nwhitehead

"the smart thing to do is to give integers zeros as their tag bits, because then, adding or subtracting two shifted integers remains a plain add or sub machine instruction"

this is brilliant, love it. stealing this idea immediately.

(Spouse's story)

Today I used Claude to diagnose a blocking bug in a Steam game I really wanted to play. It took it 18 mins, but it unpacked the Godot package, figured out the bug, proposed a fix, and gave me an in game workaround.

I didn't have to do anything! Claude figured out the structure of the .pck file by using `strings`, then wrote some Python code with some magic Godot-specific code to unpack the specific chunks it needed.

This is fun, I'm curious to try it.

An alternative that I experimented with and found to be very usable is one solar panel, a small camping battery ("portable power station"), and an Instant Pot. The total cost is not super high. The Instant Pot is power efficient and can cook a lot of food at once. Since it's battery powered you can start any time the battery is charged.

You would think so...

I asked this question in a college-level class with clickers. For the initial question I told them, "This is a trick question, your first answer might not be right". Still less than 10% of students got the right answer.

One idea is to put more pressure on writers of letters of recommendation. There is obviously already some rather weak "reputational incentive" to not give letters of recommendation to fraudsters. But it could be made a more formal policy. Letters could be required to include a statement about the candidate adhering to various formalized "good behavior" policies. Then when fraud happens there could be actual consequences for the letter writer. This would be like the culture in software development where you don't blame people for writing bad code, you blame reviewers for letting in bad code. (Idea comes from Angela Collier, in relation to other forms of antisocial behavior).

I like how you can handle lots of languages in a very systematic way, that is cool.

One thing I like about pyodide is that it supports lots of packages like numpy. How would you support packages like this with a WASI approach?

I'm using the new generation of small models to do semantic search for music lyrics. The first step is giving the model the text and asking, "What is this text about?" Without fine-tuning and only minimal prompt engineering these models can understand most languages, pull out the most relevant phrases, and list major themes and speculative ideas of what the lyrics mean. I'm super impressed with the results. Reading the answers feels like grading ambitious undergraduate student essays (both the good and the bad).

This is a good one. It is all too easy to spend years casually "interested" in a topic without developing a true foundation of the important ideas in the field.

In my case I spent two decades casually reading about economics. But then one day I ordered an economics textbook to work through the problems, and found out that I basically knew nothing and couldn't do any of the problems. A few months of slowly working through the textbook gave me a much deeper level of understanding. I could do the problems now.