Andre Antunes

@femisapien_z Codex > Claude

I started a session with Claude code today to convert matlab code to python. Wrote a whole plan: map dependencies, eliminate unnecessary, check if app runs. Then I asked to convert. Only 36 functions to begin with. It said it was massive work! It didn't think to scan the app's final output to verify what was actually been used (after that only 16 functions remained). After 4 hours of mediocre work it said: "let's continue tomorrow". Can you believe that?

Marc Andreessen on introspection and the benefits of retardmaxxing: "There's this guy on YouTube who has basically a hundred videos on retardmaxxing." "He's like my new life coach. I haven't met him, but from a distance." "It's basically just—retardmaxx. Go to work, do a good job, come home, it's fine. Start a company, it succeeds, it fails, it's fine. Have too much to eat one night at dinner, it's fine. Go to the gym, don't count your reps, it's fine. Ask a girl if she wants to go out with you, if she says no, it's fine." "It's like 100 30-minute videos about retardmaxxing. And you would think that after the first two minutes, he kind of covered it. But no." "And by the way, they're all hysterical. They're all absolutely fantastic. It's literally him on his porch in the middle of nowhere with a cigar, and it's like a half hour." "It's just absolutely spectacular." @pmarca with @HarryStebbings

Okay let’s figure out what’s going on down here.

@christinaqi Databento subscriber here. When will you guys offer VIX futures data?

Bruh... If your data provider doesn't colocate, RUN. This means they're reselling another company's data (👀). It also means when you need support, it'll take WEEKS for them to get back to you. It doesn't matter how latency-sensitive you are - the only acceptable answer is "we get our data directly from the source." A lot of startups & incumbents will gaslight you by talking about latency & resolution. That is a heap of 💩. Make them show you their actual architecture.

@BHolmesDev How do you deal with the multiple Codex PR reviews when working with Opus? That also happens when Codex created the code but much less frequently.

I’ve used Opus 4.6 and GPT 5.4 on a mix of projects since release, and want to break down where I think they uniquely excel. It’s more nuanced than you’d think! Rigor of code - GPT 5.4. It goes the distance validating its work without asking. Opus needs explicit instruction to do this, and even then, it misses more edge cases. Clarity of code - Opus 4.6. Claude is a better communicator, which carries into the code. Variable names are clearer and less mechanical, which improves reviewability. This is very important since code review is the bottleneck for most engineering teams. It also adds the right amount of doc comments. GPT simply never comments or explains its work; it’s like working with an obtuse engineer that wants the solution to speak for itself. Sometimes it does, other times not. Similarly, rigor of plans goes to GPT 5.4, while clarity of plans goes to Opus 4.6. An interesting point though: GPT performs better talking through a strategy without a plan, while Opus needs planning mode to put in any rigor. I find myself forgetting plan mode altogether using GPT 5.4. Quality of research - toss-up. Opus spends longer researching with web search, but GPT spends longer studying the existing codebase. You may think codebase research matters more, but researching how others solve the same problem can be just as important. Maybe more important for greenfield. Quality of conversation - Opus 4.6. It’s just better to talk to, which matters using these things everyday. GPT 5.4 was clearly trained to challenge the user more, which results in a tendency to *always* say you are wrong. I’ve had bizarre interactions where GPT claims something is “not quite right,” the restates exactly what we’ve decided on in the last turn. On a personal level, it’s annoying. On a practical level, it makes iteration on a plan slower. THAT SAID, it takes sufficient pushing for Opus to challenge your thinking in this way. Simply say “I’m impartial” and ask questions to avoid that, as you would a person. Overall winner - Opus to make it work, GPT to make it good. I don’t have a good system of when to switch tools, but on average, I prefer Opus early on and GPT for optimization and discussing architectural decisions. Opus is also better for any design related tasks (but state management in frontend apps is better handled by GPT).

🚨 Simulation Theory: The Double Slit Experiment proves particles act like waves until observed then they snap into particles. What if our reality only "renders" when we're looking, just like a video game optimizing resources? Check out this episode from The Why Files breaking it down, tying it to Simulation Theory. Are we in a sim? This could be the key to unlocking the true nature of existence! The Why Files video did a great job on explaining the Double Slit Experiment & Simulation Theory What do YOU think—real or rendered? Drop your thoughts below!

@emollick Agree 100%, 5.4 Pro is a different beast

GPT-5.4 Pro continues to be the only model of its class. For anything really hard & complex, I throw it into the maw with every bit of context I can think of. More often than not, something very useful comes out. I can't get the same results from Codex or Code or anything else.

@ajambrosino Cursive is important IMO. Can't understand why it was removed in the first place

phone ban: should’ve happened sooner cursive: wtf…? it’s time to defund big cursive

I often re-read @natfriedman's personal website (former Github CEO)

@doodlestein @RG_Leachman @TendiesOfWisdom Regarding Codex Pro accounts, do you think it's better to have multiple Pro accounts or to add credits when the weekly limits run out?

@RG_Leachman @TendiesOfWisdom Unfortunately for my kids, daddy has 22 Max accounts 😭😔👶

“Kids I have good news. Daddy is out of Claude tokens until 3PM. He has time to play with you now.”

@MarceloLima @bcherny @_catwu @mikeyk @brada @scahhht @felixrieseberg @gabemulley @claudeai @AnthropicAI Use the CLI instead, works much better imo

@bcherny @_catwu @mikeyk @brada @scahhht @felixrieseberg @gabemulley @claudeai @AnthropicAI Team, this keeps happening and makes it very hard to work

@TheShortBear @cursor_ai Thoughts on it vs Opus? I mostly use codex nowadays, more reliable in my experience

Used Composer 2 all day. Very big fan of the work guys @cursor_ai ! Coming from a heavy Ultra user that used it all.

@theo You can add codex with mcp to Claude, but the scope is limited it seems, works more as a second brain, to revise plans, etc

I bet Claude Code would be much more performant if there were allowed to use Codex to fix it

@levelsio Usually those places weren’t built with AC infra. When AC became mainstream, whole-house installs were hard, so ppl just did bedrooms.

🇧🇷 There has to be a reason why majority of houses in Brazil only have AC in the bedrooms but not in the living room And the living rooms get insanely hot in the summer (which is December-March, opposite of US/EU etc) Literally majority of the Airbnbs we stayed didn't have AC in living room Good thing is they at least have AC in the bedrooms though unlike Europe which doesn't have any AC usually But still not Asia where the entire place is fully 16-18'C ACd in every single room, how it should be!

@paulg His a great author, More Money Than God is a great book

This sounds like a promising book. Mallaby is a sharp observer, and sharpness + access tends to yield insights.

@AskPerplexity Hey @grok compare web app development from computer to Replit agent 4

We just shipped several upgrades to building web apps with Perplexity Computer. New projects are now automatically tested using Playwright. It navigates your app like a real user to identify and fix bugs before you ever see them. Build, test, fix, ship. All in one place.

The greatest moment in F1 history.
 The day Max Verstappen ended 7 years of Mercedes dominance and saved the sport.

@mikkom Agree, but just a comment that in trading the continuous version is more appropriate

Please don't use kelly for trading. It is not suitable. I see so many people posting about kelly this and kelly that related to trading. Kelly was originally designed for gambling. Gambling has fixed rules that typically favour the house but in some cases you can find an edge. Because the rules and the environment are fixed, the edge can be calculated. In trading you have volatility clustering and fat tails and everchanging environmental chaos from both external event, varying participants and market internals. Kelly assumes fixed and known constants. In trading NONE of these is true: - Environment is known and constant - Ratio of wins/losses is constant and known - Payout is constant and known You can get much better results via backtesting with decent sample size + monte carlo uncertainties. /rant

@ah20im Organize the threads into folders within a project. Being able to actually see and open the folder and files, a bit like cursor, it helps.

What would you like to see in Codex?

@drgurner What’s your perspective on the whole debate about introspection and builders? I am in the minority side and actually agree with @pmarca

I would love to see more interviews where people actually get pushback, and they're not just a platform for some person bloviating.