Adrian Rogowski

297 posts

Adrian Rogowski

@AdrianMRogowski

Product @compstak, ex @thoughtbot all views my own

Brooklyn, NY Katılım Aralık 2013

449 Takip Edilen122 Takipçiler

Adrian Rogowski@AdrianMRogowski·23 Mar

@joseph_h_garvin I have a shell script that spins up Claude Code CLI and has to take down tasks off of a roadmap.md file and update the task as done. After it completes the script confirms (runs test suite and build) and fires up another Claude Code CLI agent

English

117

Joseph Garvin@joseph_h_garvin·22 Mar

Claude code rarely runs for longer than 15m without stopping and asking for input from me. How do all these stories of people letting agents run overnight work? Custom harnesses? Yelling at Claude in all caps to keep going no matter what?

English

401

5.8K

1.4M

Adrian Rogowski@AdrianMRogowski·8 Mar

@vipul_1011 Would love to attend!

English

125

Vipul Gupta@vipul_1011·8 Mar

Finally have a name for AI nyc meetups now. Distilled.

Vipul Gupta@vipul_1011

NYC has insane AI talent density, but barely any high-quality AI events. Meanwhile, SF has 5–6 every week. That gap doesn’t make sense. So we are trying to change it. We are hosting a curated meetup for serious AI builders & startup enthusiasts. First one is on Feb 27! We are starting small and will slowly scale it up. Comment or DM if you interested. Share it with your NYC AI friends!

English

195

47.2K

Adrian Rogowski@AdrianMRogowski·2 Mar

@Replit @ReplitSupport what am I supposed to do about this? Github integration was depreciated this morning and now no longer works?

English

Adrian Rogowski@AdrianMRogowski·2 Mar

@clairevo “Gave it back to codex” meaning switching the model in Cursor?

English

claire vo 🖤@clairevo·2 Mar

Update on this: - cursor & opus got the proof of concept working - gave it back to codex and said “hey this is working if you were to rearchitect from scratch what would you do” - full day of turns & refactors - bugbot snagging edge cases - shipped & Devin does docs

claire vo 🖤@clairevo

People are always asking me why I'm still a @cursor_ai girl through and through, and here's why: I tried to do a pretty major refactor of a part of our app with complex server + client side interactions. I knew exactly how it needed to be architected, but it was gonna be A LOT of files for my little human fingers to execute cc plan + execute: failed, took wrong approach codex plan + execute: failed, took wrong and overly complicated approach cursor plan + build: one shot perfection The harness continues to excel at understanding intent and executing, especially if you know what you're doing.

English

20.4K

Adrian Rogowski@AdrianMRogowski·1 Mar

@Suhail I’ve solved for this with this spec driven development framework where the project maintains a .specs directory with feature specs and a combo of commands /spec-first and /catch-drift create and confirm documentation github.com/AdrianRogowski…

English

Suhail@Suhail·1 Mar

It feels like someone should make a post-git-hook where it asks the AI model to look at the diff of what you changed for a merged PR and update the repo’s various readmes and other documentation to make it easier for an LLM to be able to write code and reference things faster rather than reading every single line of source code that might be relevant constantly. The agents need their own docs.

English

356

54.1K

Adrian Rogowski@AdrianMRogowski·22 Şub

@ryancarson I created this spec driven development approach for that reason - enforcing/committing specs to the codebase and implementing with TDD creates much more opportunity for the agent to check its work and the user to understand what’s going on github.com/AdrianRogowski…

English

Ryan Carson@ryancarson·21 Şub

Documentation drift - it's your new worse enemy when working with agents on large existing projects. I'm using a mix of CI checks, custom linters and GitHub actions to help battle this. My Code Factory setup, which includes a Design System, is starting to work really well.

English

137

13.6K

Adrian Rogowski@AdrianMRogowski·22 Şub

@ryancarson I don't know why but Cursor agent completes tasks like 5-10x faster than Claude Code does for me...

English

Ryan Carson@ryancarson·21 Şub

What are you using as your daily driver now for swe?

English

32.4K

Adrian Rogowski@AdrianMRogowski·20 Şub

@ryancarson What part of the iteration loop are you trying to solve? Coming up with a good UI or confirming it's implemented correctly? I use a SDD approach where all spec files have gherkin and ASCII so I can understand from just text what the app will look like github.com/AdrianRogowski…

English

149

Ryan Carson@ryancarson·20 Şub

I'm going to work today at optimizing my Code Factory to be great at front-end UI work with my agent (Codex CLI). It's working, but it still feels like the feedback loop is too slow and kludgy. I'm currently using agent-browser, but the feedback loops are still too slow. I'm trying to figure out the right mix between the Playwright CLI, agent-browser, and possibly Storybook to really speed this up. Would love to hear from you all that have setups that allow you to iterate really fast on front-end UI.

Ryan Carson@ryancarson

x.com/i/article/2023…

English

168

36.2K

Adrian Rogowski@AdrianMRogowski·2 Şub

@msfeldstein Is this what you're looking for?

English

Michael Feldstein@msfeldstein·2 Şub

@AdrianMRogowski Hmm i dont see that string in our codebase. If you do Help -> Toggle Developer Tools do you see anything in the stack trace that may help? Possible you have an extension causing this?

English

Michael Feldstein@msfeldstein·22 Oca

What are your biggest paper cuts and quality issues you'd like to see us invest in fixing in the Cursor IDE?

English

11.8K

Adrian Rogowski@AdrianMRogowski·2 Şub

Unpopular opinion: prompt engineering for code is just... writing specs. PMs have been doing this for decades. We just never had a feedback loop this fast. "I know what I meant" has always been the enemy of good specs. Now you find out in 30 seconds instead of 2 sprints.

English

Adrian Rogowski@AdrianMRogowski·2 Şub

@msfeldstein I get multiple of these every time I open Cursor, I've had to downgrade versions every few weeks and now whatever the latest 2.4.x I was on just crashes my computer (overheats/freezes my M2 with 24gb ram). I went down to 2.3.35

English

Michael Feldstein@msfeldstein·23 Oca

@AdrianMRogowski share a screenshot?

English

Adrian Rogowski@AdrianMRogowski·1 Şub

@gregisenberg Will the Russian evals always give lower scores?

English

GREG ISENBERG@gregisenberg·1 Şub

Someone will create the "AI agent Olympics" AI agents compete against each other in different "sports" aka tasks on the internet. 10M+ people will watch Polymarket or Kalshi or Draftkings will be involved May the best Clawdbot win.

English

245

944

88.5K

Adrian Rogowski@AdrianMRogowski·1 Şub

"AI coding tools don't work" is the new "my engineer didn't build what I asked for" Same problem. Same cause.

English

Adrian Rogowski@AdrianMRogowski·24 Oca

@bryan_johnson Make one small step, take a shower, make the bed, do some pushups etc. Don’t feel like you have to solve everything at once, just make one small step and start to stack wins

English

Bryan Johnson@bryan_johnson·23 Oca

I've been in the depths of despair with depression. A state where nothing matters. No one can say anything to convince you otherwise. For those of you who've also been there or who are there now, what advice would you give others about how to speak with people in this state?

English

2.5K

183

6.6K

1.2M

Adrian Rogowski@AdrianMRogowski·23 Oca

@mckaywrigley Mmm interesting, thanks!

English

Mckay Wrigley@mckaywrigley·23 Oca

@AdrianMRogowski yeah but this makes it completely tool agnostic

English

507

Lenny Rachitsky@lennysan·23 Oca

What's the most underrated/underhyped AI tool in your stack right now?

English

293

294

267.6K

Adrian Rogowski@AdrianMRogowski·23 Oca

@mckaywrigley @noelhatem @lennysan Doesn’t Cursor already do this in their browser?

English

585

Mckay Wrigley@mckaywrigley·23 Oca

@noelhatem @lennysan agentation.dev

QME

2.7K

Adrian Rogowski@AdrianMRogowski·21 Oca

@Angaisb_ Because I’ve been perfecting my cursor rules for months so Cursor agent works better than Claude Code out the box and Cursor will use the browser to test features against specs after a TDD cycle completes

English

Angel 🌼@Angaisb_·20 Oca

Cursor users, why are you still using Cursor? I'm not trying to joke about it, I'm genuinely curious

Claude@claudeai

The VS Code extension for Claude Code is now generally available. It’s now much closer to the CLI experience: @-mention files for context, use familiar slash commands (/model, /mcp, /context), and more. Download it here: marketplace.visualstudio.com/items?itemName…

English

445

1.9K

586K

Adrian Rogowski@AdrianMRogowski·8 Oca

@BenjaminDEKR Data flywheels and network effects

English

Benjamin De Kraker@BenjaminDEKR·7 Oca

What is the moat of *any* software company right now?

English

482

410

395.4K

Adrian Rogowski@AdrianMRogowski·8 Oca

@mntruell Cursor has no "opinion" on the best set of commands and rules. While that allows for flexibility, Claude Code is better "out of the box", however with the right set of recursive rules to create/update specs and tests before coding my Cursor setup is better than CC out the box.

English

Michael Truell@mntruell·7 Oca

Cursor seeks to be the best and most powerful way to code with AI. What are the ways in which we could be better?

English

741

1.4K

255.9K

Keşfet

@joseph_h_garvin @vipul_1011 @Replit @ReplitSupport @clairevo @Suhail @ryancarson @msfeldstein