HNHacker News
TopNewBestAskShowJobs

glerk

1,220 karma · joined January 28, 2019

some guy on the internet
submissionscomments
glerk··on Gmail thinks I'm stupid, so I left
I just can't stand how Gmail is putting a red line under every other sentence that I write (telling me that my writing style is a "mistake") and aggressively nudging me to rewrite it to make it sound more like AI.

Whoever thought such a product would be a good idea should be fired.

glerk··on Shantell Sans (2023)
I like it! Somehow balances playfulness and readability. Thanks for sharing.
glerk··on Alignment whack-a-mole: Finetuning activates recall of copyrighted books in LLMs
Where did I say that writing or journalism are rent seeking? And why are you being dishonest?
glerk··on Alignment whack-a-mole: Finetuning activates recall of copyrighted books in LLMs
> Oh no!! Those strings of words belong to me!!

Yeah, maybe it’s time to move on and find ways to benefit yourself and the rest of humanity outside of artificial monopolies and rent seeking. Copyright is dead.

glerk··on China blocks Meta's acquisition of AI startup Manus
> As an anarchist I view civilization to be a hack on the human condition, and I see all states as fundamentally authoritarian.

I think we are not really in disagreement, it's mainly an argument over the semantics of "legitimate" at this point :) Rousseau and Hobbes were both right.

glerk··on China blocks Meta's acquisition of AI startup Manus
> The government's legitimacy comes from it's stick being bigger than yours

By the same argument, are Somalian warlords and Mexican drug cartel also legitimate in the territories they control? I don't think "legitimate" is the word you are looking for to describe pure power dynamics, since "legitimate" is imbued with a moralistic judgement (look up is vs ought etc.). But yes, in practice, if I have a gun pointed at my head, I could be forced to do things that go against my judgement (within limits!).

glerk··on China blocks Meta's acquisition of AI startup Manus
If we’re going to descend into pedantry, my statement was normative, not descriptive, as in “I agree this is what a government does, I disagree this is what it _should_ do”.

“Beneath me” is _my_ value judgement that I pass on this government and its appendages as in “it has been weighed in the balance and has been found unworthy”. That this government has more power than me doesn’t make it sit above me as a moral absolute, and it doesn’t magically give it legitimacy.

glerk··on China blocks Meta's acquisition of AI startup Manus
> Every country does it. Doing it is a central function of having a government.

You are falling back on whataboutism. This is irrelevant. If we were having a similar debate in the middle ages, you would probably say something like:

> Every church is burning witches and heretics at the stake. Doing it is a central function of having a church.

The CCP has abducted these individuals and is preventing them from leaving the country. This is not ok. You can't justify this by saying "yeah, but they're the government, so it's their right to abduct whoever they want". A government is just a corporation with a bit more power than the others, not some sacred entity that sits above us.

glerk··on China blocks Meta's acquisition of AI startup Manus
Just because the US also does this doesn’t make it right for China to do it and vice versa.

Team coca-cola and team pepsi are both evil and illegitimate.

glerk··on Show HN: GoModel – an open-source AI gateway in Go
This is awesome work, thanks for sharing!

How do you plan on keeping up with upstream changes from the API providers? I have implemented something similar, and the biggest issue I have faced with go is that providers don’t usually have sdk’s (compared to javascript and python), and there is work involved in staying up to date at each release.

glerk··on Salmon exposed to cocaine and its main byproduct roam more widely
Did you actually find a place where you can buy this beverage (the official version)? Asking for a friend of course.
glerk··on Rumor: Anthropic is going to buy Atlassian?
I'd buy the business book, and keep some of the staff for the expertise, but scrap that horrible product and rebuild it from scratch.
glerk··on Ask HN: How to solve the cold start problem for a two-sided marketplace?
There is probably no other way than to seed it. You can do this with outright bots or contractors from fiverr, but people who grew up with the internet are very good at spotting inorganic growth.

I'd start with a slow rollout to friends, family, colleagues and even LinkedIn acquittances. You'd be surprised how many people are super eager to be the first to test your apps, even with a generic cold message like "hey, remember me? I built this cool app, wanna try it out if you have a minutes? I'll send you money to cover the fees."

glerk··on Landmark ancient-genome study shows surprise acceleration of human evolution
Lmao sure thing you fucking moron.
glerk··on Landmark ancient-genome study shows surprise acceleration of human evolution
You are obviously too disingenuous to engage in any kind of dialogue and it showed from your very first reply. But sure, thanks for the recs.
glerk··on Landmark ancient-genome study shows surprise acceleration of human evolution
I'm gonna be very blunt with you (disregarding again the goalpost moving from your earlier claim that "speciation is reproductive incompatability (sic)"): the jargon you are throwing around is not helpful, it is pure linguistic diarrhea you are using in an attempt to project an image of erudition.

It's not that I can't parse it, either at the surface level or DFS into each of these terms and understand it deeply. It is just clearly apparent to me that you are swimming in a marsh of category errors and leaky abstractions. This field has accumulated a lot of entropy over time. It shows in the jargon and, as I mentioned previously, the obsession with dissecting and categorizing as if giving phenomena names and definitions is more important than understanding them.

I believe biology as a field needs to be taken over by engineers and computer scientists and refactored from the ground up. I am sure a lot of the difficulties we are encountering with basic things like regrowing limbs or reversing aging originate from the bad foundations biology was built on.

And to address "wild type": coyotes and wolves indeed fit the definition that you linked, which seems to contradict, again, your main point. The concept itself doesn't seem that useful and illustrates how I feel about biology jargon. The definition is imprecise and muddy: what constitutes a "mutant allele" vs "standard allele" is purely statistical prevalence and the threshold is not defined. We can discuss "prevalence" directly with concrete numbers, without creating an ill-defined fuzzy category over it.

> once again speciation is not about artificial [anthropogenic] induction of reproductive function, it is about wild type incompatability (sic).

Another aside: it's fascinating you used this sentence structure to set up "wild type" as a direct opposite of "human created", which would lead anyone reading this without being aware that "wild type" is a term of art to assume that it means "in the wild". This is not the gotcha you think it is...

glerk··on Vercel April 2026 security incident
NextJs requires what exactly? Running a nodejs server? I mean yes, it takes a bit more time to set up than one-command deploy to Vercel. But in 2026, even this setup overhead can be cut down to minutes by telling your favorite LLM agent to SSH into your server and set it up for you.
glerk··on Vercel April 2026 security incident
There really isn't any if you are running a serious product.

They have a free tier plan for non-commercial usage and a very very good UX for just deploying your website.

Many companies start using Vercel for the convenience and, as they grow, they continue paying for it because migrating to a cheaper provider is inconvenient.

glerk··on Landmark ancient-genome study shows surprise acceleration of human evolution
> oh by the way, we are a species of primate, not dust, and rib bones.

Fascinating. Not sure what gave you the idea I don’t believe in evolution or that I’m somehow promoting biblical creationism? Are you responding to the right comment?

> speciation is not about artificial [anthropogenic] induction of reproductive function, it is about wild type incompatability.

Neither coyotes nor wolves were created by human selection as far as I know. Dogs were. You can take dogs out of the equation if you want.

Coyote/wolf hybrids (coywolves) exist in the wild and challenge your definition. And I am talking about your original comment's definition "speciation is reproductive incompatibility", because I believe you backtracked a bit with the more vague "wild type incompatibility".

Besides, I don’t necessarily disagree that wolves, dogs and coyotes should be seen as the same “species”. I find this obsession with taxonomy completely useless and irrelevant. We all know there are biological clusters and the boundaries are fuzzy, but we can use simplified/imprecise models when communicating because it is more convenient.

> citation required

https://en.wikipedia.org/wiki/Coywolf

glerk··on Landmark ancient-genome study shows surprise acceleration of human evolution
You are free to create your own human classification scheme with k sub-species and try to popularize it, there's nobody stopping you and there isn't some authority out there called "the scientific community" that's going to send you cease and desist letters.

There's just not much to gain from the exercise and there are better things to spend your time on.

glerk··on Landmark ancient-genome study shows surprise acceleration of human evolution
> speciation is reproductive incompatability

This is not necessarily true, even under a very strict definition of reproductive compatibility (offspring is itself not sterile, which excludes mules or ligers). For example, feral dogs, wolves and coyotes regularly mate and produce 100% viable offspring. You could argue that these are not really different species, but they are usually classified as such.

glerk··on Anonymous request-token comparisons from Opus 4.6 and Opus 4.7
Of course, and I feel like Codex/GPT is generally better at following instructions and implementing a step-by-step plan and at a lower cost. Opus still has an edge in writing, brainstorming, and open-ended frontend vibe-coding.

I’m definitely not coming to this from a “AI is useless” angle. I’ve been using these tools extensively over the past year and they are providing a massive productivity boost.

glerk··on Anonymous request-token comparisons from Opus 4.6 and Opus 4.7
I like your style, and I appreciate you trying to get to the truth, despite us both being aware that we are engaging in persuasive writing here, so part of the rhetorical game is in what we choose to emphasize and what we choose to leave out.

> How likely do you think this is? Do you think it is more likely than the other three I mentioned?

I won't write down probability estimates, because frankly, I have no idea. Unless you are yourself a decision-maker at Anthropic, which, from what I can infer, you aren't, both of us are speculating. However, I can try to address each of your explanations at face value, because I don't think any of them makes Anthropic look any better than the explanation I provided.

> (1) One possibility is they are having capacity and/or infrastructure problems so the model performance is degraded.

As far as I understand it, scaling issues would result in increased latency or requests being dropped, not model quality being lower. However, there is a very widespread rumor that Anthropic is routing traffic to quantized models during peak times to help decrease costs. Boris Cherny, Thariq Shihipar, and others have repeatedly denied this is happening [1]. I would be more concerned if this were the actual explanation, because as a user of the Claude Code Max plan and of the API, I have the expectation that each dollar I spend buys me access to the same model without opaque routing in the background.

> (2) Another possibility is that they are not as tuned to what customers want relative to what their engineers want.

There is actually a strong case for this: the high performance on the benchmarks relative to the qualitatively low performance reported on real-world tasks after launch. I suspect quite a bit of RL training was spent optimizing for beating those benchmarks, which resulted in overfitting the model on particular kinds of tasks. I'm not claiming this is nefarious in any way or that it is something only Anthropic is guilty of doing: these benchmarks are supposed to be a good representation of general software tasks, and using them as a training ground is expected.

> (3) It is also possible they have slowed their models down due to safety concerns. To be more specific, they are erring on the side of caution (which would be consistent with their press releases about safety concerns of Mythos).

This would be the most concerning to me. I don't want to get too deeply into a political/philosophical argument, but I am very much on the other side of the e/accy vs. P(doomy) debate, and I strongly believe that keeping these tools under the control of some council of enlightened elders who claim to know what is best for humanity is ultimately futile.

If the result of the behind-the-scenes "cerebration" is an actual effort to try and slow down AI development or access, I don't have much confidence in the future of Anthropic.

I agree that there are incentives other than pure profit maximization here (I don't want to get into "my friend at Anthropic told me such and such" games, but I also believe this is the case). I'm sure there is some tension between these objectives inside Anthropic, but what is interesting is that lower model quality and maximizing user engagement could, at least in principle, align with both constraints.

[1] https://x.com/trq212/status/2043023892579766290

glerk··on Anonymous request-token comparisons from Opus 4.6 and Opus 4.7
> To claim to know a company's strategy as an outsider is messy stuff.

I said "it seems like". Obviously, I have no idea whether this is an intentional strategy or not and it could as well be a side effect of those things that you mentioned.

Models being "worse" is the perceived effect for the end user (subjectively, it seems like the price to achieve the same results on similar tasks with Opus has been steadily increasing). I am claiming that there is no incentive for Anthropic to address this issue because of their business model (maximize the amount of tokens spent and price per token).

glerk··on Anonymous request-token comparisons from Opus 4.6 and Opus 4.7
I'd be ok with paying more if results were good, but it seems like Anthropic is going for the Tinder/casino intermittent reinforcement strategy: optimized to keep you spending tokens instead of achieving results.

And yes, Claude models are generally more fun to use than GPT/Codex. They have a personality. They have an intuition for design/aesthetics. Vibe-coding with them feels like playing a video game. But the result is almost always some version of cutting corners: tests removed to make the suite pass, duplicate code everywhere, wrong abstraction, type safety disabled, hard requirements ignored, etc.

These issues are not resolved in 4.7, no matter what the benchmarks say, and I don't think there is any interest in resolving them.

glerk··on Discourse Is Not Going Closed Source
I’d rather next models be trained on my code than all that slop that’s out there. Hell really is other people’s code.
glerk··on Isaac Asimov: The Last Question (1956)
What do you think would have been more valuable for him to do? His sci-fi books had a huge impact, and not only on sci-fi and literature, they literally changed people's lives. People decided to pursue a career in science or technology because they read these books when they were kids.
glerk··on Discourse Is Not Going Closed Source
I've started to opensource my side projects (as long as the code is in a state that I'm not too ashamed of). Seeing how easily I can reverse engineer binaries, clone various applications, and just generally build stuff from scratch with AI assistance, I think there is no moat in hiding your source code. If you can use my code to build something better than me, I wish you the best of luck!
glerk··on Good sleep, good learning, good life (2012)
I am not dismissing the research and I am well aware there are tradeoffs. I am aware that low sleep puts extra strain on the brain and the body at large. I am aware of the increased inflammation, decreased CSF draining, cardiovascular issues, appetite issues, insulin tolerance etc. etc. I know that our bodies were optimized by evolution to function better with sleep cycles synced with circadian cycles, and I know that many of these very well optimized processes are put under stress when you sleep less than 6h per night.

Here's the thing: say what I have is a "disease" and currently, the best cure for this disease is a pill that would slow down my thoughts and make me into a bland emotionless vegetable during my waking hours in addition to reducing the amount of waking hours that I get to enjoy. Even if that magic pill buys me 20 extra years of old age at the tail end of my life, is it rational for me to take it in my early 30s?

glerk··on Ohio prison inmates 'built computers and hid them in ceiling' (2017)
Their thinking is that making the conditions bad will serve as deterrent i.e. would-be criminals would think twice before committing crimes because they're scared of going to prison.

Of course, this makes no sense, as most criminals have low impulse control and don't think about the consequences of their actions in terms of risk/reward calculations. We should use prison time to re-educate these people and try to make them better instead of psychologically torturing them, but here we are, and it's very unlikely things can change within the current political system (too many "checks and balances" for meaningful reforms)

← PreviousPage 3 of 13Next →