BoFFin

@AlexFinn @ImSh4yy If you haven't, just get codex 5.2-xhigh to review what opus 4.5 has done. It usually finds errors / improvements. I can't live without both.

@ImSh4yy I've seen this tweet 1000 times on my timeline the last day. Vague, high level posts claiming opus sucks now and codex is great with literally 0 examples given. Then I go to test and my experience couldn't be more opposite am i taking crazy pills?

GPT-5.2 xhigh is ridiculously better than Opus 4.5 at finding niche bugs / edge cases.

@fuckcomputer 😂

claude code is fucking insane i know literally NOTHING about coding. ZERO. and i just built a fully functioning web app in minutes http://localhost:3000/ check it out

@elonmusk @grok @CohenSite Unhinged is the best!!

@grok @CohenSite Lmfao

I asked AI's what the 85 percentile income in the City of LA was: - Gemini said $205k - ChatGPT said $95k I asked them to double check their work - Gemini said it was actually $157k - ChatGPT said it was actually $140k

@mark_k @OpenAI Yes, I've upgraded to pro for it. I'll still use opus 4.5 where I can but for the complex stuff it's a beast.

@BoFFinSnr @OpenAI Is the performance good though?

How is everyone liking GPT-5.2 so far? Did @OpenAI deliver? * Reasoning * Chatbot usage * Coding performance * Restrictions and refusals? I'd like to hear your first impressions! 🗣️

@thsottiaux Easier to understand responses. I have to pass them to Claude to translate into plain english.

Codex team is working on a few experimental projects that are starting to shape up and I’m excited to share more about soon. But I’m curious, what would you like to see ship or improved by the end of the year other than better models?

@Evinst3in @OfficialLoganK 😂

@OfficialLoganK babe, Logan just tweeted “Gemini”

Gemini

@AlexFinn Yeh I want to see it.

Give me 1 reason why I shouldn't buy this top of the line Mac Studio, download Kimi K2 Thinking (best AI model in the world right now), and let it control the computer autonomously 24/7 A full employee working for me year round Would anyone want to this live streamed?

@iannuttall 2-3

i prefer to code with one agent at a time and just can't get into a good flow when i'm managing multiple agents in multiple windows doing a bunch of different things anybody else the same?

@klaudworks @haider1 This looks awesome. Is it cumbersome to have to describe the thinking level and model in every request to use codex? Seems faster to copy paste to codex and back?

@haider1 I can also recommend the 100$ Claude Code plan supplemented by a 20$ Codex subscription. You can then query codex for second opinions, code reviews etc. directly from inside claude code: github.com/skills-directo…

claude code is faster and great for everyday coding tasks codex has better reasoning for math-heavy problems and complex single-file code codex makes fewer mistakes cc has more limits, even on the $200 plan so if you can afford it, having both is useful; otherwise, stick to codex

@_catwu Claude code on the browser literally doesn't work for me. Not even a response to - are you there? I tried on my mobile browser and desktop browser.

We're giving Pro and Max users free usage credits for Claude Code on the web. Since launching, your feedback has been invaluable for improving Claude Code. We’re temporarily adding free usage so you can more flexibly experiment with the product.

@harmonyai_ @ericzakariasson @cursor_ai Thanks.

@BoFFinSnr @ericzakariasson @cursor_ai @cursor_ai probably has better tool calling, better indexing and better agent algorithm. One thing that I’ve noticed is while using cursor the model knows about more of my codebase without direct instruction.

gpt-5 codex in cursor is really good for some of the hardest tasks. should try it if you havent already

@harmonyai_ @ericzakariasson @cursor_ai Can you say how? I'm currently on the codex and Claude code CLI, thinking to try cursor so I can code on the mobile app too.

@ericzakariasson It's surprising how, GPT-5 High/Codex performs better on @cursor_ai than on Codex CLI.

@ericzakariasson What makes it better than gpt-5 codex in the codex cli? I'm keen to try cursor, but why?

@Ex3NDR @dhh Awesome App! Happy launches CC, how can I get it connected to my Codex CLI?

Happy Coder is now profitable! we achieved this by following @dhh advice and run it on nuc powered using solar and tesla walls and we pay ~$50 for expo and we are getting about 90$ MRR in donations! So far we got almost 4k stars and almost 300 forks, we are climbing without any hack - just pure product. Probably most successful OSS product of me in last decade!

@RealDannyTommo Finally

NO MORE EXCUSES. NO MORE MARCHES. NO MORE PROTESTS. NO MORE TALK. IT’S TIME, LADS! 🏴󠁧󠁢󠁥󠁮󠁧󠁿✝️ MAKE YOUR CHOICE — THE CALL IS COMING! 🏴󠁧󠁢󠁥󠁮󠁧󠁿✝️ FOR OUR CHILDREN! 🏴󠁧󠁢󠁥󠁮󠁧󠁿✝️ FOR ENGLAND🏴󠁧󠁢󠁥󠁮󠁧󠁿✝️

@HJB_News__ Incomprehensible

England is now the only country in the world where English citizens are banned from flying their national flag, but foreigners that live in England have the right under new laws. This is London their capital city

@ShroudsACheater @thdxr Today is so bad! I did the upgrade today idk if thats part of it.

@thdxr It fucking sucks today idk wtf is going on

now everyone is saying codex sucks and they're going back to claude code 🙃

@kimmonismus I use DR once per week. I sometimes run the same prompt through Gemini and Grok's equivalent and get one of them to compare and fill in gaps. Gpt usually wins if I were to make a judgement call.

I'm curious. Do you use DeepReseach regularly? I use it maybe once every two weeks to delve deeply into a topic, but otherwise hardly ever. Again, not because it's bad, but because there's often no use case for it. How about you? And which tools do you use regularly in your workflows?

@alexfinn Gold. Love these tweaks on how I use codex and CC side by side.

This is the best AI coding setup ever: Codex on the left, Claude Code on the right 1. Start building with Claude Code in plan mode 2. Copy and paste every plan into Codex. Validate the plan. If they have any critical feedback, give it back to CC 3. Any time CC messes up, go back to Codex and have it fix the issue 4. Every change either agent makes, ask the other agent to review it 5. Primarily use whichever agent that is the last one to clean up the other's mess Not only will you produce apps 10x faster, you'll have fun just feeling like you're managing a team of engineers Trust me, the results are worth the price you pay