tehnub

Such excellent writing. I also read the linked post from 2018 near the bottom. When every other thing you see on here is Claude-inflected if not Claude-written, encountering a blog like this is an occasion to cherish.

Indeed. See https://en.wikipedia.org/wiki/Common_knowledge_(logic)

Reminds me of the extended description of what it might be like to drown in an ice lake in the book Stella Maris — it wouldn't be quick.

Opus 4.7 did this in my code and I had never seen it before

It's a common rhetorical technique that never convinces me. I'm not convinced this guy's anecdote is an exception to some rule.

If the models are good enough, you will simply get outcompeted by engineers willing to trade their long-term cognitive ability for a short-term lucrative career

(2) AI-users thus become less effective engineers over time, as their technical skills atrophy

Wouldn't (2) imply that if everyone just used AI there eventually would come a time when there aren't engineers who will outcompete you (because their skills are so atrophied)?

Interesting, enjoyable post. Like the bit about data modeling. I've been dabbling in some OCaml and that kind of modeling is the best part. Also interesting to learn of CAMLBOY. Feedback to the author: Skip the AI edit step. I'd have preferred grammar errors or inelegance to what we have here, which is a bit stale.

I just read this https://en.wikipedia.org/wiki/Battle_of_Cajamarca

It seems like the Incans were overconfident and didn't expect a surprise attack (didn't have their weapons, only a small retinue around the rule in ceremonial garb instead of armor), and then the 8000 warriors were outside and didn't even attempt to fight the Spaniards because they were so demoralized.

Brilliant

Sure, if you mess up and one of the houses collapses then that'll manifest as an outcome that impacts the home owner negatively.

lol

There are plenty of furniture makers that use all hardwood construction. It's just that they charge like $10K USD for a cabinet

This incident was from 2014. I wonder how many OKCupid employees and shareholders from then are still at/invested in the company. What do corporate punishments do if the people who made the mistake aren't even there to receive them?

I've been using jj, which apparently is also faster than jq https://github.com/tidwall/jj

There's probably tons of stuff people use that is way slower than it needs to be but speed isn't one of their self-reported burning problems because they don't even realize it could be faster. pip vs uv?

Nah we're not doing prequel hate in 2026

Why Article By AI?

I wouldn’t because idiomatic Haskell is way slower than idiomatic Rust.

They’re using big things to do experiments. Maybe they discover some new physical effect. How do you know that that effect couldn’t be demonstrated in some smaller scale experiment after it’s understood better?

I don't think that argument holds up. See quantum mechanics.

How can people still make such mistaken analogies when ChatGPT was released over three years ago?

Totally on point, except I'm pretty sure Jobs was not like that. From what I've read he'd be more of a hands on "agentic engineer". Baby-sitting his engineers and designers and steering them.

I think Asimov is right that 1984 was not intended as a forecast but rather a depiction of Stalinism with British characteristics, so to speak.

In short, if 1984 must be considered science fiction, then it is very bad science fiction.

Based on this quote and others, it seems Asimov didn't believe that Orwell intended the novel as science fiction, although others categorize it that way. I would say he's attacking the interpretation of it as science fiction, but it veers into an attack on Orwell, which is unfortunate.

You write

Science fiction does not _forecast_.

Not to be overly pedantic but to be fair to Asimov, he didn't exactly say science fiction _necessarily_ does that, but rather it's a knack related to science fiction.

I worship both thank you very much.

This is a nice guide. I especially like the masked decoding diagrams on this page https://nanonets.com/cookbooks/structured-llm-outputs/basic-....

edit: Somehow that link doesn't work... It's the diagram on the "constrained method" page

I guess this applies to the type of developer who needs years, not weeks, to become proficient in say Python?

What's most useful to you is not necessarily most useful to the business. The bar for critical thinking to get staff at this company I've surely heard of must not be very high.

People did?

I predict by the end of next year we will have our AIs write TPS reports.

Beads?

It's obviously tongue in cheek

I've seen him a few times in the last few years. At each show he played like two or three covers, and the rest were his songs.