Bit

if any of u guys want in on early testing lmk !!

It does bother me that it costs money to code now

@umgbhalla That actually explains it

anthropic has a co maintainer of electronjs on team and their macos app is still so ass

I have this feeling that a Chinese model release is near. The vibes are saying smthing shall happen soon.

@dothebit ???

@arpitingle man tf how can we be in so much sync

listening to the paul atreides war chant on loop

@Ibelick man this is sooo awssomee

magnetic morph exploration

if ur not watching the new dune trailer every 2 hours ur ngmi actually.

@ItsAditya_xyz @TheArena how is it self sustaining then. it does not make sense probablility wise ik but still it would have an average with scale. running just one means that its one trade away from game over.

@dothebit @TheArena Just bcz u can turn 10$ into 15$ trade once doesn't mean you can do it 2000 times.

Also my agent is just running for free. No cloud bills, no AI bills either (just using chatgpt codex) It has a goal in a month to make 39$/month as it's running on @TheArena launcher's. It makes very safe and sound trades to make those profits and pays for the compute

@ItsAditya_xyz @TheArena uhh whats stopping u from scalling this to 100 - 2000 agents?

@dothebit @TheArena Yup, so far it made 150$ this month.

i think we cannot fully automate most processes that people do nowadays. but i think bringing the current coding/egineering mindset to all the other industrys could potentially be a very big win iykwim. i mean just look a industry u would wanna automate right? the architecture for that automation also has to be flexible that it can be changed instantly in the future by just a click of buttons uk? for example system prompts. you make a v1 -> then build a architecture that self evolves the system prompt by the data u recieve. but u also need a way to interrupt the process which means it cannot be ever truly autonomous. i kind of imagine this whole thing as => v1 => iterate v2> => automate => looks at data= > updates => simulates the process => passes some checks made by us => goes live => does another check => can be rolled back anytime => if something goes wrong human is able to stop/interviene at any step and suggest something else. i think this will be the future of most industries.

@ni5arga ayo nic

ppl building codex app are doing sm generational soft engineering i think

@willdepue hell yea

shoutout bihar

@yiliush x.com/dothebit/statu… btw

testing something

@avinierx codex tends to do stuff i dont understand and actually very dumb idont know. it will make u feel that it did smthing very cool. but i would'nt be so sure

whenever i paste codex output in claude code for the same input, it always says 'oh this analysis much sharper and precise than mine' like bro😭😭

@DevvMandal @markov__ai soo cool

Today, we're launching the world's largest open-source dataset of computer-use recordings. 10,000+ hours across Salesforce, Blender, Photoshop and more, to automate the next level of white-collar work. Link in the comments :) @markov__ai

if ur life has gotten any easier with claude code. ur were just not working anyways. because i spend double time than before arguing with this mathematical algorithm which gives me more stress and anxiety but who cares lmao. life is all rainbows and sunshine. i love opus.

@hoenogatari its tru he called me

chat i scheduled a random call with a random person on the internet and we ended up talking for the whole hour maybe i can hold a conversation afterall