Finding signal on Twitter is more difficult than it used to be. We curate the best tweets on topics like AI, startups, and product development every weekday so you can focus on what matters.
Page 1 • Showing 20 tweets
<rant> MY KINGDOM for one agent that can actually do e2e design -> pixel-perfect frontend code (including auth so it can login to the app). I'm currently using @paper + Claude Code + Codex CLI + Codex Mac App + agent-browser + Claude Code Chrome extension + ... It's ridiculous. I really, really, really want one ADE to own all this. Come on people. You literally have billions of dollars collectively in funding to sort this out. I'm working on a @nextjs app - this is literally burned into the weights of the models - not doing something new or novel here. Maybe it's a skill issue on my side. Honestly I doubt it though - I spend 10 hours/day in these tools and I've been coding with agents for 3 years. Part of the problem is that I want the UX to be pixel-perfect, beautiful, and elegant. I'm very opinionated about the design being perfect. Maybe the problem is that I've been a backend developer my whole career, so if I was a front-end developer, maybe all this would be easier. The point is that we need the agents to allow anyone to write perfectly *designed* front end code. Also, to be clear: I have a design system skill. I have clear and concise documentation. I don't have any documentation drift. I don't have bloated agents md files. I'm doing everything correctly, and this is still WAY too freaking hard. </rant>
What are you using as your daily driver now for swe?
Getting accounting setup for my startup. I use @mercury for banking so thinking about using their integration with @QuickBooks. Suggestions?
I'm seeing a massive speed up of Codex 5.3 right now on the CLI. Has anyone else seen that? It almost feels like Spark.
Old world: 1. startup founder finds a co-founder to close their skill gaps 2. raises money 3. launches MVP 4. hires a bunch of specialists to close more skill gaps New world: 1. startup founder builds entire product with AI 2. launches and gets basic PMF 3. hires 1-2 generalists that can run teams of agents 4. maybe doesn't even raise capital
OK, what's everyone using to code right now?
The next-devtools mcp is killer. If you're building in @nextjs check it out. Nice work @vercel_dev
We're now seeing the rise of the ADE or the "Agentic Development Environment" which is the natural child of the IDE and TUIs in this new agentic age. As you all might have seen, I've been working on terraforming my repo to be Code Factory enabled. I think the big harnesses like Codex and Claude Code will roll this out natively and a lot of the short-term kludges that we're all inserting to make Code Factories will go by the wayside and will just be part of the new ADE. I personally can't wait for this - so that I can stop duct-taping everything together. Meanwhile OpenAI shipped this skill which I think is awesome and I'm going to integrate it into my Code Factory I would thank Eric Traut but he's not on X 🤷‍♂️ https://github.com/openai/codex/commit/7…
Add PR babysitting skill for this repo (#12513) · openai/codex@7e569f1
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.
Wow. Storybook + Agentation + Codex 5.3 Spark is amazing for frontend UI work.

Anthropic could do one simple thing to start winning back the hearts of devs http://claude.md -> http://agents.md
Here's how to use your ChatGPT account with @openclaw! openclaw onboard --auth-choice openai-codex openclaw models set openai-codex/gpt-5.3-codex openclaw models status --plain h/t @AndrewWarner @calebhodges
I'm just going to say it because I'm so frustrated. Anthropic is terrible at DevRel. I think we all saw what happened with OpenClaw, but I've also seen it personally. I've DM'd and tagged a bunch of folks at Anthropic, and I've literally never received a a single reply or DM. On the other hand, look at Codex. I switched over to Codex recently and started trying it out. I love it. So much so that I've switched my entire dev workflow to the Codex CLI. I posted a few tweets and quickly @gdb @hillarycbush (who's a friend, but nevertheless, nice :D) @embirico @dkundel and more immediately started to engage and get my DMs. All of them asking for feedback and offering support. I just don't understand what's happening over at Anthropic 🤷‍♂️ (FYI I'm paid $0 to say any of this. I have received $2.5k in API credits on OpenAI but I'm using my paid ChatGPT Pro account for Codex usage and haven't spent any of the $2.5k yet.)
got something fun to announce soon :)
If you’re not frustrated when you’re working on your UI with your agent, you don’t care enough about the details. Good UI and UX is a long, painful grind.
Having an AI that schedules meetings and coordinates my calendar has been a massive unlock for me. Currently using and paying for @blockitAI I've had lovely human EAs in the past and I have to say that an AI just works better. Instant replies. Never forgets. Always working. Hard to compete. This makes me both happy and sad.
Starting to aggressively block llm accounts in replies to my tweets. Hoping this dings them on the algo.
When I post on X and I type on my keyboard, I tend to not bother to use proper grammar or capitalization. Then, when I tweet using STT (like I'm doing now with @usemonologue), it's perfectly formatted. This bothers me.
Documenting your design system in .md is a huge unlock. Then making a skill for it is even bigger unlock. So happy. Thank you @joshpuckett
Codex CLI + agent-browser (by @vercel @ctatedev and team) + agentation (free tool from @benjitaylor) is so powerful for UI fixes. You click the element you want to change, add a comment, copy the comments and paste into Codex. It just works.
