khushal @celeroncoder

this prompt has only acturally ran for like 20 mins max, it says almost 4h because i accidently lost wifi on my box & then i reconnected it and it started like nothing happened coding agents has come a long way

me when I was using caffeinate commands to keep my mac up

dudu @dudufolio

a day ago

Just got a new mac what should I install?

233 9 771 239K 136

@theo works for projects that aren't a git repo? I have a project my agent setup and it isn't a repo bad setup ik

@aicodeking @t3dotcodes @jullerino new benchmark to chase?

it's nice that the free 10$ credit you get every month on gcp on the pro plan gets banked (doesn't expire after the month) argon is pretty unusable in terms of costs there tho, the cache is 89% on antigravity, gets better with pi but still costs a lot (also these are without the introductory discount they mention in the blog) and this is just using gemini-4-argon ocassionally for reviews with claude (hallucinates a lot but some good novels findings) x.com/celeroncoder/s…

khushal @celeroncoder

5 days ago

im using gemini 4 argon from vertex ai/ agent platform whatever from gcp and they don't have the announced introductory price there and that cache hit is pretty burtal only 88% ? and this is like a single run and use it on antigravity CLI maybe if i use it with pi or

2 0 4 500 0

i should stop the skill mongering :/

with claude these tools are better

Lasse @lassejv

2 days ago

mole.fit is still the best storage cleaner for mac

1 0 8 426 3

i wasn't a fan of the inbox style sidebar on t3 code but i came to terms with it and with the new working section it's replacable for me thanks @theo

i haven't seen a better model launch video

MiniMax (official) @MiniMax_AI

a day ago

Welcome back. Space Bunny. 🚀 MiniMax-M3.1-Flash is coming.

285 566 7K 861K 719

@BHolmesDev @dexhorthy my pr workflow has this skill + pr-write skill as the default in most of my worktrees, such a good side-chat skill and great when revisiting PRs

My Grok bot and I both work in the same T3code thread I absolutely love this collaboration style

i like it in general the only opensource i've tried is openmausbot and that wan't a good experience then maybe it's improved now good features i'd love: 1. bots being able to talk to bots on other servers would be gamechanging 2. some UI bugs overall (like if i paste in something in a code block and it expands the whole message UI was expanded. 3. my bot generate an html report but html code can't be previewed 4. also i think some template or files it saving on my machine even when im talking to bots on a different server not sure about that 5. i want to be able to change the animation of a running bot 6. group chats (in docs and when i asked it says it has it but like i don't see it i can put them under a section is that the group thing? because in grok bot i've seen like proper group chats with like indications of all bots like a channel of sorts 7. direct claude usage and inject your bot tools like t3code does so i can directly use MCP on my claude code or other harness 8. reuse the machines harness binaries if possible, having ur binaries made sense but like installing and authenticating a new one was tedious if it could just reuse those that would've been great. 9. for gemini/antigravity ig you're gemini CLI not sure 10. also i set my default agent on claude and then i switched the model to anything else like gemini or opencode it errored with invalid params not sure what's that error. 11. adding a server option only shows your cloud thing the custom server addition should also be there in the app and not just in docs imo 12. being able to add MCP servers or other integrations directly in the chat and not go to the settings page that was annoying 13. i want to be able to use a grok bot template and add it directly im not sure if that's possible right now because the only import option i could figure out was to import grok bot config 14. for headless boxes as server i think browser is possible? i think you should separate browser use from computer use (also be able to import cookies there would be nice 15. for voice i want to be able to have more options than local models something like what voiceInk supports so it's there on mobile too 16. with mobile the connection is slow i think you should consider having some local cache as well and also the ability to select like a default server 17. on mobile there was this cursor problem it overlapping like being far too right like being on top of the placeholder text almost in the composer, and i don't like that small placeholder thing you do in mobile (also when i have the composer on and i scroll it goes away (android) this is all i remember rn thanks!

i installed @OpenBot_ today now that they have support for standalone server without having the desktop app installed there. it has some rough edges and is nowhere near grok bot in terms of feature parity or ux but it's still a great open-source option and is very easy to setup. im triaging my github issues using it, it's finding novel issues in documentation, product issues with the QA bot and all of those issues are getting triaged by using the t3code MCP of my multiple machines (with priority based on available resources). let's see if this workflow sticks to me (i still like going to t3code threads and following up on longer feature / bug threads but the initial messages and my workflow which is was thinking to offload to an evolving skill is now handled by the bot which is self evolving so that works out ig :/)

my oracle box (free forever btw) that i use for routines can't handle two pnpm installs together 😭

@localhost_5173 can this monitor remote machines? and maybe like expose this data to an MCP so my bot can delegate work to the machine which is more idle?

@alexdotjs wtf

btw thanks @tryreplicas for the 100 bucks of free credit

this feels legit, waiting for it to be somewhat stable those pesky typechecks getting shorter is a big W x.com/ctatedev/statu…

Chris Tate @ctatedev

4 days ago

New from Vercel Labs: tsc-ts TypeScript was written in TypeScript Then rewritten in Go We ported it back to TypeScript Compiled it with scriptc And it's 2-4x faster than tsgo

246 199 3K 486K 924

@sundarpichai @GoogleQuantumAI researchers using cats now as clickbait 😭

I'm trusting bun less and less with the slop generated stats he put out the number would be legit but that ai image/screenshot I can't detach from slop in my mind

Jarred Sumner @jarredsumner

4 days ago

Here's a little how Bun AOT currently works (and may still change a lot): 1. bun build --compile --aot <script> 2. `bun check` type-checks but allows untyped code 3. Typed IR 4. Optimization passes 5. JavaScriptCore's B3 compiler runs 6. ARM64/x64 machine code (continued)

33 8 431 28K 68