Like many terms in the ai space I’ve been trying to better understand spec driven development. And my takeaway is we should be calling it design doc driven development or spec-anchored driven development (thanks Birgitta Böckeler). I think “spec” can easily drift in meaning and interpretation.
Every couple of weeks I think to myself that I should post more of what I’ve read about engineering and AI only to remember that 90% of the useful links I discover are via @simonwillison.net
Simon is the single best source for keeping up with AI engineering.
They can have the borg speak multiple languages but they can’t pronounce the name of a little land locked country. IYKYK (Pluribus)
Anyone else having issues with claude code and the scripts folder in skills? All the examples show a relative path (python scripts/blah.py) for invoking but claude code complains because the path is relative to the project not the specific skill folder. Seems to be fine with plugin installs.
My current favourite combo is Claude code and AppleScript. “Find all tabs in safari related to [topic] and append the title, url and short description to today’s Obsidian daily note”
You’ve changed.
Maybe I’m projecting but, how have you picked directions in the past? What caused you to dive into react or bsky or write longform (and greatly appreciated) posts recently? I have to imagine there’s an emotional force, a passion/joy that guides some of that. What about that energy?
I’m getting a ton of value from the posts but it’s a slog. Because they’re long I break it up into sessions but the structure and style requires maintaining some context and history of the changes. Sometimes I come back and I’m lost requiring rereading earlier sections.
@simonwillison.net is there a way to combine a geojson with lat/long values in datasette? I want to show a bunch of markers as well as city boundaries. I can view each separately in their respective tables but not sure how to combine them.
Introducing MyCoder - a very capable AI agent-based coder that just works.
1. npm install -g mycoder
2. export ANTHROPIC_API_KEY=[your key]
3. cd myproject
4. mycoder "fix any build errors"
github.com/bhouston/myc...
We live in a world where technology is already helping us achieve tasks faster and easier than ever before. We can order groceries for the week from our laptop in minutes. No agent needed.
I think the issue with consumer facing agents is a mismatch in task scope and ability to articulate the task. I think agents show their value with tasks that take days not minutes. Most people don’t have big enough tasks or struggle to articulate the process/outcome in a meaningful way for agents.
SSX Tricky! In my early 20's I had a PS2 but couldn't afford the games. For months we played the SSX Tricky demo (2 min limit) over and over until the wee hours of the morning. I would dream of flying squirrels. youtu.be/PDFUgsE6odU?...
As an FX artist who used to spend days crafting fluid simulations (fire, water, smoke), the gen AI stuff often falls short. The coke ad always make me cringe. But, this… This, is really good.
I’d argue 90% of the time an ‘agentic’ framework isn’t needed. I had to learn the basics about orchestration and distributed systems to realize that.
The initial appeal was easily constructing/modifying a graph. Define your fn’s, define the flow/relationships.
It took time for me to understand agentic flows, thinking complex systems needed a special framework.
“According to EPRI, a single ChatGPT query requires around 2.9 watt-hours, compared to just 0.3 watt-hours for a Google search, driving a potential order of magnitude more power demand.” about.bnef.com/blog/liebrei...
From the article:
Several screenwriters who’ve worked for the streamer told me a common note from company executives is “have this character announce what they’re doing so that viewers who have this program on in the background can follow along.”
I really want to enjoy Black Doves on Netflix but I can’t get this article out of my head. It’s embarrassingly obvious in the first couple minutes of the last episode. Ugh.
www.nplusonemag.com/issue-49/ess...
I think this is why some agentic workflows produce seemingly novel results - they’re designed to iteratively explore until a condition is met. We don’t necessarily do that when chatting with an llm.
My take on this is that we’re iteratively exploring the latent space. The initial prompt and response puts us in the ball park. By asking “write better code” we’re probing the neighbouring, high dimensional space.
AI produces fewer greenhouse gas emissions than humans! Humans emit 27g of CO2 in the time it takes to write three hundred words. ChatGPT, however, performs the same task in 4.4 seconds and produces only 2.2g of CO2. www.nature.com/articles/s41...
From: kenthendricks.com/52-things-i-...
I copied the files into Claude, laid out my steps and had a mostly working refactor in seconds. What would have been a half day or more was done. But that tension still lingers. Anyone else feel this?
I still feel like I’m learning to let go of how I used to interact with code. I had a pretty big refactor I wanted to do and even though I could see the steps there was still an anxiety or resistance to getting stuck in. Then I realized I could hand this over to Claude.
I think htmx and alpinejs are good additions. Lightweight, inline, llm-friendly.
Is there any way we can see some of these experiments? I’d love to get a sense of the the problem statements, how they’re structured, and what makes for a good response. I’m not an academic but I feel like the methodology could be transferable.
At first I wasn’t sure what Command offered over conditional edges but after reading docs it makes sense. BUT, I then discovered interrupt() and that just feels like a minefield of footguns. I’ll likely put it in the wrong place in a node. Or worse, run into issues with tracking multiple interrupts.
Do kids learn about markdown at school?
I picked it up over the years because I like the simplicity. And tools like Obsidian and iaWriter are my favourites.
But AI has kinda made it mainstream and I don’t know if kids know or care about it.
Anthropic > web containers > js
OpenAI > wasm > python
