Blog

I’ve written a few times about borrowing from manual/industrial jobs to inform knowledge work. But there are also key differences.

Yesterday, I listened to a podcast where the hosts discussed specialization.

How much detail do you need to adequately describe something to someone?

Is AI is going to take your job? It depends on what AI is good at.

Some of the most important pieces of feedback you receive as an academic are lagging indicators.

Technological development is great for increasing efficiency. There are pros and cons to this.

GPT-2 was released in 2019. So at most, the best expert in LLMs has 5 years experience. Less if you think the LLM age didn’t start in earnest until GPT-3, 3.5, or 4. And each iteration brings changes in best practices for how to extract information from them.

Google won’t release the AlphaFold3 code.

An experiment with GPT-4o, in screenshots.

Google is rolling out GenAI responses to search results. That got me thinking.