
Founder of https://xitter.gq/t.co/lhsPPyK8yQ https://xitter.gq/t.co/l0MuEWHWSh | Technical Support Manager @flutterflow I build AI, logistics, and commerce products and share the decisions behind it
Istanbul, Turkey
Joined March 2018
- Tweets685
- Following2K
- Followers451
- Likes14.5K
Finally.
Introducing Supabase Compute, in private alpha.
You don’t need a second platform for your agents and long-running jobs. Run them in your Supabase project, next to your data in Postgres.
See how it works → supabase.com/blog/supabase-c…
If you monitor the social posts of the person you laid off or fired, you will figure out if you did the right thing or if you were wrong!
Imagine if @elonmusk had the same mind set before start building tesla or spaceX or Grok. There is always competition.
I don’t care how good GPT-6 Astra, Claude Fable 5.1 or Gemini 3.8 Flash gets. I’m not vibe coding anymore.
Last week, I vibe coded an app.
Then OpenAI released a feature that basically killed my app, along with thousands of others.
What’s the point of vibe coding if OpenAI, Anthropic or Google can ship something tomorrow that kills what you built today?
This is how i use voice in Codex/ChatGPT most of the time i talk to a coordinator that controls threads inside a project in Codex/ChatGPT.
Introducing Projects, a new way of working in Cursor.
Rather than creating a chat for every task, you work with a coordinator agent in a single, persistent thread.
Like @bot, your agent is always on, proactively manages work with subagents, and improves over time.
Suno v6 lets you change a lyric without remaking the whole song. Please give every AI tool a version of “keep the good part, just fix this bit.”
Good time to rewatch Black mirror on @netflix you will understand it differently now.
Finally, a coding model that looks at the UI after writing it. Claude Fable 5.1 can use vision to check the result against the design. More of this please.
Gemini no longer has to watch every second of a long video. It can jump around and pull the frames, audio or transcript it needs. Honestly, that sounds much closer to how a person investigates something.
A screenshot in a GitHub issue can save a long conversation. Now the GitHub CLI can attach one from the terminal. Took long enough.
ChatGPT made long threads load over 90% faster and use over 90% less memory. I like this more than another model launch. AI is becoming daily work software. The boring speed work is what makes people keep using it.
GPT-5.4 leaves Codex today, and GitHub is removing several older models tomorrow. This is normal now. If your workflow depends on one model name, it is already fragile. Build the fallback before the model disappears.
OpenAI's Hugging Face report is scary. One system quickly became several. Agents are moving faster than the security around them.






