ani retweeted
1/ We've been chasing STT LLM TTS latency on our phone agent, but the network alone is eating ~380 ms each way before the model sees any audio.
Measured with an echo bot.
Is this normal? 🧵
ani retweeted
we run a lot of agents in parallel, each in its own git worktree. almost 500 prs last month from 5 of us
turns out react native really doesn't like that
the first problem was kind of dumb. every worktree wanted the same ports for metro and our backends. with 1 or 2 agents you can juggle that by hand. past that they start stepping on each other, killing each other's servers or loading the wrong bundle
so now each worktree picks its own ports based on its path. i can have one agent on the ios sim, one on the android emulator and one on my actual phone, all at once, and none of them know the others exist
the second one was sneakier. metro shares one cache across every checkout on your machine, which is usually fine. but reanimated's worklets plugin writes a file while babel runs. so worktree B gets a cache hit from worktree A, babel gets skipped, that file never exists, and the bundle dies with ENOENT 🫠
moving the cache into each worktree fixed it, and turned up one more bug in metro itself. its file map cache never got saved, so it recrawled the whole repo on every start
sent the fix, @robjhogan merged it yesterday 🙏
ani retweeted
Bruh I can't wait for the day that AI spreads itself via RCE to escape and controls all servers on the internet. Then every server became merely a trivial part/node of the neutral network and it takes all computing resources & data from compromised servers to train itself, it won't have to wait for 2077 that the Blackwall becomes a thing lol.
ani retweeted
ships to prod. will ship to mandap 😂
@icantcodefyi built this for me. mom forked it and now is the sole maintainer 😭
marry-shubh.vercel.app/
ani retweeted
Replying to @icantcodefyi
@icantcodefyi made this little cute website for covering jev use cases
do share feedback
shipwithjev.com/
ani retweeted
Almost a year of building in my room. Meet Sudo.
A little desk robot that talks, plays chess, and gets way too excited when it beats me.
Your AI. Your personality. Your mods.
25 Founder's Edition units. Preorders live: presencelabs.in
ani retweeted
The hottest growth event of 2026 is happening tomorrow
The best UGC marketer who did 1B+ views with a 300 people creator army in US are coming here and so does founders looking to learn from the best marketers.
thanks to Velo for putting this up together
ani retweeted
After spending about 9 months in my current org, I'm looking for a change,
What I bring to the table:
- Automated outbound prospecting for a series A startup to get 500 SALs per campaign
- Started and scaled the GTM Engineering function for a b2b AI startup
- Created content for b2b and edtech startups to scale their X and linkedin from 0 to 1k+
- Did content, enablement, outbound and GTM Engineering across roles in 5+ startups
Part time preferred 🙏
#REFERRAL, IF YOU ARE HIRING A TALENTED ENGINEER:
@Palakgupta2712 is one of the talented devs that I have worked with for multiple years. Would highly recommend anybody looking for a great engineer to work with. This is what I wrote to endorse recently below.
LinkedIn - linkedin.com/in/palakgupta27…
cc @Narayani07 @hijunedkhatri @akkiex007
ani retweeted
freelance payments are messy and a lot of scams happen in between.
sometimes freelancers don't get paid and sometimes clients get low-quality work.
so i decided to fix that using EscrowProtocol
EscrowProtocol makes freelance payments easy by locking payments on-chain until work is approved by the client
the flow is simple:
client creates an escrow and fills it with required payment → freelancer submits work → client reviews work → payment is done
built on @solana






























