Kyle Aster

Out here spinning up VPS's and running commands I have no understanding of like it's 2005.

@tael@yiff.life @mattly@hachyderm.io I don't think it's too complicated. LLMs will read documentation and instructions. People usually won't.

I've been working on describing how I've been building new things with LLMs. I call it my product builder. It works really well for me.

https://github.com/kneath/product-builder

A friend turned me onto this short a few days ago. In case you ever wondered why I live in the mountains and treasure our public lands so much…

https://vimeo.com/198061884?fl=pl&fe=ti

There's a discontinuity right now in how people are experiencing this whole AI-built-software thing. Some of us came to this game to write code. Some of us came to the game to make our computers do weird stuff.

Neither are right or wrong (you need both to build a successful commercial enterprise), but for a couple of decades we've been on the same team out of necessity — even when it didn't need to be a commercial thing.

Now we kind of don't need each other, but we're still in the same room.

I really enjoyed Search Engine's take on what it means to "replace" drivers with AI. Happy to see other people in the world who love context and difficult problems.

I get frustrated with so much talk about AI "replacing" humans because we've been doing that for centuries (the lamplighters!) and we've never reconciled what "jobs" mean in our modern capitalist society. Is it in service of real work? Or a social contract so we can pay rent?

https://www.searchengine.show/are-you-a-good-driver/

Accidented into having dinner with one of my idols last night and still pretty jazzed about it. A good reminder to keep showing up.

Get after it.

https://www.youtube.com/watch?v=f4c2nAl0jew

@Brilliantcrank@mastodon.social It's absolute insanity and I'm loving building software more than I ever have before. Better software than I've ever built (by myself). I did the same a month ago and it changed everything I ever thought about "AI": https://github.com/kneath/knyle-share

I see so many people saying they're burning tons of tokens on agents "understanding" their app, and I have to wonder: have you asked the agent to write docs + design patterns for your app?

An incredible number of problems can be solved this way w/ LLMs. They're little context goblins.

@yvg@indieweb.social My honest theory is the “shadow ceo” era bred a lot of leadership that were desperate to make their mark on the product. They made a series of incredibly poor product decisions that resulted in unstable, unprofitable, unmaintainable features that sidelined the core product. Were actions kind of cool? Sure. Did they make ops 100x harder? Definitely. Were they profitable? Not really. Could say the same for a lot of what happened. Just my theory from the sidelines, though!

Sometimes I wonder if @github is nerd sniping me into creating a new GitHub after a decade or two of being out of the game.

Here’s my suggestion to those who remain:

- Data integrity is non-negotiable
- Code should render
- PRs should render
- Issues should show correct counts
- Login link should render in <300ms
- Login flow should work
- Registration flow should work
- It’d be great if it worked on bad connections, at all

Maybe add a new zen of GitHub “It’s not a product if it doesn’t work”

Lisbon is just a stupid beautiful city.

@Luke@typo.social Thinking about you, in Lisbon.

Here’s one counterpoint to ai-is-destroying-humanity. My agent is always present, always trying. I can have an intellectually stimulating discussion that leaves me with a new understanding about the world.

The vast swarms of humans are so dissociated they can’t perform basic societal functions. People can’t walk in shared spaces. They can’t hand over objects. They blast full volume phones of infuriating tik toks in quiet settings.

@tclark That man does have a way with words

Three decades of programmers building riddle machines has almost entirely unwound in the space of two months.

I know a lot of you that follow me here are pretty anti LLMs, but I gotta say — if you're still digging your heels in for philosophical reasons you are missing out a phenomenal sea change that's happening right now.

Is it perfect? Are there not terrifying futures ahead of us? Don't be obstinate. The world is complicated.

As someone who grew up writing CSS for Netscape Navigator 4.0 and IE 5.5 on Mac — it is impossible to describe how miraculous this feels.