Notes

This is the kind of medical advancement I get excited about.

These posts are the best. A great write-up of the computers in Jurassic Park.

RSI is here.

Any attempt by AI lab leaders to curtail model release or restrict usage by the government is almost always an effort at regulatory capture of a market, and we should stand against it.

Really cool, and an incredible implication for drone warfare. The world is not ready for a battlefield of millions of drone swarms autonomously targeting and destroying targets.

People should settle more. Paul Skallas is on to something here and it’s something my wife and I talk about a lot. Marriages have thrived for thousands of years that were not bonded by temporary Eros feelings. When I hear people talk about being single and having trouble finding their person, I just assume they’re evaluating the wrong signals around compatibility and long-term companionship. It’s great to feel lovey-dovey for your spouse but I also feel lots of other things too. All of them carry me through.

I think I’m getting an Impulse Stove when we remodel this fall. This thing is amazing-looking.

Working in a solutions architecture role sometimes means you will have teams come to you and say, "we don't want to give you that permission in a system for that particular integration." Cool! But just accept it won't work. But then they don't love that.

I can offer workarounds with usually more friction, but the integration is the integration.

It always surprises people.

"What do you mean you can't read the table of data I refuse to give you read access to?"

The best is when people will say, "well, it works when I do it from this api over here." And then you have to gently point out that the api doesn't actually have the data they think it does, or isn't supported for partner integration, or a dozen other little things that are always annoying.

I'm really sympathetic here, honestly. It's annoying when you get stuck in binds of wanting to lock things down and also provide useful access, but sometimes those things are mutually exclusive.

Trying GPT-5.6 Sol on High, and honestly, I haven't been too impressed yet. I asked it to do some simple research, and it lied and said it checked all these sources, and then didn't do any of it.

It did a terrible job of looking up the info, and then when I called it on its laziness, it basically said, yeah, I didn't really try very hard. Really lame stuff.*

Update: Just tried their new knockoff cowork clone in chatgpt/codex and it did much better. Agent harness matters a lot here.

I have started jiu-jitsu. I’m going to be so sore tomorrow.

People swear in general conversations a lot more now. Twenty-five years ago I rarely heard anyone curse casually. Now it’s every day for many people I’m around. I don’t like that it became so casual. We should go back to less swearing in public.

Highly entertaining 9yo photos are taking place.

Data still matters and we're running out, from what I can tell.

The easy good data has been collected. Synthetic data is doing its best to fill the gaps. The next big task will be backfilling real-world data that never lived well on the web the way code did.

The other day I was talking with a family member about how good AI has gotten, and it really has, but he was complaining how it wasn't as good for him and people in his orbit of agriculture.

As we talked about it, we started talking about a niche hobby of looking for arrowheads with AI. So we tried it! It did terribly. Now, this was not a rigorous test, but to me it showed how challenging the data gap is for information that never lived naturally on the web for training.

The next frontier of AI seems to be specialized hardware and specialized data. For now, people will collect and curate this data, but it strikes me that the next generation of mechanisms will need to allow for AIs to source their own data and build tooling to discover and learn this information.

Regime change is always hard.

So it begins.

There are three arrowheads in this photo. I asked Fable on Claude to find them. It got one right with the #2 finding. It missed the other two and had 2 false positives. Can you find the others?

My son and I finished The Well of Ascension today. My son has been absolutely sucked into the series. It's 38 AR points, too. It also helps my son with Book It, which he's pumped to be doing as well.

We start The Hero of Ages tomorrow!

I had to upload a large file today from a computer at our church. It was a video of my son's first sermon. I hate having to do the whole login song and dance with Google Drive. So today I built my own tool called warp to more quickly and easily pick up files from unowned computers and grab them later.

I love living in the future with AI when building becomes essentially free. It feels like becoming a machinist for tools. Any weird custom tool you want is just a few prompts away.

It's the 4th of July weekend. I hope you're watching The Sandlot.

My son preaching his first sermon!