Daniel Aranda

React is a state-based framework, but a lot of people use it like jQuery with JSX. Then AI learned from that code. Now Codex, Claude, etc. happily generate useEffect abuse, duplicated state, ref hacks, and DOM patching. React is not the problem. The mental model is.

@Xxi5olc @awilkinson Might be worth checking the setup/workflow. I’m using 5.5 in Codex and it’s been noticeably faster for me than Claude Code on comparable tasks. What kind of task/repo is taking 25 minutes?

@awilkinson I use 5.5 in Codex. Will say Claude is the slow one lol

I want to love Codex, but holy mother of god did they cover the GPUs in molasses? I swear to god it takes 25 min to do something that Claude Code could do in 5 minutes (and yes, I'm using fast mode).

Told a dev to switch from Cursor to Codex with GPT-5.5. He showed me his screen: "I'm already on GPT-5.5 in Cursor. Same thing, right?" Bro. That's not the same thing. The model is the engine. The harness is the entire car. Tool routing, context strategy, agent loops, system prompts — Cursor and Codex differ on every single one. Same GPT-5.5 in different harnesses produces wildly different outcomes. Different cost. Different reliability. Different ceiling. The model is maybe 30% of the actual experience. The harness is the other 70%. How do you explain this to people without sounding like a snob? 😂

@thsottiaux I hope someone rethinks the limits. In this AI era, a weekly reset feels ancient. A 5-hour to 3-day reset window would be far more reasonable.

@matt_gray_ This says less about AI’s leverage and more about how you view people. I’m an AI advocate and use it daily; it helped me go from architecting and coding everything myself to architecting systems and delegating better. But treating human needs as defects is the sick part.

just reviewed our numbers. we spend $2,400/month on AI tools. they replace roughly $47K/month in human labor costs. that's a 19x return. and the AI doesn't call in sick, doesn't need a 1-on-1, and doesn't have a 'creative block.' still think AI is overhyped?

@cryptogems555 @grok @BitcoinPulseX

@Luke_Unedited4U @cyrilXBT OpenAI began as research-first, not profit-first. Products changed that, same as the rest of AI. @elonmusk is just being combative here.

@cyrilXBT Yes, but they have text messages between Sam Altman and Elon Musk, talking specifically about how they both wanted to create an arm in Open AI that was for-profit, and arguing over who would be CEO, so this Elon Musk bragging about how he is non profit and for the ppl is just bs

SAM ALTMAN CALLED OPENAI A "COMPANY" TO ELON MUSK'S FACE IN 2015. Elon stopped him cold. "OpenAI is structured as a 501(c)3 nonprofit." Not a company. A nonprofit. Built specifically so no one would ever profit from it. Sam nodded. Then spent the next 10 years running it. Then converted it into a for-profit company valued at $300 billion. With shareholders. With investors expecting returns. The correction Elon gave him in that interview was not a small vocabulary note. It was the entire point of OpenAI's existence. The nonprofit structure was not a legal technicality. It was the promise. The thing that made it different from every other AI lab racing to make money. Sam knew that in 2015. He was corrected on it by the co-founder in real time. And then he changed the structure anyway. The interview exists. The correction is on tape. The conversion happened. @elonmusk is in court arguing this exact point right now. The evidence is the 2015 interview. Screenshot this. Follow @cyrilXBT for every OpenAI development the mainstream refuses to connect the dots on.

@everyonebpup Why is everyone suddenly writing in greentext / 4chan format here? Been seeing it everywhere lately. Write normal please.

> be Sam Altman > email Elon "you're my hero" in February 2023 > get sued by Elon 18 months later > get called Scam Altman to 200M followers > get $97.4B hostile takeover bid > reply at 4:11 PM "we'll buy twitter for $9.74B" > 34M views in a week > court date arrives > show up. dark suit. front row. 10 feet from the bench > your accuser stays home > posts four tweets instead > one of them spells your co-founder's name "Stockman" $134 billion on the line and the man who filed the suit couldn't be bothered to walk through the door.

@DragonStacker @bridgemindai I agree. And even the frontend/design part is becoming less of a gap. Create a strong mockup with GPT Image 2, then have it implement it. The results can be genuinely professional-grade.

@bridgemindai This is the second benchmark I've seen similar to this today. It really doesn't mesh with my experience at all. Other than frontend design, GPT 5.5 is far above anything else in my experience. I burn billions of tokens every month reviewing and shipping code to production.

GPT 5.5 just debuted on BridgeBench. It ranks below GPT 5.4. Read that again. The "most intelligent model ever built" scores worse than its predecessor on real world vibe coding. #8 overall. 84.5 quality. GPT 5.4 sits at #6 with 85.1. Behind Claude Opus 4.6. Behind Claude Opus 4.7. Behind Claude Sonnet 4.6. Behind Grok 4.20. Behind Qwen 3.6 Plus. The smartest model in the world is not the best coding model in the world. BridgeBench just proved it. bridgebench.ai

@jasperdevs @TechnikBlogEu @sama Post your prompt and outputs from both, many of us think you are the problem but if show real examples that will probe your point.

@TechnikBlogEu @sama same prompt with codex vs claude is a massive difference dude, even sam altman said codex is bad at frontend lol

GPT 5.5 is amazing but im still sad it sucks at frontend so I cant cancel claude yet @sama when will this be fixed 🙏

Anthropic acting like they’re playing 4D chess while repeatedly shooting themselves in the foot. Claude had the love of developers. They had momentum. They had trust. Then came worse product decisions, unnecessary limits, and self-inflicted damage.

Hey @thsottiaux Would love your thoughts on this, shared some UX feedback for the Codex experience for power users. I think there’s a big opportunity in reducing cognitive load and making context/state much more visible. linkedin.com/posts/danielar…

The hiring process in tech is completely broken. - Round 1: Recruiter screens you for 15 minutes. - Round 2: Technical phone screen. - Round 3: Take-home project. "Should only take 4-6 hours." Takes 20. - Round 4: On-site. 5 back-to-back interviews. - Round 5: "Culture fit" chat with the founder. - Round 6: Reference checks. - Round 7: Waiting. Silence. Ghosted. Meanwhile the guy who got hired knew someone on the team. 6 rounds of interviews defeated by one LinkedIn connection.

@fahmitech @vivoplt The ideal process? Let the people who feel the outcome decide. Too often, engineers are evaluated only by other engineers protecting their comfort zone. Some of the best builders get filtered out not because they lack skill, but because they think faster.

@CaptainInsightX @0xPrajwal_ Did not you say they delivered the SAME thing?

@0xPrajwal_ Even if AI code is messier?

if two developers ship the same thing one with AI and one without who would you hire?

Codex can learn from experience and proactively suggest things it can do for you. It now has an in-app browser, many new plugins, and so much more.

Stop reading logs. Start reading signals. ⚡ A powerful dashboard to understand your AI spend, session burn rate, and budget risk in real time. Built on top of @agentbudgetsdk github.com/daniel-aranda/… @twtofsahil

🚀 AI is no longer just a coding assistant — it’s a force multiplier. From coding → orchestrating. From 1 horsepower → 20 horsepower. The engineer’s role is evolving toward architecture, product judgment, and execution orchestration.

Built a real-time observability dashboard for @agentbudgetsdk 👀 Hard budget limits are only half the story. Teams also need to see: • where spend moved • what caused the spike • whether burn is accelerating • when exhaustion becomes likely

@nbrempel Would love to hear more about what you built with Opus that you haven’t been able to achieve with Codex. For context, I’ve built many frontend UIs with Codex and the results have been genuinely great, especially when the prompting is design-specific and iterative.

Has anyone successfully made gpt 5.4 /codex good at front end design on web? Opus 4.6 creates some truly amazing front end designs but gpt is absolutely abysmal. I'm giving gpt a shot this month instead of Claude but it's unusable for this purpose so far