tangled

I cancelled my subscription. I tolerate advertising on free products, but find it pretty distasteful when I'm already paying $15/month. This reeks of trying to extract every last piece of value from the customer.

Yes, customers are sticky, but every customer has their limit.

And of course, this is very much dependent on the switching costs. Which sucks, because tech companies are great at making this as high as possible; the EU is trying to do something about this with their data portability legislation.

Anecdote: I worked at a company providing infrastructure as service. We lost some bids the first time around, likely for a variety of reasons (price, features, ...). And, some customers came "back" to us after using our competitors' offerings and being burned by their reliability. Yes, switching costs were real, but the pain of losing their customers was even higher.

Each to their own, and the market will decide in the end? If a company is going to (implicitly or explicitly) incentivize everyone to directly ship Claude’s output straight to production, and prioritize features above everything else, that’s a choice. Providing that I have a long enough runway, I’d love to have this company as my competitor? Yeah, year 1 will suck, because they’ll pull ahead, but if I can stay in the game until years 2 and 3…

Aside: this reminds me of running a “negative split” in a long distance race, where you aim to run the second half faster than the first. It’s very hard to do this because you have to be willing to let everyone else in your pace group pull waaay ahead, and running above race pace in the beginning feels “free” with all of the adrenaline. But if you do manage to stay disciplined, it’s a fantastic feeling to reach half-way with gas in the tank, and then start to reel in all those runners who sped by in the beginning.

What am I missing here? I think of fireworks as an inference provider serving open weights model. The value that they primarily provide to customers is that (i) they improve reliability by balancing across a bunch of clouds/neoclouds, (ii) they get better pricing by buying capacity in bulk, and (iii) they reduce operational costs. So far so good.

I can also see the argument for providing a post-training service from a customer acquisition perspective: "hey, we can fine-tune this open weights model, so it both gives better/more predictable results than OpenAI/Anthropic and also is cheaper. And btw, once we've won your business, please run this model on our infra."

But what I'm struggling to understand is fireworks spending a bunch of money (on salaries and compute) releasing a frontier model that is going to rapidly fall behind the frontier. Is this "just" advertising for them, both for customers and also for hiring? Or are they actually trying to stay on the frontier? If so, to what end?

As others have said: this was not a good faith discussion. Huang was using the interview as an opportunity to “talk his book”, and Klein was using it to sell his own Serious Interviewer brand. Both were getting something out of this.

Given this arrangement, Klein had a limited ability to push back (well, he could, but that would mean a bunch of future interviews wouldn’t happen). Instead, he left it up to the viewer to make their own inferences. And some people are going to see Klein’s lack of push-back as a legitimization of Huang’s views.

After watching the first twenty minutes or so (I couldn’t watch any more), I had a much lower opinion of Klein as an interviewer who is willing to ask the tough questions no matter who he’s talking to.

Europe absolutely needs a home-grown AI lab, especially with Pax Americana looking increasingly shaky.

LLMs embody value systems, and American and European values are not the same (yes, there are overlaps, but also key differences).

More nefariously, I can also imagine LLMs that silently degrade their reasoning when used in a national security context of a non-US country.

So, Mistral may not be competitive with OpenAI and Anthropic, but in many contexts that doesn’t matter. And, perhaps this gap could be closed with more funding (the three billion funding figure is a rounding error next to US labs). I’m sort of surprised that the EU isn’t stepping in to support them.

My default position is that making money takes precedence over everything else. Yes, some people inside a company may say “we care about doing the right thing” and they might even mean it, but if that comes into conflict with making money, then they tend to lose. Maybe not totally, or immediately, but in the end. The only effective way to prevent (this that I’ve seen) is to have legislation with teeth. It’s probably not a coincidence that after Mark Zuckerberg had to start personally signing off on adherence to the privacy program mandated under the 2020 FTC consent decree, privacy started to become Very Important.

Having previously worked for several years at a Big Tech company, I have seen many humans precisely tailor their work to maximize their scores during performance review. The evaluation criteria are written down, with examples, so... that's what people work at maximizing, almost entirely ignoring everything else. These really are human "paperclip maximizers". And, at first, it's shocking to see.

Of course, there are some things that aren't exactly written down, but which you should either do just enough of, or else be able to plausibly deny doing (ignorance is a good cover for this), so that's what people do. For example, during oncall, you investigate just enough to clear the alert and show that you attempted to understand the problem. Of course, you don't really try to understand the problem, because that would take too much time away from your paperclip maximizing.

Which is all to say: I don't know anything about OpenAI culture, or why nobody stopped this sooner, but I have seen examples in other organizations of people not really wanting to understand too much.

This feels like one of those rather rare projects that is both sailing pretty close to research, and also yielding industrially useful results -- I love it! Normally I'd expect to see something like this coming out of one of the big tech companies, where there are enough advertising megabucks to pay a small team to work on a project like this (providing someone can make the business case...). So I'm curious: what was the initial motivation for this work? Assuming this is not a passion project, who is funding it? How many person years of work has this involved? What is the end game?

To quote the NYTimes:

“Despite the desperate humanitarian crisis, a survey conducted in May by the Institute for National Security Studies at Tel Aviv University found that 64.5 percent of the Israeli public was not at all, or not very, concerned about the humanitarian situation in Gaza.”

“About three-quarters of Israeli Jews thought that Israel's military planning should not take into account the suffering of the Palestinian civilian population in Gaza, or should do so only minimally, according to another recent survey by the Israel Democracy Institute, a nonpartisan research group in Jerusalem.”

https://www.nytimes.com/2025/07/28/world/middleeast/israel-d...

Yes! Email in profile.

(I worked for Yelp for over eight years, knew large parts of the codebase, and never once saw any evidence that it was possible to pay to remove bad reviews.)

I’ve pondered the implications of running an ‘installation art’ project where a bunch of these are deployed around a city, with aggregated data visible over a web UI (assuming that this doesn’t violate any laws.) The aim would be to raise public awareness of pervasive surveillance, and perhaps catalyze changes to the law.

I'd be interested in seeing the buy vs. build analysis for this project. Are there any pre-existing projects that have similar features? And assuming that this project is used to process classified data, what impact does this have on the selection process? e.g. is it possible to use closed-source solutions?

A lot of this applies to good security teams as well :) For example, reaching out to offer consulting services at an early stage in projects, building goodwill by advising instead of mandating (where possible), and constantly communicating best practices.

In California, property taxes are usually only reassessed on sale. There is also an inflation linked increase, but it's capped at a maximum of 2%. So you can end up with people who have been living in houses for 20 or 30 years who are paying an order of magnitude less in property tax than if they'd just bought the same house today. This also means that if you buy a house in a down market then you're likely to save a lot in terms of property taxes. Disclaimer: I am not a tax adviser etc.

Because then you would be bringing additional stored energy into the competition (beyond what's in your body)?

I'm guessing that the rules prohibit using some of the pedal power to charge a battery during the initial part of the run, and then using that stored energy to provide an additional speed boost :D

This is approach is great for greenfield development, but when it comes to software maintenance I find myself having to think in terms of the target language so that I can formulate the right long-term solution for the codebase in question.

I'd be interested in seeing why instagram didn't implement the suggested user service as a batch process that populated some KV store. Online evaluation is great for guaranteeing up-to-date results, but the tradeoff is that you now have to bring up and maintain a fairly critical production service.

> The SU service fetches candidate accounts from various sources, such as your friends, accounts that you may be interested in, and popular accounts in your area. Then, a machine learning model blends them to produce a list of personalized account suggestions.

> This enabled us to reduce the number of instances of the Suggested Users service from 720 to 38.

I used to have a Happy Hacker keyboard and write Java (and other languages) in a customize emacs environment. These days I use a regular mac keyboard (so I can easily transition between my macbook and desktop setups) and run IntelliJ. And I feel like I'm a whole lot faster writing production-quality code. There are definitely a few annoyances when using the IntelliJ emacs key binds, but this is more than compensated by being able to do things like automatically download and jump to referenced packages and having the IDE suggest code improvements ("I've noticed that you could be using the Java 8 lambda syntax here..." / "Do you want to remove these unused imports" / "This value may be undefined"). And now I can't write Java without IntelliJ, and I'm going to fail my next whiteboard interview ;)

I'm not sure it is fatalism. Could the past fifty years or so just be an anomaly? Is it that liberal democracies are inherently unstable as they scale up? Already, it's basically impossible to have a single-earner family, whereas I think that was pretty common in the post WWII years. And education costs continue to rise (my parents were paid a stipend to become doctors in the UK... those times are long gone now).

Do you actually believe that there's much difference between the republicans and democrats (that's a serious question)? My cynical take is that both parties are bank-rolled by the same set of people :) If I had my billions, then I'd pay both sides to ensure that my voice were heard whatever the outcome.

I find it hard to imagine abrupt corrections happening in the western world. Everyone is just so very controlled. But then maybe people thought the same thing leading up to the French revolution 'only' 200 years ago? From my limited knowledge, the French had a stable aristocracy with a controlled populace. And then just a few years later the new powers were executing tens of thousands of citizens. That's crazy.

I've only used ascenders when climbing up free hanging rope, and they work very well for that purpose (I've done several hundred meters of vertical ascent in a single trip using this technique).

An alternative to this is the mechanical ascender / 'jammer'. I've used the Petzl version when caving: http://www.petzl.com/en/Sport/Ascenders

They're practically bullet proof, although they can take a little longer to bite when coated in mud :)

> one guest left a New York restaurant with the impression Bernanke, 60, does not expect the federal funds rate, the Fed's main benchmark interest rate, to rise back to its long-term average of around 4 percent in Bernanke's lifetime, one source who had spoken to the guest said.

http://mobile.reuters.com/article/idUSBREA4F0OG20140516

The article is about restarting HAProxy without any downtime. HAProxy restarts are needed when adding new service instances or adjusting configuration options. This is a different and much harder problem than gracefully restarting load balanced service instances.

It's interesting to read Xavier's annual statement on why there will never be multi core support in OCaml: http://mirror.ocamlcore.org/caml.inria.fr/pub/ml-archives/ca...

No, the author claims he has no choice: "By voting in a Tory government, you basically did not leave us any other choice." This feels like a dishonest representation of the situation.