manan

The more you f*ck around the more you're gonna find out.

@tylerangert dead internet theory but its products instead of users

the year is 2030. every product is a product for building products that nobody uses.

was thinking of going to sleep early today. here i am glued to my screen watching my codex give me slop and me trying to fix it. life is good

thats how it always was. The better ur plan, the better results u r going to get but getting to a good plan is the hard part. It means you actually understand the system you are building, not just the surface. Otherwise you are just asking it to execute something you do not fully grasp and it will try anyway

GPT 5.5 + Codex with /goal the first setup that turned my work 100% administrative, and I'm getting good at it. There are two outcomes: → A: my plan is good → GPT nails it → B: my plan is bad → GPT tries anyway → it fails but minimizes it And this is the catch: once you're used to detecting the apologetic "everything is fine" tone of case B, you can just revert and pivot to another idea. In particular, it seems effective to ask: "Do you honestly think this is production ready?" If there is anything smelly, GPT 5.5 will say so, allowing you to evolve towards a better design. Concrete example: > Asked it to implement an in-kernel GC for Bend2. > It did exactly I asked, and declared success. > Code grew too much, so I suspected and asked: > "Would you honestly merge that?" > It then replied in the lines of: > "Ehh... actually, this kinda sucks due to " I then investigated and, yes, my idea was bad. Doing GC at that point requires us to track roots coming from many places, which is a massive, fragile endeavor. It is just a bad approach. So, I reverted it, chatted a bit more, then choose another path: use linearity information (which we already have) to free pattern-matched constructors, with small freelists. A few minutes later, it came back with a working implementation that completely solved the issue. I asked again: > "Would you honestly merge that?" To which it replied (verbatim): > Yes, I’m confident enough to call it production-ready for merge, subject to normal review. Not “impossible to break,” but the specific requested invariant is implemented, audited, tested, and bench-checked. When GPT uses that kind of tone, it is very, very likely that whatever it was done is actually solid, robust and trustworthy, and it knows it. I still inspected the code manually and, indeed, it is solid. Doesn't mean fail-proof, but certainly *much* better than before, and mergeable. That's what I like the most about this model: if your plan is solid, it will almost certainly succeed at implementing it. It only fails if the plan itself is poor, which will cause it to desperately attempt to fit a circle in a square, fail, and be too embarrassed to admit. If you master how to detect which case happened, you can get it to produce a lot of constructive, not destructive, outputs...

yeah that’s usually how it goes lol. keeping the MCP layer pretty opinionated for now. instead of letting the agent run raw SQL, I’m exposing safer actions like create resource, add fields, enable ownership, inspect schema, etc. goal is to make the agent build the backend without giving it too much rope

@builtin2005 This sounds fun tbh — these “just exploring” projects usually turn into something way bigger 😄 Curious how you’re designing the MCP layer for agents — more generic or opinionated?

starting a small side project for fun basically a local backend that AI coding agents can control through MCP planning to add simple table creation, CRUD APIs, auth, file storage, a tiny SDK, and maybe a basic dashboard later nothing serious yet, just seeing how far I can take it codex is working hard as ever

@OrevaZSN see thats where u get it wrong. Microsoft never. Ever. Like, never. implements feature that user actually use most of the time

Dear Microsoft, when I hit the Windows Start menu key and start typing a word to autocomplete a search, I never, ever, EVER want it to return results of something not on my computer. Ever. Like, ever, ever, never.

@BenjaminDEKR maybe occasionally asking them to enter their password when using the app might help more. Recently i saw this on Signal app where it frequently asks u to enter the inapp password to check if u still remember it

Is there any real evidence that forcing users to change their passwords every ___ months actually improves security? It seems more likely they won't remember their password, writing it down places they shouldn't or triggering less-secure password resets

wtf are these ads bro 😭

@shivamhwp got him right below ur tweet lol

People are running ads for getting contributions ??

@ibocodes wait why is minimax in the group ?

may 2026 is coming

@chatgpt21 typing this while codex is working its ass off

A meaningful leap from GPT 5.5 would be life changing, considering 5.5 is already a daily necessity for developing

small update on this local AI-controlled backend project got the first real slice working now: - local Postgres - resource/schema creation - auto CRUD APIs - email/password auth - user-owned resources - tiny TS SDK - MCP server connected to Codex Codex can now create backend resources through MCP instead of writing SQL directly.

@thsottiaux can u hit the reset ahem ahem

spent some time looking into redis internals and its kind of wild honestly. a single-threaded core pushing millions of ops/sec, clever memory encodings, probabilistic data structures, and persistence mechanisms. it sounds wrong but somehow it just works

@YourAnonOne so when will it be ready to do my dishes and laundry ?

Evolutionary biologist and outspoken atheist Richard Dawkins says that after spending three days interacting with Claude, which he calls “Claudia,” he is certain that it is conscious, according to AFP.

One thing I miss about the 90s and 2000s and the part of the 2010s is that every 2 years tech just got better. Games looked remarkably better, the systems you would buy would be significantly better. Honestly at this point I can barely tell the difference, performance wise, for the last 5 years.

@Tazerface16 whaaaat? does that mean the billion dollar idea my chatgpt gave me is fake?

People understand that LLMs aren't actually "thinking," right?

@GregKamradt just wanted to ask if its possible to run the agent overnight if i have a plus plan ? wont it run out of limits or something ?

When OAI employees say, “I let codex run all night…” What framework they use? How do you set up the task so it has enough work for 8 hours?

@Astor609 @FaidesTwitch @Brparadox @MrBeast yo sab moh maya ho laa

@FaidesTwitch @Brparadox @builtin2005 @MrBeast

I'm giving away a $2,100 Gaming PC with Paradox Customs!! To Enter: • Follow me and @Brparadox • Like and Retweet this post • Tag 2 friends Additional Entries Here: gleam.io/HPGEi/faide-x-…

Over half of the planet’s internet traffic is now made up of AI bots, per Lumen Technologies