ketzo

"All mental models are incorrect; some mental models are useful"

"Trojan horse"? 95% of people currently access AI via web or mobile app; those are pretty JS-dominated, no?

It’s already becoming politicized, in the lowercase-p sense of the word. One is assumed to be either pro- or anti-AI, and so you gotta do your best to signal to the reader where you lie.

I just think you’re totally wrong about that, to be honest. It’s very strange to me to paint every AI researcher as an unfeeling robot who does nothing but read PDFs, who’s not remotely interested in how their work will ever be used by real people.

No, frontier model researchers are not learning about the concept of AI from Twitter.

But it’s one more piece of visibility, one more sense of “hey this company is actually doing stuff with their models,” that absolutely contributes to where these insanely in-demand people choose to do their work.

> Leading edge chip designers at TSMC didn't end up there because they ended up playing graphic intensive video games

What a baffling and wrong example! I mean, I guess I don’t know about TSMC specifically, but tons and tons of technologists have an origin story that rhymes with “I was really into computer games, then got curious how they worked”

This is my whole point: engineers are real people, and real people are fascinated by the uses of AI!

I feel like you underestimate the degree to which AI developers are influenced in just the same ways as everybody else.

Capturing mindshare matters no matter how you do it. I would bet my life savings that more people will join OpenAI to work on their models as a result of the Ghibli moment than will join Google as a result of this particular (incredibly impressive!) Gemini iteration.

Because we’re humans! And we’re all pretty easily impressed, even those of us who actually build these dang things.

Right — which is to say, the vast, vast majority of people heard about Ghibli and not Gemini.

Particularly with technical writing, I think you can definitely get away with both.

“How I Reduced My Postgres Query Latency By 100x With A Single Index”

Even in the title, I can tell you the punchline (if you wanna make your DB access faster, use an index!)

but an interested reader still wants to figure out how exactly your solution works, and you can tell them some interesting details along the way

“just enforcing unique constraints does help certain data types, but it’s not a big performance boost most of the time”

while finishing on the kicker

“Since my hottest endpoint by far was for individual users querying orders which were still ongoing, I created an index on the user field for the orders table, and included a status filter in the index, which took p90 latency from 10s to <100ms!”

> unless you are vastly incompetent at basic police work.

Boy do I have news!

We already have multiple news stories about xAI receiving 10,000s of GPUs that were originally meant for Tesla. That's about as close to "public company giving private company owned by CEO money" as you could possibly get without cartoon sacks of cash.

> Even by Tesla & Musk lawlessness standard that would be too wild.

Not to be overly snarky here, but, uh: what standard? They have proven over and over that they could not give less of a shit about either societal norms or the laws of the nation.

The linked article can tell you that the ChatGPT app has been downloaded more than 600M times, so approx 1 in 13 humans has actually downloaded ChatGPT to their phone already, let alone heard about it.

They have WAUs in the 300m ballpark — average people already are going out of their way to open a standalone AI.

Based on other comments in this thread it sounds like this takes place in SF’s Mission neighborhood — among other things, the warmest and least foggy area in the city!

I wonder about how to do this in my own, much chillier neighborhood… guess getting started in the summer is a big key (the author hits on the broadening/deepening event split re: weather)

If you think “garage coffeeshop” is a fast track to wealth, boy do I have news!

There were a bunch of headlines last week about DOGE firing half of NOAA then rehiring them. I don’t think you should consider anywhere sacrosanct.

I mean, I don’t think it’s a fairytale idea; the implication of the opening discussion on wealth inequality is that the US would need to raise taxes fairly significantly to do so.

The question is really just whether the political will to do so exists. If that’s what you’re referring to as fanciful, I guess I can see where you’re coming from.

Well, that's sort of the author's point -- every framework is focused on something other than simple salary numbers, which definitionally means it is focused on something other than the dollar value provided to the company.

We recognize that they can still be a useful tool, because 1) assigning dollar values to employees is an inexact science that's hard to standardize over an organization and 2) ICs like at least some direction about what they need to do to be eligible for more money aside from just "provide more value"

Wait, what? Utilities all over the world have rate systems that offer higher or lower prices based on who's paying. "Water is fungible" doesn't mean that you can't charge Amazon more for it than you do the little old lady down the street.

I'm not saying that the current rate systems in Mexico, or California, or XYZ place are perfect -- far from it! But there's no reason they couldn't be changed to fit society's goals.

> It’s a meat grinder that chews up developers, sysadmins, and infosec pros

…and designers, and PMs, and sales and marketing and support and…

I know that, charitably, the point is that engineers often have higher leverage than other individual contributors at a tech company.

But the whole point of a union is to join together employees (as opposed to employers) for greater leverage. Don’t make your tent smaller for no reason!

Okay, I was being a little snarky. But that quoted portion doesn't really contradict my point; that is in no way an exploration of good reasons why Apple might want to make these choices.

I think this is an excellent addition to the perspective of the article! Why do you not think so?

> Anywhere Apple trusts itself is a place where they can trust a third party.

This quite literally could not be further from the truth, and to suggest that it is true reflects such a comprehensive misunderstanding of both the fundamental nature of computer security and the practical realities of the world in which we live that it's not really possible to continue the conversation productively

That’s not the point, though. Any method by which apple exposes APIs to Samsung, Google, etc:

- requires immense development effort and expansion of security surface area

- STILL offloads trust to Samsung, Google, etc

The hyperbole here is a little hysterical. Apple doesn’t totally lock out third parties. In the smartwatch example, it is a very specific set of features which involved passing data (which users expect to be e2e encrypted!) back to Apple. That’s an extremely hostile security environment! Product tradeoffs would absolutely have to be made in order to support arbitrary third parties! I don’t think it’s fair to just demand that Apple make their product worse without at least exploring the balance.

> Apple claims their restrictions on competitors are only about security, privacy, crafting a better experience etc etc. At least that’s what they tell you as they tuck you into bed.

Ah, yes, the author is clearly interested in an in-depth discussion of the tradeoffs in allowing 3rd-party users access to data that you tell your customers is 100% always encrypted.

You're being sarcastic, but isn't this all just... correct?

Yes, I do trust the company that developed Secure Enclave more than I trust random BLE firmware in a $49 Alibaba watch.

More importantly -- my great-uncle can trust the same thing, because Apple has spent decades building that trust. Consumers generally should not trust random hardware. Apple is not random hardware.

No, the vendor's BT stack would be responsible for broadcasting any responses back to the device -- like, in the article, "send text messages, or perform actions on notifications (like dismissing, muting, replying)"

The way people on HN sometimes talk about "management," you'd think the universal experience is 1% good, 9% benign, 90% actively seeking the downfall of civilization

Managers seem like a good example of the "toupee problem" -- the ones you notice, and really remember, are the bad ones; the best, you might never see at all.

Most informed “dark pools” comment

Is this gonna be the thing that finally makes me tried Rails? Ruby syntax really is just nice.

Yeah, I think a lot of people talk about "fixing hallucinations" as the end goal, rather than "LLMs providing value", which misses the forest for the trees; it's obviously already true that we don't need totally hallucination-free output to get value from these models.

My 2c would be:

- tech-oriented publicists / marketers definitely do know that HN exists and matters

- there are “benign” ways to game HN, like releasing at an optimal time for readership / upvotes

- there are less nice ways, like sock puppet accounts / botting

- all of these things are certainly done regularly

- it doesn’t mean that everything you read was an astroturfed psy-op or whatever.

The ~3 times this has ever happened to me, it made my week. Cannot recommend enough.