Thoughts on AI

Looks interesting. Yet another thing to put on the list to check out.

This was one of the easiest Rails upgrades I’ve done yet. Also, since I’ve been posting about using AI agents, and here’s what deploying the newly updated app looked like.

You just _know_ Doctorow is right: we’ll be getting sold stuff by AI with little visibility in no time. It’s hard to see what the incentive structures are, and I have no doubt that moral clarity alone is not enough to deter the leaders of those companies from tapping that keg.

As part of my ongoing experiments using AI agents, I created two specialized agents that work together to automate an entire release process—from test verification through production deployment. Read more about it here.

Over the past few days, I’ve been experimenting with a new-to-me workflow: using [git worktrees](https://git-scm.com/docs/git-worktree) to let AI agents try multiple parallel implementations of the same feature. It’s been interesting enough to share.

We focus on the times when LLMs do a bad job, but since humans are deliberately doing a bad job pretty regularly, this might be a good use for the tech.

I found this episode of “On With Kara Swisher” to be an excellent listen. Yann LeCun, Meta’s chief AI scientist, is a fascinating guest and Swisher did a good job interviewing him. It’s definitely worth a listen if you’re at all interested in LLMs and AI.

It looks like you’ll _have_ to pay the $200/mo if you really want to use o1 for much. I’ve been using it this weekend on the $20 plan and got the following: “You have 5 responses from o1 remaining. If you hit the limit, responses will switch to another model until it resets December 10, 2024.”