lorenstewart

A look at different ways to work with agents with reference to cultural context.

Sometimes you shouldn't follow conventional "wisdom"

A maintainer refactored the Analog app, so it's doing much better than before. The post has been updated.

There are two measurements: the size of the index page, the size of the board page. You are quoting measurements of the board page.

Related: 3G === spotty 4G (and you know from your own experience this is true)

I don't really know what to say except that you're wrong here. There are many studies out there about loading time affecting conversions/sales. I link to the most recent study done by Speed Curve about the damage slow loading apps do to business metrics.

Not wanting to believe something is very different from the thing being untrue. The differences between 13kB and 500kB are quite real and quite measurable.

Jeez! Y'all are quite harsh. I wrote this in sections and then tried to stitch it all together. I did my best to remove repetition, but overall it's my writing edited with the help of AI to moderate some of my personal biases. The value here is both checking out the code (written by me with the help of the maintainers of each framework) as well as tables for lighthouse scores and bundle sizes.

I personally think the most important parts of the post are these two related parts: (1) the technofeudalism section (2) developing lightweight apps so a business does not feel the need to have a separate native codebase along with a specialized native team ($$$). I was hoping this post was a gift to many people by doing all this work for y'all, or at least providing a foundation from which you can fork and modify for your own use.

It is an open question for everyone if start up time on cellular is something you need to worry about. All the other non-sense here misses the point.

I don't have anything deployed bc it uses AI which costs money. If you get yourself an Open Router API key you can use that in the .env files. There are instructions to get the apps running locally; it should be straightforward but let me know if you have troubles

Four part blog series, start here: https://www.lorenstew.art/blog/llm-tools-1-chatbot-to-agent

Full application with authentication and streaming is all set up. All you need is an Open Router API key. Then update the logic in the MCP servers to suit your needs. You'll have your own system up and running quickly.