Hardoudou

I'm only going to use GPT for review, the model is really not good for professional dev. For small projects where we don't care about optimization, maintenance, or performance, why not; for math, why not.

GPT 5.5 That's it for me I'm done I'll never let it write code again. It's really not a good model for professional use The ridiculous things it does, the unnecessary complexity of the code just one prompt, I can see how unsustainable this is in the long run Claude 4.7 Easy Win

@melvynx Ta combien d'instance simultané ? Le x5 j'ai jamais réussi a atteindre le quota de 5h même en spamant non stop avec deux instances

We can all say it... Claude 20x is dead. The previous "feel unlimited usage" doesn't exist anymore. Usage increases quickly and scales really fast... People who discover Claude Code now: you missed a time that will never come back, I think.

@mateodr_ En vrai je penses que c'est même l'inverse il seront l'un des derniers en vie. OpenAI vont mourir avant Il se diversifie pas assez et les modèles ce démarque pas assez dans une catégorie précises

Anthropic va mourir dans 6 mois. Du moins c'est ce que j'entends partout sur Twitter. Car beaucoup pensent qu'ils vont être juste moyens partout et qu'ils vont s'écrouler. Moi je pense que c'est une stratégie. Comme ils ont cash infini + tous les tops profils qu'ils veulent bosser chez eux. Ils voient un peu leurs features comme des lead magnets. De un, ça permet de tester le marché. Voir ce qui prend ou pas pour après améliorer le produit. Certaines vont exploser. D'autres vont mourir. Mais ils ont les moyens de tout tester donc ballec. De deux, ça crée un momentum bullish. Ils font peur à 100% des startups. Et du coup, les VCs et investisseurs veulent mettre leur cash chez Anthropic et pas la petite startup du coin. Donc même si ils ship des features éclatées, ça attire l'argent et l'attention vers eux. Et quand ils trouvent ce qui marche, ils scale avec des ressources infinies.

I've made at least 120 commits in 10 days, and written tens of thousands of lines of code. I've never been this productive. And to produce code of such high quality Honestly, I can't get Codex to work

Between $20 of Claude and $200 of Codex, I prefer Claude even if I can't develop much with it. This month I got the $100 subscription for the first time and I have unlimited quota and I'm happy I've made so much progress, 20x faster than last month on my project

On very low-level tasks, Claude 4.7 has so far been giving me exceptional, well-structured answers. And it solves things that Codex never managed to do in 30 prompts. Claude 4.6 managed it by hacking something together, but the code was awful and only partially worked.

@theo 4.7 is giving me exceptional, well-structured responses so far. And it's solving things that Codex never managed to do once in 30 prompts. And Claude 4.6 managed to cobble something together, but the code was awful and only partially worked. It was a complex tasks

@theo In the month I've been using Codex, I haven't approved a single line of code. It's buggy and unusable. It doesn't code cleanly. It codes too much. And it codes in an overly academic way. Too many false positives. Claude 4.5, 4.6, 4.7 and far better without a doubt

Is Opus 4.7 the best model from Anthropic? No, that’s Mythos. Is it the best model we can use for code? No, that’s GPT-5.4. Is it the best model in the Opus line? No, that’s 4.5 It’s the best model released today I guess?

@RayaneRachid_ Enfaîte c'est assez marrant tout le monde a cracher dessus, et maintenant tout le monde en dit que du bien twitter est tellement bipolaire mdr Le modèles pour moi et super bon pour l'instant rien a signaler que des perfect pour l'instant

Les avis que je vois sur Opus 4.7 mdr c’est catastrophique pour l’instant moi cava il performe bien mais on verra si c’est le même pattern encore ou pas

@TheNabeelKhan @bridgemindai Do you have the $20 subscription? I don't know if it's my project's documentation that helps, or my way of working. for example, I provide the file paths when i prompt, which makes things faster and consumes less

@TheNabeelKhan @bridgemindai There's a really big issue with your session; I'm running 4.7 High and Fast And I'm sending multiple prompts per minute "Earlier, after what I'd say was 100 prompts over 5 hours, I was at 80%

Claude Opus 4.7 has been out for two hours and I'm already rate limited. 100% session usage. $200/month Max plan. New model, same problems. Anthropic shipped the best model on BridgeBench and you can barely use it. Some things never change.

github.com/Hardel-DW/exte… Produces the exact same worldgen, byte for byte. All the technical details are in the Readme. That represents approximately 120 lines of Java code. I hope this helps. It's difficult to work with the existing dimensions in datapack >.<

@Mojang @Minecraft An concept for reducing the size of the dimension data-driven. It shrinks the overworld from 8 MB to 250 KB. This allows datapacks to add/remove biomes to existing dimension without conflict. Cleaner Java code. A potential for Datagen. Here's the source code

@Kappaemme1926 I gave in to the $90 plan for the first time since the beginning, and upgraded from “2 prompts” to unlimited it's so awesome! I'm spamming nonstop I've even discovered multiple agents and I still can't get past 50% for weekly and session

Today I cancelled my Claude Pro subscription. Claude is insanely good. Probably one of the best models for coding and reasoning. But the usage limits run out way too fast. So I made a decision. For now I’m going all in on Codex only. Not sure if it’s the right move yet. But I’d rather master one tool deeply than constantly hit limits on multiple ones. Maybe it’s the right call. Maybe not. We’ll see. what do you think?

For Claude, With x5 I can’t even reach 50% of the 5 hour mark And he produces flawless code on the first try every time explains things perfectly. Plus he offers code that’s future-focused he takes the time to use official APIs or existing ones rather than reinventing the wheel

I canceled my Codex subscription and gave up, I was wasting too much time, even with the resets and unlimited quota. I’m losing so many hours of work just explaining things many times, It generates code that works, but it’s hardcoded really awful, not really suited for the future

For me, OpenAI is unusable, I have to revert it 7 or 8 times before I get a decent answer and even that’s far from perfect. It’s often not very practical. Codex isn’t good at explaining things it can do good research but does a poor job of explaining it.