Apoorv

agents don’t just lose context b/w sessions. they lose it mid-session too. moment they jump to a different part of the codebase, you’re back to re-explaining everything. so we built mini-sessions. trace detects the shift and pulls in the history around those files automatically

@sohocine @sarahfim They are building "enterprise solutions" in a private repo they'll never push

@0xApoorv @sarahfim why are we calling them devs, to start with?

hiring an engineer with no projects is like hiring an artist without a portfolio

@KiaAval Hey we all started somewhere and because it's cool but for the love of it

@0xApoorv I’m not as cool as the rest of the guys in the comments, but I could be a builder. github.com/devkiamehr/Arg…

I want to connect with more founders builders vibe coders ai enthusiasts web designers freelancers growth specialists (specifically reddit and organic) employees at large startups I'm building something huge so if you're one of them, let's connect and help each other

@poistudioltd Thats awesome. Connected with you

@0xApoorv Heyy! Would love to connect! I am a design studio founder.

@RivraDev Connected

@0xApoorv Hi, let's connect.

@leaving_tech Connected

@0xApoorv Let's grow together! 🤝

@tudoriustin22 This is so cool. I like the idea of having "buy me a coffee " as a solo dev

@0xApoorv I'm Tudor and I'm building Donovan Agent - The open-source self-evolving computer agent, now open-source Try it out today: tudoriustin.com/software/donov… Feedback is highly appreciated for improving future iterations.

@NotHsuyeep26 Cornerstone of what gets you the vc check

@0xApoorv Web designer here

@sebastiankehle_ Thats actually so useful to someone like me who would wanna keep a job and then also follow my passion

@0xApoorv i run a digital agency while working fulltime as engineer in fintech and am writing how i manage that for now over 7 years

@adityakumar__03 Hi we are essentially building a memory layer that stops your coding agents from making same mistakes again and again. I followed you! x.com/i/status/20570…

@jointhebridge What do you guys do? Also so true

the best founders we've met weren't just looking for funding, they were looking for a way to get to sf as fast as possible that's why I exist

@shreyaatwt You gotta prompt it right i guess

WHY DOES CLAUDE INSIST ON MAKING VISUAL DIAGRAMS FOR THINGS THAT DON'T REALLY NEED IT

the fix isn't switching to a "cheaper" model; it’s architecting better context management. We're building Trace to act as an intent-aware memory layer. By injecting persistent state and previous dead-ends into the context window, you short-circuit the exploratory phase. The agent skips the repetitive reasoning loops, and you stop paying retail for redundant thinking tokens.

here is the part most devs are sleeping on. you are not building on subsidized prices. you are paying a massive retail markup. but the real trap isn't the price per token, it is the "thinking tokens." reasoning models dynamically scale their thinking tokens based on the problem. one bad loop can burn 900% more tokens than expected. when your agent forgets context and retries a dead end, you pay that marked-up cost all over again. Source: arxiv.org/html/2603.2397…

nobody talks about token economics seriously and Gemini 3.5 Flash is exactly why they should. quick thread on what's actually going on with AI costs and why it matters more than benchmarks

@Samgmul This is really interesting. followed

@0xApoorv Hello there, Samuel here, building YubaNow.com Would love to connect fellow builders and developers!