My favorite phrase to say to Codex/Claude Code is “Pareto improvement” I use it everytime I’m planning, reviewing, or prodding the agent to improve and cleanup my codebase. A Pareto improvement is a change that makes one aspect of a system better, without making any other aspect worse.

Finally found a productive workflow for QA’ing frontends with agents. Can’t believe it took this long for me to find a working setup. In Codex & Claude Code, there’s a few ways the agent can help QA your app in a real browser: 1. First-party Chrome extension 2. In-app browser (only on Codex) 3. Playwright MCP 4. Chrome DevTools MCP The first two sound tempting, because they’re first-party to your agent. But for manually testing a web app, I think the best option is DevTools MCP. It not only lets the agent drive the web browser, like all the rest do, but it’s got access to the full suite of dev tools in Chrome, including performance/network tracing, reading console logs, and debugging. Works super well with Codex /goal: you can ask it to improve perf to hit certain metrics, and it’ll not stop until it hits those metrics.

Addicted to telling Codex or Claude Code to “get a subagent to do [task]”, especially since it keeps the main thread clean and you can continue chatting in it. Makes me wish regular ChatGPT and Claude had this. This pattern is helpful for ideation/research/planning tasks too!

Just tried Claude Design. Burned through my Pro weekly limit in one session, but boy was I happy. I don’t think Figma is going to be replaced by this anytime soon. It still works best if you give it hand-designed layouts, color choices, type choices, etc). That said, the experience is miles ahead of Figma Make. Only because it makes a serious effort of structuring your design system and it makes creating one from existing artifacts super duper easy (GitHub repo, Figma files, brand assets, copy samples). I’m an indie maker, and putting together a design system has always been a bit of a chore, but this has legitimately taken a lot of work of my plate. Now, if only Figma worked this way! Autogenerate or update a design system from the decisions I’ve made in my mockups and my repo. A boy can dream!

I’m excited for the agentic revolution remaking software as we know it. The argument for user-configurable or user-extensible software has always been a bit flawed, because it assumes that if someone finds off-the-shelf software distasteful, or disrespectful of their needs, then the answer is to make tweaks atop that software. But software expresses affordances and conceptual models that are tightly coupled to a data model, API semantics, and architectural capabilities. Which means if said software is wrongly-opinionated for a use case, it’s unlikely that making userland tweaks or extensions will help much. When agents become users that consume APIs directly, software that’s designed in a fashion disrespectful of their users’ goals, needs, or mental models will become more intolerable. Any hard limitation of those APIs or underlying data models will become more stark. When users’ experiences with software used to be mediated by UIs alone, it’s always a bit opaque as to how that software can evolve or be modified to work differently in a way that more closely matches your goals or the way you think. In that world, a different UI seemed to be the answer. But at the end of the day, poorly-made software is hamstrung by its fundamental design at the system level. The UI can only express what the system is designed for. Competition between software tools will become much more perfect, when these design flaws are quickly discovered, as agents run into a wall with an API that can’t express the user’s intent.

Folks like to compare Anthropic to Apple but I think OpenAI has more of that Apple design DNA. The Codex macOS app such a treat to use with a ton of little design details. They ship new workflow features more slowly compared to Anthropic, but they do it more thoughtfully and usually with a considered UX Anthropic UX feels more like Microsoft in this case: drawers chock full of stuff, lots of it powerful, but they aren’t necessarily well thought out from a usability perspective

Codex sub-agents are such a treat to use I love how you can steer each one individually Makes it possible to start with a coarse ask, have the agent break up the work into small orthogonal parts, then deep dive into each one It’s also super cute that each sub-agent gets a name like “Erdos”, “Halley”, “Hubble”, etc: it’s a small detail but it makes it possible to refer to them easily!

I think people are making too much of a big fuss over “taste” and how that’s a comparative advantage for humans over AI models I think “taste” is simply an act of prediction over learned patterns of human preferences (I go into detail more in the article) Corollary of this: AI models can pattern match over a larger dataset of human preferences than any single human will ever be able to, and will win out over humans every time. Humans only have an advantage over AI models insomuch as much of these human preferences are still tacit and granular and difficult to put into words Anyone who’s tried to prompt an AI model for a song, an image, or a video exactly to taste will know how difficult it is to explicitly state what your preferences are. But recommender algorithms are already better at recommending and curating content than human curators, using proxy signals like likes, views, etc Perhaps it’s a matter of time before someone cracks the code of how to infer a user’s preferences (without explicit prompting) for creation (not just curation) When that happens, there’s no need for a middleman human to make creative decisions. The end-consumer can simply ask for what they want, and the AI models will provide it, to taste.

This is beyond cool. Designs collocated with your codebase! Visual designs are a huge part of spec’ing out a UI If specs should be collocated with the code so that agents always have the context of what they’re building or maintaining, then it makes sense for the design files to be collocated with it too

I had this thought too, while working on a reasonably complex feature that necessitated spiking out multiple approaches to a problem Agents let us try many different approaches quickly and that’s a good thing But not having failed approaches be easily accessible as context to the agent is a missed opportunity It’s one thing to describe in a spec .md file that an approach didn’t work, it’s another to have that available to the agent to inspect at anytime

Why didn’t Adobe do this with Firefly?