Kromsten retweeted
*8th time. Your data in incomplete
xitter.cf/larry0x/status/2037174…
Kromsten retweeted
check out all the hackathon submissions
what is your favorite?
madscientists.io/hackathon/s…
Kromsten retweeted
Anthropic won't pause AI development.
This is a blatant lie unless they fire their whole army of AI engineers.
They'll keep working and give strong AI to huge corporations and government while arguing that it's too powerful for plebs to use.
Dangerous effective altruism.
Kromsten retweeted
You will never guess from this chart alone the year in which Barcelona stopped issuing new Airbnb and tourist-use licenses.
(answer in the next tweet)
Kromsten retweeted
layer 7: why specific curves are chosen
not all elliptic curves have hard discrete log problems. curve selection is security engineering
secp256k1 (Bitcoin, Ethereum transactions):
→ y² = x³ + 7 over a 256-bit prime field
→ Koblitz curve (a = 0) for efficiency
→ chosen by Certicom, adopted by Bitcoin early
→ ~128-bit security against classical attacks
BN254 (Ethereum ZK precompiles):
→ Barreto-Naehrig curve
→ pairing-friendly (embedding degree 12)
→ used for Groth16 verification on Ethereum
→ originally ~128-bit security, now estimated at ~100-110 bits due to improved NFS attacks on the pairing target group F_p^12
→ still used because Ethereum precompiles support it and migration is costly
BLS12-381 (Ethereum consensus, Zcash):
→ Barreto-Lynn-Scott curve
→ pairing-friendly (embedding degree 12)
→ designed AFTER the NFS improvements that weakened BN254
→ 381-bit base field chosen specifically to maintain 128-bit security even with improved attacks
→ the "12" is the embedding degree, "381" is the field size
ed25519 (many non-Ethereum systems):
→ Edwards curve over a 255-bit field
→ designed by Bernstein for maximum implementation safety
→ constant-time arithmetic, no exceptional cases
→ not pairing-friendly (no pairings possible, which means no Groth16)
Pasta curves (Pallas/Vesta, used in Halo 2 and Mina): → designed as a 2-cycle: Pallas's scalar field = Vesta's base field and vice versa
→ enables efficient recursive proofs without pairings
→ not pairing-friendly (by design, since the goal is pairing-free recursion)
BabyJubJub (used in Circom circuits for in-circuit signature verification):
→ twisted Edwards curve defined over BN254's scalar field
→ designed so that curve arithmetic is NATIVE inside BN254 circuits
→ enables efficient EdDSA signature verification in ZK circuits (~6K constraints vs ~500K for secp256k1 ECDSA)
each curve is a specific point in the tradeoff space:
→ security level (determined by curve parameters and the best known attacks)
→ pairing support (determines which proof systems are possible)
→ field compatibility (determines how expensive in-circuit operations are)
→ implementation safety (determines how easy it is to write bug-free code)
Kromsten retweeted
A technical interview is like homework: nobody uses it, everybody hates it, and it is nothing like the actual job!
How about giving engineers an issue in an open-source repo?
If @FFmpeg merges your PR, you are hired!
(inspired by watching the @lexfridman interview; but any open-source project can be helped this way!)
Everybody wins! Open-source wins!
Employer wins (no more time spent on stupid interviews)!
Engineer wins (open-source contribution, good karma from actually building something people use)!
You get new problems in a real project. You test collaboration.
Why is this not done????
Kromsten retweeted
My dear front-end developers (and anyone who’s interested in the future of interfaces):
I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept):
Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow
Kromsten retweeted
It’s incredible we were able to build a fully functional version, a testament to Timewave’s engineering excellence.
Everything has been open sourced under Apache 2.0, which can be found on our GitHub if anyone is interested: github.com/timewave-computer
TIL that parameter existed
If anyone needs a random one for Meetup that is very helpful
There used to be an option to sort events by date. They removed it from the interface but functionality is still there.
Simply add "&sortField=DATETIME" to the url of the search page
Google just made a subtle but massive change
Last month, Google quietly removed the num=100 search parameter.
This means you can no longer view 100 results at once. The default max is now 10.
Why does this matter?
- Most LLMs (OpenAI, Perplexity, etc.) rely (directly or indirectly) on Google’s indexed results, alongside their own crawlers.
- Overnight, their access to the “long tail” of the internet was cut by 90%.
The fallout:
- According to Search Engine Land, 88% of sites saw a drop in impressions.
- Reddit, which often ranks in positions 11–100, saw its LLM citations plummet. Its stock dropped 15%.
For startups, this is brutal. Visibility just got harder. Reddit as part of AEO just changed entirely.
It’s no longer enough to build a great product you need to crack distribution first. Because if people can’t discover you, they’ll never get to evaluate you.
Most engineers seem to always neglect this reality, but a mediocre product with great distribution will always beat a great product with mediocre distribution.
As Peter Thiel says:
“Most businesses get zero distribution channels to work: poor sales rather than bad product is the most common cause of failure. If you can get just one distribution channel to work, you have a great business. If you try for several but don’t nail one, you’re finished.
Superior sales and distribution by itself can create a monopoly, even with no product differentiation. The converse is not true. No matter how strong your product — even if it easily fits into already established habits and anybody who tries it likes it immediately — you must still support it with a strong distribution plan."
Distribution > Product
(h/t Adarsh Appaiah on LinkedIn)
