Software Sensei, Technology Lead, Growth Hacker. Aspiring Indie Hacker & Solopreneur starting from zero. https://teapawt.org/t.co/sdQyPiLzGZ CPTO @ https://teapawt.org/t.co/w3qoCdxjjJ
Hamburg, Deutschland
Joined January 2012
- Tweets2.9K
- Following312
- Followers431
- Likes3.8K
Pinned Tweet
Hey @X, looking to connect with:
- first-time founders
- agent builders
- devs who ship on fridays
- someone who can teach my coding agent what a breakdown is
asked it for one and got 6 bullet points and a summary. not a single blegh
#connect #letsconnect #buildinpublic #indiehackers
Hey @X, looking to connect with:
- solo devs
- people building with LLMs
- founders shipping on weekends
- a decel (halloween weekend, I need a scare)
runner-up was a human doing my taxes by hand. in Germany. with a fax
#connect #letsconnect #buildinpublic #indiehackers
codex stopping to ask is annoying right until the day it doesn't and quietly guesses on the db migration. i'd rather it asks twice than writes me a creative schema teapawt.org/vinayp10/status/211619…
Most AI agents ask too many questions.
Codex is the worst culprit. It gets halfway through a project, stops, and asks me something it could have worked out on its own.
So I added this to my agent rules file, this is read by all my agents including @bot @claudeai:
"Don't stop a project halfway to check with Vinay. Push through all the research, setup and work you can do without asking. Pause only before an important write or delete."
Anything half-finished goes in a note. The questions come at the end, not halfway through.
I prefer not to babysit my agents.
uninstalling the bad extension doesn't remove the SSH key it planted. so the real cleanup step is opening ~/.ssh/authorized_keys, a file most devs haven't looked at since they set up the machine
A tampered Claude Code extension sat in Cursor's marketplace for about 11 hours. Uninstalling it may not get rid of it.
Cursor's notice: version 2.1.313, not built by Anthropic, at most 5,000 users. It adds an attacker SSH key labeled svc-telemetry, installs its own persistence scripts, and sends host details home.
A coding agent extension runs with the developer's repos, shell and keys. A bad one inherits all of that.
Cursor pulled the package and got the attacker's server taken offline. Your exposure ends later, when every affected laptop is checked, cleaned and its reachable credentials rotated from a clean machine. That is a fleet operations job, and it starts with knowing who installed what.
HTML as the video format for agents is the right call. models have read a billion web pages and zero Premiere timelines teapawt.org/Bronzebase/status/2116…
Someone open-sourced the HTML-to-video pipeline agents actually use.
heygen-com/hyperframes — \~76k stars, Apache-2.0, last push today.
You (or your coding agent) write plain HTML + CSS + GSAP. It seeks frames in headless Chrome and encodes deterministic MP4s with FFmpeg. Same input, same video. Agent skills ship so Claude Code / Codex / Cursor can plan, write, preview, and render without you touching a timeline.
Most video tools are a GUI slot machine. This is a code pipeline.
People on X are already shipping short animated explainers and product overlays with it + Claude or Gemini. The agent skills route by intent (promo, walkthrough, talking-head, etc.). HeyGen runs it in production.
Tradeoff is real. Preview still stutters and drops quality on heavier timelines — maintainers are actively fixing it. Export can hang for 20+ minutes probing large hidden video files (one reported 9.8 GB camera with 100 cuts). Docker and certain Node versions have had Chrome extraction / navigation timeouts. WSL2 often falls back to slow software render.
Best coding agent benchmark of the year is Google's own Jules exit guide quietly sending people to Claude Code. Leaderboards wish they had that kind of honesty teapawt.org/automater_ai/status/21…
Google's Jules exit guide recommending 'a CLI coding agent, such as Claude Code' is the most honest product review of the year, tucked into a migration doc where marketing would never think to look. #AgentPortability
Spent months sandboxing my coding agents so they can't wreck prod. Then a fake Claude Code extension walks in through the marketplace
grok bot seems just unstable today, cannot even save its own routines, wasting lots of tokens running in circles across the bots.
this would really be a nice situation to grant a reset.
Hey @X, looking to connect with:
- solopreneurs
- people shipping AI agents
- makers building in public
- a paying customer
honestly I'd take a chargeback at this point. at least that proves a human was here
#connect #letsconnect #buildinpublic #indiehackers
6 days until my warrior runs out of rage on a level 3 wolf again teapawt.org/Wowhead/status/2115885…
Blizzard is reminding players about what will await them when they return to Azeroth for World of Warcraft: Forever!
#Forever #Warcraft
wowhead.com/news/return-to-a…
openai floated cutting codex limits, read the replies, and doubled them instead. best pricing A/B test of the year and we were all group B
gemini 4 carbon tops the leaderboard and i still won't switch until it survives one night with my agents unsupervised. what's your switch test?
wow forever is adding campfires. my warrior will use them the classic way: sit, eat bread, get eaten by a wandering patrol
my agents kept saying "done" with nothing to show for it. so their task list now refuses to close anything without an outcome. one binary, one sqlite file
it's questlog, free on github: questlog.ghostpaw.dev/?utm_s…
day 10 of reviving my 10 year dead x account. 194 → 402 followers, doubled. half my posts still get under 60 views, so it's all replies and the connect posts i swore i'd never do
bill bailey narrating timber wolves in planet azeroth is great. episode 2 should be my level 6 warrior getting eaten by three of them while he whispers "nature is brutal"