DailyXplorer

I built Custom CSS Injector for myself because I was tired of CSS extensions that looked outdated, felt heavy, or simply didn’t feel modern enough to use every day. The goal is simple: a lightweight, accessible, clean extension to inject custom CSS per domain without unnecessary complexity. It also supports one-click export/import, so you can move browsers and keep all your custom CSS rules organized by domain. Link in description.

GPT 5.6 was already spotted !

Delete 1 from > Claude Code > Codex > Antigravity > Cursor

@thsottiaux Symphony ?

We will ship again this week. Codex has achieved escape velocity and will keep improving rapidly.

@Angaisb_ @OpenAI @ChatGPTapp @thsottiaux @sama

This is why I'm asking OpenAI to make image generation for Pro users 2k instead of 1k Look at the details, if you zoom in the first one looks weird but the second one doesn't (at least not as much) First is ChatGPT with Thinking, the second is API (2k, high)

A lot of people burn through Codex usage by defaulting to the newest model every time, but the $20 plan can go much further if you switch models based on the task. For pure coding, GPT-5.3-Codex still feels extremely capable and often more than enough. GPT-5.4-mini is the model I think people underestimate the most: fast, cheap, and surprisingly useful for small fixes, refactors, debugging, and explanations. Save GPT-5.5 or GPT-5.4 for the harder agentic work. For everything else, 5.3-Codex and 5.4-mini can carry a lot of the workload without eating your limits too quickly.

developers.openai.com/codex/speed #image-generation-usage-limits" target="_blank" rel="nofollow noopener">developers.openai.com/codex/pricing?…

Useful Codex usage tip: speed, images, and model choice all affect your limits more than it may seem. Fast mode is documented as ~1.5x faster, but it costs more usage: GPT-5.4 Fast uses 2x, while GPT-5.5 Fast uses 2.5x. Codex image generation also uses gpt-image-2 and counts against your general Codex usage when using a ChatGPT subscription. If GPT-5.5 is your 1x baseline, GPT-5.4 is roughly 0.75x, GPT-5.3-Codex around 0.5x, and GPT-5.4-mini around 0.25x. Use GPT-5.5 when quality matters, but switch down for routine tasks if you want your limits to last longer.

I built Custom CSS Injector for myself because I was tired of CSS extensions that looked outdated, felt heavy, or simply didn’t feel modern enough to use every day. The goal is simple: a lightweight, accessible, clean extension to inject custom CSS per domain without unnecessary complexity. It also supports one-click export/import, so you can move browsers and keep all your custom CSS rules organized by domain. Link in description.

AI image generation still gets criticized as if it only produces flat, uneditable images you can’t really work with after generation. But that argument is becoming weaker every month. With tools like Canva Magic Layers, you can now turn a static generated image into editable layers: objects, text, colors, layout, composition. The real shift isn’t just better images. It’s that generation is becoming an editable workflow.

@Nate93260 @BastiUi C'est bien réalisé en plus

@DailyXplorer @BastiUi J’suis mort de rire c’est exactement un taf que j’ai dû faire pendant mes études

Le nouveau logo de Lacoste, vous en pensez quoi la team ? 🤔

@clestise @BastiUi Oui, c'était du troll

@DailyXplorer @BastiUi T’es au courant que le logo simplifié c’était l’ancien hein?

@delveroin Shooty, an application that completely replaces Shottr for taking full screenshots on my Mac.

WHAT DID YOU BUILD TODAY? Post your work

For me, it’s GPT Image 2. Not because it wins every single niche use case, but because it’s the most useful overall: better realism, stronger understanding of real-world context, cleaner prompt following, and it’s already part of my ChatGPT Pro subscription. The fact that it’s also integrated into the ChatGPT/Codex ecosystem makes it even harder to beat. It’s not just an image model, it fits directly into the workflow.

Be honest, which is the best Image Generation model?

🚨22,36€ d'épargne de précaution💸 ❓Est ce que c'est une bonne idée d'avoir peu d'épargne de précaution ? 🤔 Garder trop de cash, c'est peut être juste une façon de se rassurer. Moi, je suis investi au maximum 📈 Curieux de voir vos avis 👇

DeepSeek V4 Pro is not the model that makes premium AI subscriptions obsolete. It is the model that makes them harder to justify on price alone. It still isn’t my first choice for serious software engineering. For coding, I’d still take GPT-5.5. But DeepSeek keeps proving that high intelligence can be delivered at a much lower cost than people expected. That changes the market. OpenAI, Claude and the rest can still charge more, but they now have to justify it with reliability, tooling, agents, product depth and real workflow value. The price of intelligence is going down. The price of trust, execution and ecosystem is what remains expensive.

@R2Cdev_ @Angaisb_ x.com/DailyXplorer/s…

Sonata today? @Angaisb_ ChatGPT music????

@ai_for_success x.com/DailyXplorer/s…

OpenAI Drop GPT-5.5.