Shawn Simister
@narphorium
Building AI powered tools to augment human creativity and problem solving in San Francisco. atelier.dev Previously @GitHub Copilot, @Google, 🇨🇦 narphorium.com
Keyset on the left, mouse on the right, like Engelbart's setup in the Mother of All Demos
My transparent Creator Micro keys arrived, so now I can make atelier.dev decision points map to physical buttons
I guess what I really want is a portable Dynalamp dynamicland.org/archived-med...
Just realized that my portable table lamp would make the perfect vessel for a voice-powered AI assistant
Anything can be a walkthrough. It's just a views: walkthrough entry in the Markdown front matter and then every ## Subheading becomes its own step
Atelier walkthroughs can use the Chrome MCP in Claude Code to build UI walkthroughs with live screenshots of the app. Grounding the walkthrough in execution exposes gaps in the implementation that you wouldn’t catch with code review.
You can verify any task on the atelier.dev kanban board and it will automatically build a walkthrough of the acceptance criteria grounded in deep links back into the code that it wrote.
Every time an agent explores your code base it builds up its own mental model of how the code works, and then throws that model away when the session ends. Code walkthroughs turn that model it into a step-by-step guide to bring you up to speed on any part of the system.
I feel like over the years, we've sort of given up documenting our mental models of code and just accepted that everyone who reads the code builds their own model from scratch.
In 1985 Peter Naur argued that a program is more than just its source code. "Programming As Theory Building" explained how we build theories of the code which help us debug and refactor it but those theories rely on knowledge from outside of the code. pages.cs.wisc.edu/~remzi/Naur....
finally a dedicated device for the 20X per day you need to open MS authenticator 😄
I sketched this out a few years ago. The HTML vs Markdown debate is conflating substrate with information density. The real question is what kind of feedback an artifact actually invites. Hi-fi invites parameter critique. Lo-fi invites paradigm critique.
Most importantly, everything is editable so I don’t need to run another prompt just to get the agent to make some simple adjustments. If it’s faster by hand, I just edit it by hand
I can add new annotations and ask for revisions to existing ones at any time. The tool follows my thinking process instead of trying to impose its own workflow
Each group of edits can contain many individual changes. I can accept some changes and reject others. It’s not all or nothing
This is how I use AI to augment my design process. Instead of having the agent show me a giant diff of changes, Atelier links each edit back to the original thread of feedback so I can review them in context
The CLI is great for a lot of things but reviewing Claude Code plans is not one of them. So much scrolling back and forward just to figure out what the agent changed
I just released v0.3 of atelier.dev with the new inline revisions feature available for everyone to try
Revising a document with an agent should feel like collaborating with someone sitting next to you reading the same page. You highlight what needs to change, the agent suggests revisions, and you decide. That's what I'm building in Atelier
I want my workflow to feel more like this. One big reconfigurable space for deep work
When the agent finishes a task I launch atelier-verify to automatically verify the acceptance criteria and move it to the done column if they're all met
Not a fan of this new effort selector in Claude Code. It's cool that the models can adapt to different tasks but it's annoying that you need to think about which setting to use before you send each prompt. Why can't I just define which tasks are high effort in CLAUDE.md?
Plans update to show how many tasks have been completed. Tasks show how many acceptance criteria are met.
I added realtime notification so that when I'm running multiple agents in parallel I can scan the board and see which ones need my attention. Clicking on the yellow text automatically switches to the terminal for that agent
