Dan Luu

5.6K posts

Dan Luu

Dan Luu

@danluu

Active on https://t.co/WG71Nrs60M; also trying out https://t.co/fGOzbSxVHi. No longer read replies or notifications here now that tweetdeck is gated.

Katılım Aralık 2008
43 Takip Edilen45.1K Takipçiler
Dan Luu
Dan Luu@danluu·
Looks like I spoke too soon about the AI not being superhuman. The current Azul world champion played against it and thinks it's better than him at higher difficulties and a top 100 player played against it at default difficulty and thinks it's better than him at default.
English
1
0
12
1.7K
Dan Luu
Dan Luu@danluu·
Not to overstate — I saw someone vibe coded the same project then declared programming dead after they finished, but their bot loses to MCTS + simple heuristic. Mine was in the same state when I just had an LLM in a loop with instructions to improve the result.
English
3
0
19
5K
Dan Luu
Dan Luu@danluu·
It's sort of amazing how quickly you can do things now. I wanted to try writing an alphazero-style AI for Azul. With no AI background, it took me maybe 2-3 hours to (2-3 days wall clock) to beat the best public AI I could find, training on my laptop CPU: danluu.com/game/tile/
Dan Luu tweet media
English
3
4
96
6.6K
Dan Luu
Dan Luu@danluu·
I've heard you can email jeff@ to escalate further, so I'll do that if the scam continues, but it's interesting that escalated support has no power to either do something about a serial scammer or stop what looks like a scam in progress even when they can see it's happening.
English
0
0
7
2.6K
Dan Luu
Dan Luu@danluu·
Just had a really interesting experience with escalated Amazon support, where I talked to them about a seller who appears to be a serial scammer based on seller feedback. The support person agreed but said there's nothing they can do about it, their hands are tied.
English
2
1
8
2.8K
Dan Luu
Dan Luu@danluu·
BTW, the above is a huge talk anti-pattern. In the median 10-minute talk I've seen, the speaker introduces themselves for 30-60 seconds and it's not uncommon to see the intro go to 2 minutes. Why are you spending 10% of your talk introducing yourself? Get to the content!
English
3
2
42
6.9K
Dan Luu
Dan Luu@danluu·
@josephscott You can also do comparisons like React vs. Vue vs. jQuery In terms of Core Web Vitals (which Google says is used as a ranking signal) we have jQuery > React > Vue. Per above, this is highly confounded with what users do, but I still find these silly comparisons fun
Dan Luu tweet media
English
0
1
16
3.5K
Dan Luu
Dan Luu@danluu·
You can also see CWV, INP, etc. Those are confounded by what users do, so they're less about the host, but per danluu.com/slow-device/, there's value in looking at the default experience as well as what users actually do due to what's made easy or hard. h/t @josephscott
English
2
1
6
3.8K
Dan Luu
Dan Luu@danluu·
TIL archive.org provides a performance dashboard with slice and dice capability E.g., you can sort big WordPress hosts by TTFB. Below are the fraction of sites that are "good" (I'd prefer to see the distribution, but this is easily available) lookerstudio.google.com/s/ha530x6JKeE
Dan Luu tweet media
English
4
11
46
10K
Dan Luu retweetledi
apenwarr
apenwarr@apenwarr·
@danluu I think this part in particular is really underrated. Once you can go really fast, things become possible that were impossible to contemplate before. Tailscale wouldn’t exist if we had done things the regular way. It’s supposed to be unthinkable.
apenwarr tweet media
English
1
5
49
0
Dan Luu
Dan Luu@danluu·
A version of Missile Command for the Commodore 64 where the bottom of your screen is the game state in memory and missiles cause memory corruption: csdb.dk/release/?id=13…. In the video below, a missile broke my controls and caused my cursor to get stuck moving down and left.
English
6
32
190
20.5K