lekevicius

tbf they turned previous ChatGPT macOS app into Codex, so getting a "new" Codex user might mean someone using ChatGPT just updated their app.

Amazing, exactly what Workers lacked, I was quite annoyed that a worker would spin up when 99% of my requests then return a cache. Waste of 3ms, times millions.

I'm not a native speaker (Eastern Europe), and my scores are 20, 20, 17, 18, 15 - more aligned to your native speaker model.

Known generative-AI crawlers are disallowed in robots.txt. This is a research catalogue assembled from primary sources; it is not training data, and a model fine-tuned on these paragraphs would launder out exactly the part — the citations — that gives the prose its value.

This reads like distaste for LLMs - but generally website reads (and is designed as!) very LLMy.

Feels like CoffeeScript for C, in the best way

I always knew that egg came first.

Do you think it would be possible to achieve 1:1 ELO:bytes? Even smaller, but can be less smart.

https://lekevicius.com

Is there enough of new blood on HN? For me it was the best place, my favorite website, when I was entering startup scene. Loved it. I don't think a lot of young founders I know ever go here...

Even if that's true, that sub-minute videos are not the apex content, that only goes to prove inevitability. Every idea will be tested and measured; the best-performing ones will survive. There can't be any coordination or consensus like "we shouldn't have that" - the only signal is, "is this still the most performant medium + algorithm mix?"

I do disagree that some of these were not inevitable. Let me deconstruct a couple:

Tiktok is not inevitable.

TikTok the app and company, not inevitable. Short form video as the medium, and algorithm that samples entire catalog (vs just followers) were inevitable. Short form video follows gradual escalation of most engaging content formats, with legacy stretching from short-form-text in Twitter, short-form-photo in Instagram and Snapchat. Global content discovery is a natural next experiment after extended follow graph.

NFTs were not inevitable.

Perhaps Bitcoin as proof-of-work productization was not inevitable (for a while), but once we got there, a lot of things were very much inevitable. Explosion of alternatives like with Litecoin, explosion of expressive features, reaching Turing-completeness with Ethereum, "tokens" once we got to Turing-completeness, and then "unique tokens" aka NFTs (but also colored coins in Bitcoin parlance before that). The cultural influence was less inevitable, massive scam and hype was also not inevitable... but to be fair, likely.

I could deconstruct more, but the broader point is: coordination is hard. All these can be done by anyone: anyone could have invented Ethereum-like system; anyone could have built a non-fungible standard over that. Inevitability comes from the lack of coordination: when anyone can push whatever future they want, a LOT of things become inevitable.

They can appeal.

It’s of course possible, but the spec is trying to avoid this. There are multiple “masking patterns”, and the algorithm should choose one that gives fewest “penalty points”. Large single color areas are a lot of penalty points, so QR algorithm is trying to avoid them.

I’m curious! What are its trade-offs?

You are right, but a better place to follow progress is https://ethtps.info/. It shows TPS across all layer-2s.

Current network demand doesn’t really require them, but after EIP-4844, “100 000 TPS” should be possible.

That “good summary” video is anything but, and contains numerous incorrect claims, starting with not being aware of differences between tokens and chains. He claims that Binance hasn’t provided any “Token addresses”, when they provided their ETH addresses (that contain tokens).

HN already has a “reply” link below the comment. Adding “upvote” would solve both the timing (noticing upvote after reading the comment) and usability (solves touch target size on mobile).

In addition to all the recommendations about supporting his Patreon, I also recommend following him on Twitter and clicking the bell icon: he only tweets when new articles are published, a very worthwhile notification.

I host a nationwide radio show that we also publish as a podcast. Podcasting is not radio, as much as I'd enjoy that. We are the only show I'm aware of that publishes "show notes", links to things we talk about on the radio, encouraging listeners to check those things out. We are also deeply tied to the schedule: some of my favorite podcasts vary greatly in length, our show is exactly 25 minutes every time, with a reminder "what show you are listening to" around 12 minutes.

Radio is still great at having massive, immediate, but highly passive audience. In every other regard, podcasts are so much better.

I found one particular example of the opposite change quite annoying. In the TV Show “Suits”, the premise is that a character Mike has incredible photographic memory, can to read books and evidence at unbelievable speeds. As the show went on, this unique trait was almost completely removed. I think by season 3 it was just gone completely, turning the show into a regular law drama.

New API, and already mixes camelCase (committeeReport) and snake-case (house-communication).

It might be this sentiment -- that emails should be like letters, with a greeting, ending, signature, etc -- that pushes all communication to more "SMS-like" media. We even get support from companies over DMs.

I personally use email like SMS. I might say "Hi" on the first email, and might end with my name if we never chatted before, but otherwise, straight to the point.

Shinzo Abe marked as "probably alive", so clearly some data is out of sync with WikiData.

I've always felt they did Xbox / PlayStation strategy, selling headset below cost and earning on game store fees.

Perhaps the math is not working out?

I think name servers is one of the best applications for a decentralized ledger. It _can_ work without a central party, and I think it might be better without one. Something like .org controversy might not have happened without a central party.

What a strange implementation of "format on too many missed passcodes". Apple (on iOS and watchOS) implements this, but after some amount of failures, phone gets into progressively longer lockdowns. So maybe after 3 failed attempts you have to wait 2 minutes, after 4th 5 minutes, and before the final (formatting) attempt you have to wait something like 12 hours. This prevents "kid playing with the phone" problem.

I see this as securing against "unknown unknowns". No software can ever be "100% bug free". If you can identify areas that are more likely to contain yet-undiscovered vulnerabilities and turn them off in advance, the device becomes more secure.

Assuming the node network and mining remains intact... now you own a pretty valuable and difficult to replicate infrastructure. At least some people will want to own the "native token" of that network.

GP - Thank you, you explained the same things I would have pointed out.

We have a lack of curiosity and skepticism, and too much passion. There is quite a bit of echo-chamber effect going on, entrenching current view. Suppressing an emotional topic might help later on. Hopefully in 5-10 years cryptocurrencies will either be boring or dead: either way passion will have died out and we can properly discuss it.