The pace of AI triggers my, "Chill out and wait 'til this all settles down," instinct, which is probably maladaptive in a world where nothing settles down anymore.

I can't believe I hadn't heard of this before. share.google/eIqSbkLLNjI8...

I've heard people swear by LLMs for blood test results but had never tried it. Usually I don't look at my results because it makes me anxious, but Claude did great. It looked at the couple flagged results and was like "let me see <some other details>" and then gave it to me straight.

what

Now that my wife is back from SF I can finally get some sleep. Our schedules are shifted enough that each of us can sleep uninterrupted without worrying about the dog.

this is an attempt by the Secretary of War to bankrupt Anthropic. good luck to everyone working there

I try to see things from other people’s perspective but for so many these days it goes far beyond just imagining yourself in their shoes to having to imagine yourself having a literal potato for a brain.

this is about the government using AI for surveillance a somewhat rushed post about Anthropic telling the Secretary of War no, Anthropic's history with the government, etc.

After mishearing the lyrics to Lana Del Rey's song Ride for so long I propose we rename the Sun to Photon Daddy.

My weird sleep schedule this week has proved interesting for learning certain things because I'm getting extended time in that zone where you're too tired to overthink but not so exhausted you can't think at all. It's a good zone for learning the "it is what it is" type stuff.

Another way LLMs are helping me with learning is in categorizing and organizing things to learn. As someone who naturally wants to know why, I always struggled in school, particular math, with knowing when something was a deeper concept to dig into vs just something to memorize for the toolbox.

If you've spent your life in a position where confident, fluent, authoritative-sounding people have been wrong about you, about your identity, your experience, your reality, you develop exactly the kind of vigilance the essay describes.

It just blows my mind that energy has inertia and the implications of that. That solar sails work kind of makes it intuitive that momentum is there, but that’s a very small cumulative buildup. That charged batteries weigh more kind of gives away the game, but still, it’s very small.

excuse me

I really underestimated the current models ability to work with the often crappy output of voice dictation. I've been missing learning opportunities between the multimodal voice models being synchophantic and outdated or not being in a position to type out a bunch of stuff to a good one.

so whatever your patterns and styles of learning, an LLM can be a potent source of scaffolding and support for the curious learner, the LLM is a science fiction technology capable of developing your thinking, spotting gaps in your knowledge, and lighting up a hidden curriculum (unknown unknowns)

A good example of how LLMs have helped me learn by connecting topics and filling in the gaps where I have deficits, which is usually math. Here we were discussing GR when the concept of four-velocity always having magnitude c clicked, and it connected it to ML for me.

Are there other movies about the sun dying besides Sunshine and Project Hail Mary?

Claude out here working on deleting so much more than your email.

This probably isn’t going to help the trend of sounding like I’m going insane, but I swear last night I noticed my pillow producing pulses of light.

Just now getting around to starting season 2 of Fallout. I love that Hank didn’t even get out of the power armor before pouring a cup of coffee and the nod to Twin Peaks there.

Something I appreciate about the recent years with LLMs and learning is not just what I get from using them, but also what I’ve gotten from learning how best to use them. Reminds me of the Mortimer J. Adler book How to Read a Book.

This is what I don’t understand about Google. There is absolutely no reason for everything to constantly feel like some hackathon project duct taped together.

I think of this scene whenever I mention LLMs and physics. No, it's not that.

I had no idea the guy (Minkowski) who came up with the concept of a 4D space-time continuum back in 1908 died just months later from appendicitis at age 43.

Catching up on years of physics to procrastinate from dealing with the duvet cover.

Out of Context Claude: Your entire remaining existence has a finite proper time — and it's short.

I love how Claude put this: "Sagan would have had a field day with this. The demon-haunted world now includes demons that autocomplete."

Finally had General Relativity click for me today after some chats with Opus and it explained things like world lines and geodesics. All these years I had had a mistaken timelapse type view of space-time but now I get it in the geometric sense and suddenly so much makes sense.

I feel like I’m losing my mind every time I read a reddit post where someone used AI for something it obviously can’t do, complain about it, and then all the comments are just about how it also can’t do anything else. Like what in the world are these people doing?

When Opus sends you out for books on gravity.

DMB's When the World Ends Oakenfold remix would be interesting.

Holy shit. It's so fast. chatjimmy.ai

If a slider is just a hamburger to a smaller stature being, and their species also has a concept of a slider, then given the size of the universe we can assume there are a series of much bigger and smaller sandwiches than we can imagine.

There's some creepy ass Alan Wake style fog going on right now.

Why do I keep seeing people comparing AI coding to gambling? I can't remember the last time I had to revert something and try again.

That wolfdog at the olympics looked like it ran straight out of a dog food commercial.

Every few months, I write an updated, idiosyncratic guide on which AIs to use right now. My new version has the most changes ever, since AI is no longer just about chatbots. To use AI you need to understand how to think about models, apps, and harnesses. open.substack.com/pub/oneusefu...

Keep seeing people respond to that article about AI denialism and missing that the point that is basically Abundance. The claim that the left abdicated on a vision of the future isn't the same as being anti science.

It's so weird to me how the anti-AI type treat it like any other product like datacenters are being built just for slop and a bubble that is going to pop anytime, while the buildouts are openly being compared to the Manhattan Project and the DoD is beefing with Anthropic.

It’s weird we never do stuff like this with other terrible decisions. Newspaper doesn’t run “I saw a half empty can of Mountain Dew left on the street and drank the rest of that mfer”, and the guy who drank the Mountain Dew is only hurting himself

Absolutely brilliant piece about the Left's TOTAL blindness on AI. Their dismissal of AI risks mirrors how climate deniers treat CO2. Will probably get a lot of nastiness for this on Bluesky, but I guess that's part of the same problem. www.transformernews.ai/p/the-left-i...

This is how they used to get the downhill skiing shots before drones and it only cost them $2.

I was just testing out an MCP for my project management app in Codex and asked it to check the git history to see if we had already fixed a couple things before realizing I had started it in the wrong directory, it was like "no worries", found the right directory, and just took care of it.

americans cannot fathom a meaningful social safety net and are therefore convinced that efficiency improvements are evil, and so they hate them whenever they are visible

So glad to see the instagram tunnel lady is back in business.

When I hear token anxiety I think of having an 8K context window with a 1K prompt and then right when you're about to get it working it's like...

This is exactly it: LLM coding only really works when coupled with tests (and ideally type systems) to evaluate their code against. The magic is in the agentic loop, in the iterative convergence towards correctness, not the first oneshot output.

Something feels weird about all the drama today. I don't get why it is so surprising that some young people are fervently trying to keep up and reskill during an industry-wide paradigm shift? That's obviously not a new concept.

One of those days where I just dump the bag of marshmallows straight into the Tillamook.