phren0logy

RIP Popular Mechanics

I'm not a coder, I'm a medical doctor. I see some interesting parallels in how medical students sort themselves into specialties by cognitive style to this new rift in programming with LLMs.

Some people like the deep work, some like managing a steady rain of chaos. There's no one right answer. But I'll tell you that my classmates who are happy as nephrologists are very different to the ones that are happy as transplant surgeons.

I tried it out a bit. Like SQLite, it can be used as an embedded db in a flat file. It will be interesting to see what Apple does with it. Their blog had some great examples using BAML.

Compared to what?

Most of this hype appears to be coming from grifters who aren't actually connected to the project. So, it's there, but not the fault of the people doing the work.

This has come up in a few recent statements by the project lead, including scammy memecoins and name-sniping. One source:

https://www.theregister.com/2026/01/27/clawdbot_moltbot_secu...

Gas Town has a very clear "mad scientist/performance art" sort of thing going on, and I love that. It's taking a premise way past its logical conclusion, and I think that's fun to watch.

I haven't seen anything to suggest that Yegge is proposing it as a serious tool for serious work, so why all the hate?

Still a compelling value. I keep eyeing it to see if I should switch from 1Password, but I also worry about the complexity of transitioning the whole family over.

I enjoy reading this the same way I enjoy reading "Boatmurdered" - https://dwarffortresswiki.org/index.php/Bloodline:Boatmurder...

I think I'll hold off before jumping in, but it's fun to watch from the sidelines.

I had plenty of oral exams throughout my education and training. It's interesting to see their resurgence, and easy to understand the appeal. If they can be done rigorously and fairly (no easy thing), then they go much further than multiple can in demonstrating understanding of concepts. But, they are inherently more stressful. I agree with the article that the increased pressure is a feature, not a bug. It's much more real-world for many kinds of knowledge.

I tried a number of distros and settled on Omarchy because it has a coherent design and nice aesthetics, but it has some weird quirks about messing with my dotfiles on updates. It's so new I suspect this will be ironed out soon.

The recent direction of MacOS has been a good excuse to try out a few new linux distros. As someone who was away from linux for a while, the degree of UI customization continues to be both amazing and a little overwhelming, but it feels more polished than before. Taking a look at Niri and hyperland, it's hard to feel satisfied with the UI of MacOS.

Heads up - there are a couple of screenshots in this repo, but no code. This is not an open source project. Cool idea, though.

Thanks for sharing, and am always interested in more Nigerian culture since reading Rosewater.

The analogy I have used is “AI as sous chef.”

The hardware is genuinely interesting

I have heard the same concern about undermining trust from some NPs who completed their training before this more expedited route was available.

Many states now have unsupervised, independent practice for PAs and NPs from the first day they are issued a license. There is variation by state, however, and some still require physician oversight. The amount and quality of that oversight also varies considerably.

I would argue that supervised training with regular feedback on performance is different than job experience, but certainly both are relevant.

Major medical organizations have been advocating for years for more physician residency spots. Unlike NPs or PAs, some residency is required for physicians to be licensed.

Right now, there are not enough residency spots for every US med school graduate.

It's quite a lot more than a year - in primary care, it's more like four additional years of training for physicians, and 15000 supervised clinical hours for physicians (vs 500 to 1500 hours for NPs). The gap can be wider in other physician specialties, because many have longer residencies than the primary care programs. For example, child psychiatry training is four to five years (depending on the route you take), making it longer than the three years of family practice residency.

Here's a chart looking at training for MDs vs NPs in primary care. It is from a physician organization. https://www.tafp.org/media/advocacy/scope-education.pdf

Disclosure: I'm a physician.

One popular approach to saving money is to replace physicians with nurse practitioners and physician assistants, who have less education and training. The article does not discuss this element, and I'd be interested to see if that is a factor in patient outcomes. There's less data on this than you might expect.

ETA: From my post lower down, adding for visibility:

[The training gap is] quite a lot more than a year - in primary care, it's more like four additional years of training for physicians, and 15000 supervised clinical hours for physicians (vs 500 to 1500 hours for NPs). The gap can be wider in other physician specialties, because many have longer residencies than the primary care programs. For example, child psychiatry training is four to five years (depending on the route you take), making it longer than the three years of family practice residency.

Here's a chart looking at training for MDs vs NPs in primary care. It is from a physician organization. https://www.tafp.org/media/advocacy/scope-education.pdf

Thanks. I’m between Omarchy and MacOS mostly, but sometimes Windows, too. Seems worth putting some time into it.

Now that 20225 has become “The Year of the Terminal”(R), I have been looking at how managing my dot files in a more coherent way. I was thinking straight up git vs stow, but I should would be very interested in comparisons from those who have used this tool.

An important issue. how does this compare with llm-guard, and the ability to create a “vault” to later de-anonymize?

I’m a psychiatrist, so if you consider that a significant bias I’m disclosing it.

While there has been a level of diagnostic expansion that I don’t think is helpful, it’s also important to consider:

What’s the psychiatric equivalent of a sprained ankle?

Does something have to be catastrophic to warrant a diagnosis?

If you use Docling, you can set your OCR engine to OCRMac then set it to use LiveText. It’s a good arrangement. You can send these as command-line arguments, but I generally configure it from the Python API.

I have been using VibeTunnel from the iPad. Probably not good for a phone, but on a slightly larger screen it's great. The tailscale integration makes it super easy.

Sure, it works pretty well. Work on one part at a time, then generate clear documentation for each part. Keep the api simple. Probably all of this is a good idea with our without LLMs.

Nice! I like Docling or minerU for local pdf conversion - might be worth a look.

Holy moly. I did not see that coming, but it makes sense. I’m enjoying the terminal-based coding agents way more than I ever would have expected. I can keep one spinning in the background while I do #dayjob, and as a bonus I feel like a haX0r.

2025 is the year of the terminal, apparently?

For my prototype purposes, it’s great, and Claude code the most fun I’ve had with tech in a jillion years.