Hari Singh

I used Devin and OpenHands before Codex came out. The accuracy is on another level. OpenAI has tuned the model behind Codex really well. In my opinion, Devin or any other company won’t come close. Even with minimal context, Codex delivers the best result for price. I’ve already closed over 50 PRs with it.

@karrisaarinen @linear Linear is awesome please build confluence replacement so we can move our docs as well

At this point, it feels like most teams building great products are on @linear (even if not mentioned here, they're probably on or currently trialing): Polymarket Perplexity Supabase Cash App Coinbase Substack Mercury Raycast Lovable OpenAI Cursor Vercel Replit Ramp Boom Brex Clay

With codex-gpt 5.3 you can just build opinionated IDE for your self using electron!

@dakshgup working with this big boy. Currently its 4/5 but I will get there!

what’s the most number of revisions you’ve made on a PR to get a 5/5?

@OfficialLoganK Why not just create tabs in the page vs the option in the sidebar?

We just made paying for the Gemini API 10x easier : ) You can now upgrade to a paid Gemini API account without leaving AI Studio, track your usage, filter spend by model, and much more to come!

@paulg 100%! and people that nitpick on UX will win.

Prediction: In the AI age, taste will become even more important. When anyone can make anything, the big differentiator is what you choose to make. paulgraham.com/taste.html

@jamesacowling Reviewing PR > writing tests

Your job is now writing tests

@ryancarson coder.com? devpods?

nvm. this is waaaaaay too slow and the vm doesn't have enough ram. back to local dev and managing multiple checkouts or git worktrees. also, i want to really give codex a shot because 5.3-codex + 5.3-codex-spark seems

@samlambert Sam I know few ppl who would 100% migrate from azure flexi server. It would be good to write up a migration guide. When I migrated my database it was bit pain full I had to use pg dump.

We are triaging migrations like crazy right now. Please include as much about your workload as you can when you email us.

@svgoiboi @supabase 😂😂😂I can't imagine our db going down for even 1 minute! Just use @PlanetScale !

can you pls fix your outage @supabase i have 4k users reliant off of you guys as a db and it's been down for 1hr+

@pbteja1998 When it comes to managing large amount of servers aws shines. I’d suggest looking into ecs fargate or ec2 instances. What ever you go with make sure patching the servers is easy.

Turning MissionControlHQ.ai into a SaaS is... ....95% just figuring out what infra to use so that it is reliable, and won't just stop working. Figuring out how to seamlessly push updates to user's servers without anything going down... even after the user gets onboarded. Figuring out how to manage versions and states of plugins and extensions across 100+ servers at once... It's only 5% actual code, rest of the 95% is just getting the infrastructure right.... and it's much harder than I initially anticipated. But slowly getting there... I think we almost figured out the right balance. I don't like making so many people wait for it, so trying to get it out as fast as I can... hoping that everyone will understand.

@mitsuhiko C# is the best language for llms.

This weekend I was thinking about programming languages. Programming languages for agents. Will we see them? I believe people will (and should!) try to build some. lucumr.pocoo.org/2026/2/9/a-lan…

@MicrosoftHelps @msPartner I’m trying to sign up for the Microsoft Partner Program, but my request keeps getting blocked. Then I go to the help page, and it requires a workspace to contact support-but I don’t have a workspace because my request is blocked. Who am I supposed to contact? This is a dead end no product should lead customer to an DEAD END!

@pbteja1998 @PaddleHQ Careful you don’t want to be held liable security issues if you managing the vps.

Mission Control as a SaaS is almost ready!! It was more complicated than I thought to make the experience right… Got approved by @PaddleHQ But I still need to do a few more things before I make it live. Want to give early access to a few people to make it even better, before I launch it to everyone. This is going to be BYOK model. You bring your own Claud key. Every thing else will be provided by me. If you are willing to pay $99/mo to get early access, let me know…

@sama 100$ plan plz

How would you prefer us to charge for Codex?

@sama @embirico can we get a 100$ plan for codex cli plz.

@aidenybai I bet you will like codex cli with 5.3 codex -high model. Its very smooth.

alright, fuck it time to try out and rank every single terminal coding agent - claude code - codex - cursor cli - opencode - droid - amp - gemini anybody im missing?

Ok, I used Codex CLI when it first came out, then switched to Claude Code. But coming back to Codex damn, it’s on another level. @embirico and the team have cooked up something awesome. GPT-5.3 Codex is 🔥. Claude code feels clunky! Rust for TUI is the future.

@pbteja1998 I think you are getting distracted. I’d stick to growing sitegpt if I was you. You really do have an amazing product with sitegpt.

Mission Control is spreading like wild fire now!! I have seen so many people talking about it - creating their own versions of it... This is so awesome to see!! Making it even easier for everyone to create Mission Control. It will just be 1 click and you will have my Mission Control system and your own Jarvis with you. No need to buy Mac-Minis or setup and manage infrastructure. You will have an entire Squad of AI agents (led by your own Jarvis) working towards a common Mission that you set. I am also going to use Mission Control to grow Mission Control.