Everybody has been talking for years about how declarative programming is the future, but nobody anticipated it would mean declaring, “Claude, make me an app.”

I finished yesterday’s NYT crossword in 3:17!

If mifepristone caused erections we would not be in this situation. Politicians only fight against women’s health, never against men’s.

Jay Z’s “Empire State of Mind” is so good that I can sit through a whole YouTube video even after suffering 110% exposure by street merchants in Times Square. www.youtube.com/watch

The greatest practical downside to agentic AI in programming is the unpredictable delay between request and response. When I worked at Apple in the late ’90s, we would start a build at night to be sure it was fresh in the morning. During the day we would play video games while waiting.

AI offers incredible powers to people today who ALREADY know how to do their jobs. The real risk, IMO, is that a whole generation will never learn how to actually do the jobs.

I just left the sheet music to “Linus and Lucy” on the piano and now I hear intermittent attempts to play it.

Inspired by my kid’s impromptu crossword creation, I decided to take another stab at construction myself. Now my kid is test-solving it. I’ll submit it to the NYT first because of course I want that glory, but then I’ll see if anybody else wants it!

I admit I was anxious to post about my discomfort referring to trans or non-binary kids in the company of their parents. I thought that people who have a much harder time living a life where their identity is constantly challenged might find my concern too trivial. I’ve had good feedback so far. 🙏

Be right back. Just placing seashells on a beach spelling out that we should “86” — and by 86 of course I don’t mean kill, because that would be the stupidest interpretation – the Supreme Court justices who voted to eradicate the Voting Rights Act. And I know we CAN’T 86 them, but still. Seashells.

A very modern social challenge is learning about your kid’s friend being trans or non-binary, and wanting to respect their pronouns, but not sensing any buy-in from the parents. I’m always wondering if they’re misgendering their kid “for my sake” or, because they don’t know yet, or worse…

Unbeknownst to me, my son (17yo) learned enough about crossword construction to make a puzzle, print it with MY APP, and leave it as tribute on the table, waiting for me when I returned home. Very good for somebody who has never had any training or practice, as far as I know!

If you’re a parent who, for whatever reason, is lucky enough to have a piano in your home, the sweetest sound imaginable is the sound of your kids playing the piano.

No idea how likely to be true this is, but I thought it was interesting that AI knowing so much about a particular framework or technology might extend its lifetime, regardless of future frameworks' merits. bsky.app/profile/m…

I’ve had great experiences with Apple’s “Express Replacement” in the past, but now that I’ve been waiting 10 days for a replacement, with no updates, I’m wondering if other people haven’t had the same stellar service I used to enjoy.

Assistant TO the regional CEO.

Not surprising, given the number of times we agreed about it on Core Intuition, but I agree 100% with @Manton’s judgment of acceptable semantic versioning. The only one I suck it up and defy (holding my nose) is “2.1.10”. www.manton.org/2026/04/1…

As my oldest son gets ready for college, I’m thinking about the fears surrounding going out on one’s own. Having to, all at once, pay all the bills while hoping for some relief. He’s lucky to have a backstop in me/us, but so many kids don’t. College should be universally funded.

The biggest marketing coup in the AI era is Anthropic calling its agentic, command-line tool, “Claude Code.” It’s a diversely capable model that can be used for ANY matter of tasks, but calling it “code” made it feel like the perfect choice for coders. OpenAI’s Codex doesn’t hit the same mark.

AI skeptics often express a fear about how it might do something bad without you knowing. You can’t “trust” it. It’s absolutely, 100% true. But the question of whether you trust it or not is entirely in your hands. You CAN review every word, code snippet, etc., before it does it. If you choose to.

The utility that Claude built for me features a tmux (terminal multiplexer) managing orchestrator that dispatches all of my download plugins to do their thing, and monitors the result. It lets me dive in to their logs, connect to the live tty, or kill the pertinent process.

That statements downloading workflow is an example of a pattern in my use of AI: ask it to build the necessary tools for me to do what I want, with my private data, when I want. Don’t ask it to DO what you want. Ask it to AUTOMATE doing what you want.

For years I’ve been grappling with the problem of maintaining a local archive of all the digital documents that used to naturally archive themselves by showing up on paper. I got Claude to help me orchestrate the whole thing, and it’s glorious. Feed this prompt to your LLM of choice: gist.github.com/danielpun…

My kids have been incredibly successful so far. I’m sure they could have done “better” by somebody’s standard, but I never wanted to stress them out. They enjoy doing well, and many times that means getting an A, but sometimes it means getting a B. They are perfect.

I “just” got back from the ACPT: America’s crossword tournament. It takes a while to sink in, because it’s a surreal experience living, eating, drinking, and most importantly SOLVING in the midst of so many amazing crossword enthusiasts. Even the celebrities there are accessible. So cool, so nerdy.