Jeremy Lin Retires After 15 Years That Included ‘Linsanity’ With the Knicks (NYT):

The journeyman played for eight N.B.A. teams and won one championship. But he is best known for a brief stretch on the Knicks where he electrified fans and the nation.

How to feel old #3892: Linsanity for 13 years ago and I remember it like it was yesterday!

How to resist everyday temptations: A guide we can all use, but probably won't.

You're Overthinking Packing:

The amount of planning and thought that needs to go into the enterprise is surprisingly minimal. Count how many days you’re going for, then bring the same number of shirts as the number of days minus one (unless you have access to laundry, then probably less), 2-4 pairs of pants, a couple nicer dresses if that’s your thing and a change of shoes. Something to sleep in at night and a bathing suit if you’re headed somewhere warm. I always bring an extra pair of underwear and socks because sometimes I like to change throughout the day. Then experiment by trying on a bunch items to ensure everything goes together. Frankly, it’s not that different than figuring out what you’re going to wear on a day-to-day basis, which I trust you do all the time.

Some thinking about packing, maybe some tools (I like using packing cubes and a travel scale is a must when flying budget) are surprisingly useful but yes, the curve drops sharply.

My journey from ADHD skeptic to Adderall enthusiast:

At 22, I thought ADHD was fake. An excuse for underachieving kids to get “accommodations” for procrastinating on their homework. From what vague knowledge I had of stimulants like Adderall, I regarded them with the same scorn as the accommodations.

Seventeen years later, I credit Adderall with enabling me to build a 10x happier, healthier, more virtuous version of myself (to the chagrin of the countless Twitter trolls decrying my “meth addiction” in reply to this recent viral post). Here is my story.

What Is Man, That Thou Art Mindful Of Him?: God and Iblis debate if human intelligence is hitting a wall. Hilarious and insightful.

if you meet the singaporean on the road:

All this effort — fifty years of non-stop toil — turning a fallow wasteland into fertile earth, and where are all the crops we have to show for it? Where are all the local companies that we can point to and be proud of? Where are our Ericssons and Nokias?

LLMs Will be Like Ozempic for Golf:

Rob was fantastic but it’s not as though I can just do a TrackMan excursion all the time. Yet I remained curious about why my swing was what it was. I had decent power, but lacked the ability to square the ball up with anything approaching consistency. Was I destined to always have this problem?

And so I turned to LLMs, feeding the TrackMan stats into GPT. Based on 12 numbers, from one swing, GPT had me clocked. It knew my strengths and weaknesses. It fully understood the specifics of my poor technique. I’m sure Rob could have walked me through as much, but his time is limited. The machine had all the bandwidth in the world to deal with my “over the top” swing, how to fix it, and any other questions I might nag it with.

A day later, my swing was different and self recorded video sent to Google’s Gemini confirmed the change. Swing errors that were decades in the making were corrected in the span of minutes. I’m not saying that I’ve suddenly made a leap from “Struggles to break 100” to “scratch golfer.” I’m just saying that a process that could have been expensive and arduous was instead efficient and relatively cheap. I apply the LLM’s fix, and it tells me whether I’ve actually applied it. The feedback is instant and objective.

LLMs will be like Ozempic for a lot more than golf. Ability to ask unlimited questions without feeling embarrased or paying by the hour is a big deal. Imagination is the only thing limiting us.

GPT-5: The Case of the Missing Agent

It’s hard to say exactly why, even with all this progress, current AI models are still so hopeless at dealing with open-ended real-world situations. GPT-5’s inability to recognize that it was incapable of playing Minesweeper may indicate that its reasoning abilities do not generalize well. Its decision to spend 5 solid hours beating its head against the unimportant side goal of sharing a spreadsheet suggests a lack of training on the importance of setting priorities. The repeated factual errors in Gemini 2.5 Pro’s writeup of its merch store experience (click the link and look for “Editor’s Notes”) suggest an inability to keep track of key information over an extended project. Claude losing track of the fact that it is not a person is a reminder that in some ways these models really are just shallow imitations of human behavior (even as they demonstrate deep capability in other areas).

So many benefits and so many limitations.

Reading with AI:

Reading a non-fiction book from cover-to-cover is not efficient. I used to say that I read books “from the outside in.” I look at the book flap to find out about the author, who wrote the blurbs, and the subject matter of the book. Then I read the introduction and conclusion in order to get the main ideas. If I have read something by a different author that seems relevant, I look for that author in the index, and I head to those pages.

and

Once again, I believe in “Stop, Look, and Listen.” I start by asking the AI to summarize the key themes of the book. For each theme that the AI lists, I stop and try to put it into my own words. I test my understanding by feeding my words into the AI, in order get confirmation that my interpretation is correct. Another way that I ensure understanding is to suggest possible examples or ask the AI to provide examples.

How to think about AI progress: Reproducing the entire post as it really is worth reading:

The Zvi has a good survey post on what is going on with the actual evidence. I have a more general point to make, which I am drawing from my background in Austrian capital theory.

There are easy projects, and there are hard projects. You might also say short-term vs. long-term investments.

The easier, shorter-term projects get done first. For instance, the best LLMs now have near-perfect answers for a wide range of queries. Those answers will not be getting much better, though they may be integrated into different services in higher productivity ways.

Those improvements will yield an ongoing stream of benefits, but you will not see much incremental progress in the underlying models themselves. Ten years from now, the word “strawberry” still will have three r’s, and the LLMs still will tell us that. There are other questions, such as “what is the meaning of life?” where the AI answers also will not get much better. I do not mean that statement as AI pessimism, rather the answers can only get so good because the question is not ideally specified in the first place.

Then there are the very difficult concrete problems, such as in the biosciences or with math olympiad problems, and so on. Progress in these areas seems quite steady and I would call it impressive. But it will take quite a few years before that progress is turned into improvements in daily life. Again, that does not have to be AI pessimism. Just look at how we run our clinical trials, or how long the FDA approval process takes for new drugs, or how many people are reluctant to accept beneficial vaccines. I predict that AI will not speed up those processes nearly as much as it ideally might.

So the AI world before us is rather rapidly being bifurcated into two sectors:

a) progress already is extreme, and is hard to improve upon, and

b) progress is ongoing, but will take a long time to be visible to actual users and consumers

And so people will complain that AI progress is failing us, but mostly they will be wrong. They will be the victim of cognitive error and biases. The reality is that progress is continuing apace, but it swallows up and renders ordinary some of its more visible successes. What is left behind for future progress can be pretty slow.

Yet another periodic reminder that MR and The Zvi are both must-read for everyone.