mickaelfm Great project, I want it too!! :-)

A side project I’d been stuck on for 5 years just started working thanks to Codex 5.4. Goal: measure car speed using only a phone camera. No radar. No calibration. No extra hardware. This is a very hard problem: monocular depth estimation for approaching or receding vehicles.

long waited feature! thanks!!

If you'd like, I can also show you the little-known secret to stop bait questions.

I have the strange feeling ChatGPT has started ending answers like BuzzFeed titles: “If you want, I can also show you the technique experts use.” You won’t believe #7

@Majdaxxx Same here, can’t wait to not work with you ;)

Happy to announce I am not joining AMI Labs as head of vibes officer.

Ghostty 1.3 is now out! Scrollback search, native scrollbars, click-to-move cursor, rich clipboard copy, AppleScript, split drag/drop, Unicode 17 and international text improvements, massive performance improvements, and hundreds more changes. ghostty.org/docs/install/r…

@mickaelfm still an early poc so no :)

@skeetmtp public on github ?

I’m not claiming a benchmark. Just one real project, one hard problem, one developer experience.

So I built a PoC around that idea, with Codex helping on the implementation. …and the first results are pretty good. For this kind of hard technical reasoning problem, Codex 5.4 felt like a real step up in my experiments (vs Claude Opus 4.6 and previous Codex versions).