Assorted Links

• I finished book eight of Dungeon Crawler Carl. Loved it. I have recommended the series to a lot of different people now with a pretty good hit rate. If you haven't read it yet, you should.
• This was a fun read https://tastecooking.com/i-want-to-live-like-costco-people/
• I've been in the mood for something casual to play so I picked up Fields of Mistria. If you've played Stardew Valley it'll be familiar. It has enough differences to be interesting and a good way to unwind.
• I am very far from being an expert in K-Pop but I loved the way this story was put together: https://pudding.cool/2026/05/kpop-generations/ I'm generally a big fan of any examples where the web really shines as a medium for telling stories.
• Speaking of which, here's another one, this time with a more technical goal: https://myzopotamia.dev/navier-stokes-fluid-simulation-explained-with-godot

Things I'm working on

• I've been working on a project applying the whole 'auto research' idea to compression (think zip files etc). I wanted to see how things went when the task had a less clear gradient signal and how well AI could iterate towards a solution. Write up coming soon.
• I've also been experimenting with using Claude as a coach for running and learning Japanese. I would say both are good but not great. They collate a lot of public info and communicate it well but I've called it out on a few mistakes in both cases which make me hesitant to use it for things where I have zero prior knowledge.
• I am in the middle of a week off at the moment between my previous job and a new one. Spending the time doing a lot of reading about how AI is (and isn't) having meaningful effects on engineering teams and product teams. There is a lot of noise in this space, that's for sure.

In case you missed it (my posts)

Momentum is the Safest Moat

If you are at all worried about AI disrupting your business, learn to build momentum. Your ability to execute more reps that your competition is, and always has been, the best moat. The trigger for this post was hearing that Claude Code was built in a weekend. It’s probably mildly

Elliot C SmithElliot Smith

Don’t sell the build effort.

There’s a trend I have noticed when folks try sell software, their perception of how hard it was to build the software influences their comfort in setting a price. More difficult to build, more expensive. Easy to build, we can’t charge a premium for that! This is tangentially related to

Elliot C SmithElliot Smith

Build ‘Taste’ for the critics but ‘Care’ for the fans.

There are a lot of folks out there tell you that what matters now is ‘taste’. Some loosely defined term that boils down to “know if the output is slop or not slop”. While it is probably a useful skill, I’d argue the thing that will ultimately determine if your

Elliot C SmithElliot Smith