veunes

It can certainly make refactoring cheaper, but if nobody understands why some weird workaround exists, the model doesn't magically know either

And just like old modems, sometimes the handshake takes forever, makes everyone uncomfortable, and still fails to establish a useful connection.

Being trustworthy and being perceived as trustworthy aren't the same thing. I think a lot of us who dislike small talk forget that the other person hasn't necessarily had a chance to establish the former yet.

In theory, yes. In practice I suspect a lot of these fixtures are not using integration to achieve better thermal design, they're using it because it makes the product cheaper and lets the manufacturer control the whole assembly.

I think the annoying part is that LEDs themselves can last a very long time, but the cheap driver electronics often don't

At that point you're not wardriving, you're doing a neighborhood-wide ceiling fan penetration test

16 possible fan identities for an entire neighborhood. What could go wrong?

There's something very 2020s about taking two durable, repairable household objects (a ceiling fan and a light fixture) combining them into one visually complicated object, and then making the LED non-replaceable

A lot of EV design seems to assume every familiar control is a problem that needs to be reinvented

I think there's a surprisingly large market for an EV that's basically "a Corolla, but electric" rather than another attempt to make the interior look like a concept car from 2035

Toyota still sells 10M cars a year and has an enormous manufacturing, supplier and dealer network. That doesn’t disappear overnight because BYD can sell a $14k EV in one market

This feels like Toyota optimizing for Toyota rather than for the EV buyer

The interesting part isn’t really that Toyota is making an electric Corolla, it’s that they’re apparently trying to keep one platform flexible enough for ICE, hybrid, PHEV and BEV

I think "final" here just means the unit that currently gets treated as publishable and credit-worthy: a theorem with a proof, a solved conjecture, a completed paper, etc

This seems right as long as the AI output is legible enough to reverse-engineer

Seems perfectly possible to get the worst of both worlds: more mathematics than anyone can read, and less access to the mathematics people actually care about.

Data minimization becomes a lot less abstract once every unnecessary record on disk has an actual dollar value attached to the risk

Yeah, the irony is that every extra signal added to make verification "safer" also becomes another extremely valuable thing to steal when the verifier gets breached

Regulatory retention is a valid reason for some of this data to exist. It isn't a blanket justification for every intermediary in the verification chain to retain its own permanent copy. If anything, that makes minimizing the number of copies even more important.

Yeah, this is the part I don't get either. Verification should produce a yes/no result, not a permanent archive of everyone's identity documents

Definitely needs the traditional banana for scale, although in this case the banana might occupy half the cafe (just a joke, sorry (just in case))

Wow, these are excellent. I clicked for the miniature craft and ended up staying for the tiny mouse drama

The tiny newspaper and individually assembled tulips got me. There is a strange amount of joy in details that almost nobody will ever notice unless they stop and look closely.

That is fascinating. The parakeets are basically providing an early warning system, and the dogs have figured out enough of their "language" to subscribe to it.

What gets me is that it is not just "crows remember faces". Apparently the information can outlive the original encounter entirely. You can end up on the neighborhood crow blacklist without some of the crows ever having met you before.

That would explain why they sometimes seem perfectly relaxed until you look directly at them

The iPad reflection trick is hilarious. You are basically observing him observing you while carefully pretending not to observe him.

AI definitely lowers the cost of building these, but I think Ciechanowski's real contribution is upstream of the code. Knowing what to visualize and how to make interaction teach something rather than just look impressive is the rare part

This is probably the healthiest way to use stylistic imitation: treat it as scaffolding, then figure out why it works and throw the scaffolding away

I find these stories way more interesting than polished "history of computing" summaries