HNHacker News
TopNewBestAskShowJobs

dgroshev

1,778 karma · joined September 30, 2014

submissionscomments
dgroshev··on Claude Opus 5.5
I don't think it's substantially different. I just pasted a random chunk of code and asked Opus 5.5 to comment on it:

> The Vercel target is hard-coded. That's common and not wrong, but it's opaque; nobody reading this later will know which Vercel project it belongs to, and if the project is recreated the target changes silently. A comment or a named variable would help.

> Pointing a DNS name at Vercel is only half the job. The domain also has to be added to the project in Vercel's dashboard, otherwise requests will arrive and Vercel will reject them. That step lives outside this code, so it's easy to forget.

> Finally, [CENSORED] existing only in production is slightly odd on the face of it. It may be perfectly deliberate (perhaps a single shared testing tool that only needs one public address), but if you're reviewing this rather than just reading it, that's worth confirming.

It has the same annoying cadence and writing style with slightly less prominent claudisms.

dgroshev··on AI coding has made CI a bottleneck, so we reworked ours to keep up
I suspect a substantial part of this is an avalanche of useless testing.

If you even review PRs still: when was the last time you didn't just skip over tests? And if you ever looked at tests in an LLM-heavy PR, how many of those tests tested something useful, and not just built-ins and trivial behaviours?

There's at least some awareness in the industry of how LLMs generate a lot of boilerplate in business logic. It feels like we're much less aware of how much of it is in tests.

dgroshev··on Rust is tier-1 language at Microsoft
That's true, although inconvenient for production code that needs to be delivered yesterday.

Sadly, agents don't mind generating gigatons of code instead of refactoring the abstractions.

dgroshev··on Rust is tier-1 language at Microsoft
To add to this, I find that the delta between the amount of code and pain you get with good and bad abstractions is substantially higher in rust than other languages. It's alright to muddle through in Python or TS, but with Rust bad abstractions are punishing.

LLMs are pretty bad at picking abstractions.

dgroshev··on Among European Companies That Use a CDN, Nearly 9 in 10 Use Cloudflare
Ukraine is retaking some of the territory right now. https://euobserver.com/236059/ukrainians-again-liberate-terr...
dgroshev··on Your Racist Linux Distro Is Very Nice (Scott Jennings)
More likely than not a full endorsement, given DHH's multiple references to notoriously and openly racist "The Camp of the Saints", or this quote:

> standing idly by while your livestock is devoured by predators or your neighborhood is taken over by migrants

dgroshev··on Asahi Linux on M3
I don't think "growing polarisation" is the right framing here. It gives way too much weight and legitimacy to views that have to be subjected to cordon sanitaire [1].

It's not acceptable to associate race with national identity (DHH's London post). It's not acceptable to call for ethnic cleansing (his post on Travellers). It's not "just another view, couldn't we all get along"; we can respectfully disagree on small state vs large state, on low taxes vs high welfare, on isolationism vs interventionism even, but DHH's views have to be unacceptable.

What is happening is a number of very rich, powerful, well-connected men (and sometimes women, see AfD) are trying to normalise those views and demolish the cordon sanitaire. "It's just another view" and framing this view as legitimate, just polarised, is one of their tactics. It's a moral duty of anyone remotely believing in the OG liberalism (you know, that people are equal regardless of their skin colour or ethnicity) to not let them.

[1]: https://en.wikipedia.org/wiki/Cordon_sanitaire_(politics)

dgroshev··on Which tools do Claude, Codex and Cursor choose? We measured 17k runs to find out
You're talking to the CEO of the want-to-be-SEO company [1] that did that. It is very much content marketing for their services.

[1]: https://news.ycombinator.com/item?id=49557233

dgroshev··on As AI eats the web, the internet’s collective memory is disappearing
Here are a couple examples: https://takes.jamesomalley.co.uk/p/build-gemini-build
dgroshev··on The UK's war on anonymity has come to America
A lot of clubs in a lot of places scan and store your ID on entrance.
dgroshev··on EU Age Verification Project Mandates Hardware-Bound Attestation
Indeed, Trump is a great example of democratic backsliding completely unfettered by "freedoms" often mistaken for the actual freedom. The way forward is being able to change things or regaining this ability if it's lost, not trying to carve out some bit that's not under the government's control yet. That's just a form of escapism.
dgroshev··on EU Age Verification Project Mandates Hardware-Bound Attestation
You just said it yourself: the perceived danger of overreach was averted through new laws and control over what the government is doing.

If citizens still have a capacity to restrict the government via (enforced) laws, it doesn't matter much that the communications are in the clear. You can layer something on top just to be sure, but the foundation is laws and a functioning political system.

If there's no functioning political system, the government will just do whatever, like buying the same data they ostensibly can't collect from commercial providers (who weren't restricted from collecting it because there's no US GDPR).

If a country has a functioning political system, it can just do things like ban gun ownership, introduce mandatory IDs, publish everyone's yearly income, or criminalise calling other people slurs in public and not become an uncontrollable dictatorship. If it doesn't, corruption and democratic backsliding will happen regardless. Hyper-focusing on a few tools and superficial traits of a particular system in a particular moment is myopic.

dgroshev··on EU Age Verification Project Mandates Hardware-Bound Attestation
So what? Some things are exceptionally hard to track now (like e2ee messages) that were trivial to track back then.

Every remote communication could only be had in the clear. Everyone was fine with that. Was the US in the 80s a surveillance state?

dgroshev··on EU Age Verification Project Mandates Hardware-Bound Attestation
Just a few decades years ago, everyone's names and phones in the US were freely available to anyone with a phone book. Communications were done in the clear, either via letters (hardly tamper-proof), or via phone (that anyone with access could tap).

Did every American live in a "surveillance state" back then?

dgroshev··on EU Age Verification Project Mandates Hardware-Bound Attestation
Algorithmic scroll is factually more harmful, regardless of the amount of harm as long as it's non-zero, because it does affect people. Surveillance on its own doesn't. It just doesn't, everything that you can think of in response is not solely or even significantly caused by surveillance.

Instead, limiting surveillance (which in this case is only linked to the issue at hand with what's explicitly a conspiracy theory, so the link is quite tenuous) was originally meant to limit the government's powers as a form of control over the government. Control over the government is just a part of the Enlightenment project of societies being able to shape their destinies equitably, which includes shaping political arrangements.

Americans have completely lost the ability to shape their political system (when was the last constitutional amendment passed again?), so instead the discourse turned to fetishising particular tools that might help, but were never the point of having freedom. This is a fallacy.

There are straightforward ways to have age verification without descending into a dictatorship. To use an analogy, most European countries have mandatory IDs, some even have digital IDs; many of those counties are freer than the US.

Freedom is being able to stop kids from accessing pornhub without fearing a dictatorship because you can just prevent the dictatorship through a million other means. Freedom is not paralysing the political system as technology and progress (and multi-billion dollar megacorporations) march on.

Otherwise societies just get superficial "freedoms" (like using slurs in public spaces because "free speech") while their governments, say, run a paramilitary murdering political opponents on the street.

dgroshev··on Our position on open-weights models
Did the US annex Vietnam or Iran?

There's a qualitative difference between interference and outright annexation, with borders redrawn and cultures violently assimilated.

dgroshev··on Mysteries of Telegram Data Centers (2022)
> How can one intepret it in any other way is "Russia is evil and visiting Russia is thus evil"?

Simple: Durov has no problem with constructing a narrative that has little to do with reality.

dgroshev··on Mysteries of Telegram Data Centers (2022)
Of course he would say that. Even besides the fallaciousness of this argument (there is no equivalence between an aggressor and a victim), Pavel Durov is completely untrustworthy.

There is product stuff, like misleading claims about Telegram's encryption and comparisons to Signal. In reality, for the vast majority of chats they have plaintext, unlike Signal.

There are more subtle positioning claims. Durov made a huge deal of his "exile", but I saw Telegram's office in St Petersburg with my own two eyes a year after Durov "fled the country". It definitely wasn't shut. There were even local news of Pavel assaulting a guy in St Petersburg for trying to make a photo of him a couple years later.

And then there are just completely unnecessary lifestyle claims. He said multiple times how he doesn't take any "pills" or medications. It only takes a minute to find his old photos. Male pattern baldness doesn't stop progressing without DHT suppression, and last time I checked, finasteride comes in pills. I don't understand why would he make misleading statements about something so visible, but it doesn't make him any more trustworthy.

dgroshev··on Mysteries of Telegram Data Centers (2022)
More mysteries of Telegram Data Centres: https://istories.media/en/stories/2025/06/10/telegram-fsb/

(and a follow-up: https://istories.media/en/news/2025/06/10/telegram-responds-... )

dgroshev··on Why write code in 2026
Some heart surgeons might only be able to explain to the nurses their ideas on how to carry out the surgery, but are they still heart surgeons?
dgroshev··on My favorite keyboards
That list really misses Glove80! It's an incredible keyboard, and imo is better than the Moonlander at practically everything.

Edit: on second thought, I guess some people might not like the low switches?

dgroshev··on Claude Code is steganographically marking requests
China? https://en.wikipedia.org/wiki/Chinese_police_overseas_servic...
dgroshev··on Oxide computer 3D rack guided tour
I don't mean a conversation. I understand the process and how front-loaded it is at Oxide. I mean just having an actual human being on the other side of those rejection emails, instead of sending them from an unmonitored address. Oxide's own RFD 3 says

> Candidates may well respond to a rejection by asking for more specific feedback; to the degree that feedback can be constructive, it should be provided.

…but in practice it's just boilerplate and silence. Good luck asking when no one's listening [1].

Lots of companies do that, too. The problem is that this approach feels even more unfair than when it's a more "regular" hiring flow. Oxide asks for a very high level of effort from their applicants, but you can see in these comments that at the same time they are quite far towards the lower end of how much visible effort they commit back to the applicants (delays, boilerplate, ghosting).

And sure, as you say it's hundreds-to-thousands of applications, and potentially dozens-to-hundreds of emails to reply to. But the additional time it takes to send a one line reply pales in comparison to thoughtfully reading 12+ pages of materials, which they say they do. I just don't think that adding a few percent on top of that massive effort is unrealistic. It's an active decision to save time and money on people who didn't pass the first stage; I think it's an unethical decision, but then I'm not nearly as successful as Bryan Cantrill.

[1]: To be fair, Bryan kindly advised to "DM" him for feedback in a similar thread half a year ago. There are no DMs on here, so I DM'd on bsky and tried to guess his email, but I probably guessed wrong and he doesn't check his bsky DMs.

dgroshev··on Oxide computer 3D rack guided tour
Probably 16 hours all in between research, writing, and editing, spread over a week. That might be a bit more than average, since English is my second language, and I make many passes to make sure the text works.

Got boilerplate-rejected with zero human interaction three months later.

dgroshev··on Crypto in 2026: Oh, This Is the Bad Place
> Backpack launched a SpaceX token at IPO that can be moved between on-chain and your brokerage

What is the actual societal value of this? Do you seriously believe that such a token helps price discovery?

dgroshev··on UK Home Office launches £75M 'PoliceAI' to capitalise on artificial intelligence
Laws are "reviewed" by courts, and in the UK newer laws automatically supersede older laws if there are any contradictions (with minor nuances as per Thoburn v Sunderland City Council). There's no need for the mechanism that you seem to have in mind (never mind that it doesn't work like that), we can and do change laws.
dgroshev··on UK Home Office launches £75M 'PoliceAI' to capitalise on artificial intelligence
Most people on here are in the US, and the baseline of "scary" implied by that (and many other) posts is definitely not where it should be.

"Moron" doesn't nearly do it justice. The US saw a paramilitary force lead by an open racist (he was recently hanging out with open nazis on a "remigration" conference) executing a political opponent right on the street on camera, with zero consequences. A substantial proportion of the US cheered for literal concentration camps. Budget money is pretty openly funnelled to Trump's family and friends. This is not "Trump is a moron", this is a catastrophe and democratic collapse that is not nearly in the same category as "oh but what if they use the technology for surveillance".

I suspect that this obsession with the UK is just a coping mechanism.

dgroshev··on UK Home Office launches £75M 'PoliceAI' to capitalise on artificial intelligence
Thankfully, we do have laws that change with the times over here in the UK. Just a couple months ago we had a constitutional change (abolishment of hereditary peers) and that was just another Thursday. It's fine, we don't need to rely on deliberately inefficient police force.

Substituting political process and laws changing with the times with political nihilism and fetishisation of old norms (indeed, see 2A; also see a paramilitary executing political opponents and how 2A influenced that) is how one ends up with a broken state.

dgroshev··on UK Home Office launches £75M 'PoliceAI' to capitalise on artificial intelligence
Yes, I'm saying it's a normal part of a functioning country to onboard tech into their law enforcement. If anything it's positive in this case, because it's a domestic effort and not just buying a similar system from Palantir.
dgroshev··on UK Home Office launches £75M 'PoliceAI' to capitalise on artificial intelligence
I don't see what's "scary" about this link. It's talking about pretty much standard data processing, Palantir is doing that in the US for many years now.
← PreviousPage 2 of 15Next →