zetazzed

Ok, but how do the economics of this work? Based on its settlement, Anthropic paid an average of $3000 per work they scanned based on their settlement (https://tech-insider.org/au/anthropic-copyright-settlement-2...). They and OpenAI pay billions per year for a mix of experts and normal people to label or create data. Why would they continue doing this if the value of this is immediately copied by open models? If your goal is to end the economics of generating and buying data for AI (and I recognize for some people this is really the goal) then sure, but if you want AI for various subfields of interest to continue improving then it's not workable.

Back when people made arguments for software privacy, the argument was usually "big business will still pay and consumers wouldn't have paid anyways so it's ok for us to pirate" - I actually think that was fine for business software but terrible for indie games, whose market was 0% businesses.

But in the AI case, it's not like they get to keep some of the value of their investment - it all gets cloned into models that businesses and consumers alike are happy to use. If someone knows how labs could continue to fund data creation and acquisition in this model, please do share!

I uninstalled basically all of these EXCEPT youtube, and that's what I struggle with sometimes now. I find so much great long- and medium-form content on Youtube that I want to keep it around. But I have to admit that I can sometimes fall down a time-wasting rabbit hole with shorts.

The Youtube app now has a "shorts feed limit" under "time management" that will limit your access to the auto-scroll shorts feed to a certain amount per day. Worth checking out!

I'm a good target market for pro-coffee writing, but this is just an observational study of a drinks habit. No one in the general public should have any interest in studies of this kind.

Maybe if you're deep in working in this research space it could help generate hypotheses for more meaningful research? But "we cross sectionally compared two groups with different behaviors and found they were different" is not really a result.

The NYT just did a story on Andon: https://www.nytimes.com/2026/08/04/us/ai-boss-san-francisco-...

Their AI "boss" appears to be pathologically chill and worker-friendly. Has never denied a request for a day off.

Who will invest in generating data for the frontier of AI if their output will immediately be used to train a competing model? Forget China vs. US, this applies within-country too. After exhausting all the publicly-accessible data on the internet, the frontier labs started spending hundreds of millions of dollars to generate data across a variety of fields. Just look at Mercor doing $1.2bln/year with 90% coming from the top labs (https://www.theinformation.com/articles/mercors-fast-growth-...). That pushes ahead what AI can do in medicine, science, coding, and math. But if other companies are going to free ride on this investment, it doesn't make sense to continue. So AI will largely hit a wall, frozen at the current level and work will all shift to cheaper inference.

It's like the Victoria 3 combat system. You just send an army and a general to a given front and let them get to work with no micro. Easy! But of course some percentage of the time they do something crazy like deciding to redeploy from your existential Franco-Prussian war front to a minor colonial uprising...

This is actually not such bad advice for a manager who manages other managers, though I can see why ICs find it very frustrating. If you are giving high level platitudes and counseling-disguised-as-coaching to a junior new hire, they can rightly ask WTF. But managers, especially those recently moved from IC tech roles, often do benefit from this kind of forced introspection. If they have an underperforming employee, they should bounce ideas around with a more experienced manager, but the first line manager ultimately needs to be the one deciding how to rebalance work to maximize learning or to ultimately make the call to part ways with the employee. If a servant senior leader over them is actually doing the slog of working through the hardest issues (interpersonal conflict, serious direction change needed for team, firing people, top performers at risk of leaving), the first line manager is never going to grow. Similarly "cut out the middleman" advice in the article is great for senior ICs/quasi-architects or sub-managers but potentially toxic for junior engineers who may get steamrolled by the classic "1000 urgent requests issue" that managers or potentially very senior ICs need to drive.

The total profit of ALL US health insurance companies added together was $9bln in 2024: https://content.naic.org/sites/default/files/2024-annual-hea.... This is a profit margin of 0.8% down from 2.2% in the previous year.

Meta alone made $62bln in 2024: https://investor.atmeta.com/investor-news/press-release-deta...

So it's weird to see folks on a tech site talking about how enormous all the profits are in health insurance, and citations with numbers would be helpful to the discussion.

I worked in insurance-related tech for some time, and the providers (hospitals, large physician groups) and employers who actually pay for insurance have signficant market power in most regions, limiting what insurers can charge.

Is it easy to record a voiceover or add chosen audio? (Sorry I don't have an invite code so I can't try.) I could see some room for human jokes or short human-driven songs that could use a video backdrop.

Does Disney have a positive obligation to show animal cruelty snuff films on Disney Plus? Or are they allowed to control what people say on their network? Does Roblox have to allow XXX games showing non-consensual sex acts on their site, or are they allowed to control what people say on their network? Can WebMD decide not to present articles claiming that homeopathy is the ultimate cure-all? Does X have to share a "trending" topic about the refusal to release the Epstein files?

The reason we ban government censorship is so that a private actor can always create their own conspiracy theory + snuff film site if they want, and other platforms are not obligated to carry content they find objectionable. Get really into Rumble or Truth Social or X if you would like a very different perspective from Youtube's.

Their revenue is like $670bln. If you come up with an innovation that increases that by 0.01%, say by better optimizing prices or targeting adds, you've added $60m of revenue. If you pay a star engineer $1m per year and they have even a reasonable chance of an improvement on this scale, or a similar reduction in costs, then you have a super profitable deal.

Ok but telling someone they can have GPUs online in a mere 5 years if they build as fast as SK is still going to be a very painful pill. How do we get a DC in a year?

I see an increasing number of politicians taking the position: "I supported Israel's government's actions when they first attacked, given the goals of destroying Hamas' leadership and freeing hostages, but now that it has turned into a brutal siege with mass civilian casualties on a horrific scale, I'm strongly against their actions." E.g. Macron, Angus King, and many people I know personally. And I think we need to say "Great!" The dumbest reaction is "screw you, you were for Israel's invasion and you're an asshole." Movements that want to grow should accept people who change their minds when the situation changes, they get new data, or they learn a new perspective.

Manufacturing output has more than doubled in the US since 1980: https://fred.stlouisfed.org/series/IPMAN#

Companies can choose their gross margins! Just take the cost of manufacturing, double it, and call that your price. You now have 50% gross margins. But you can't choose your total profit. Maybe almost nobody buys at this crazy price you've set. Now your margins don't seem so clever...

In the models I've seen, they still require and bill insurance. The monthly fee is a supplement for the doc practices.

Ok, I appreciate that timelines in this space are long. But the opening phrase:

"Toyota Motor Corporation (“Toyota”) and Waymo reached a preliminary agreement to explore a collaboration focused on accelerating the development..."

reads a bit like a parody of corporate speak about a project nowhere close to happening. Did they agree to deploy? Or reach an agreement to collaborate? No, that's too strong. They will EXPLORE collaborating on ACCELERATING development.

Ōdī et amō. Quārē id faciam fortasse requīris. Nesciŏ, sed fierī sentiō et excrucior.

I hate and I love. Why I should do this, perhaps you may ask... I know not, but I feel it done to me, and I am wracked.

I feel like LLMs are the same as the leap from "world before web search" to "world after web search." Yeah, in google, you get crap links for sure, and you have to wade through salesy links and random blogs. But in the pre-web-search world, your options were generally "ask a friend who seems smart" or "go to the library for quite a while," AND BOTH OF THOSE OPTIONS HAD PLENTY OF ISSUES. I found a random part in an old arduino kit I bought years ago, and GPT-4o correctly identified it and explained exactly how to hook it up and code for it to me. That is frickin awesome, and it saves me a ton of time and leads me to reuse the part. I used DeepResearch to research car options that fit my exact needs, and it was 100% spot on - multiple people have suggested models that DeepReearch did not identify that would be a fit, but every time I dig in, I find that DeepResearch was right and the alternative actually had some dealbreaker I had specified. Etc., etc.

In the 90s, Robert Metcalfe infamously wrote "Almost all of the many predictions now being made about 1996 hinge on the Internet’s continuing exponential growth. But I predict the Internet, which only just recently got this section here in InfoWorld, will soon go spectacularly supernova and in 1996 catastrophically collapse." I feel like we are just hearing LLM versions of this quote over and over now, but they will prove to be equally accurate.

In a past startup, we had at least one person apply to our regular job postings with a cover that transparently said "I know this is a full-time, long-term posting, but I really want to be a contractor for a bounded time." Since it was a great fit and they were available right away (and we urgently needed more people), we made the "hire" and ended up working together for a while. Only worked because it was quite transparent and up front in the application though.

Algebra in 8th grade is back this year. https://www.sfusd.edu/about-sfusd/sfusd-news/press-releases/....

I have no other knowledge to vouch for SFUSD either way though.

The Making of the Atomic Bomb by Richard Rhodes! Lots of detail, but it also shows how many geniuses it took working over years to really make it work.

I'm still surprised the "dumb phone for kids" and "dumb watch with basic comms" markets are so underdeveloped (from my perspective). I would love my kids to have (a) gps tracking, (b) ability to send texts/calls to like 5 predefined numbers, (c) tell the time, and nothing else. But watches all seem to have games or weird gamified fitness trackers (Google's new fitbit for kids). Or they are super kid-ish, like bright blue with animal icons, and would be revolting to my older kids. That would make it easy to wait until 9th grade for a more feature-rich phone, though maybe still not unfettered access.

Does anyone have a basic watch/dumbphone solution for older kids that they like?

Controversially, Rawls asserts that you clearly want to maximize the worst outcome in this case (the "maximin" principle). Economists strongly disagree as it doesn't take into account probability. If you have the potential to live in Junkland, where life is moderately unhappy for everyone, or in Omelas, which is utopia for 1,000,000 individuals and pure torture for 1, Rawls asserts that you should obviously choose Mediocreland. But if you are randomly assigned a place in Omelas, you have really great odds of coming out quite well. The book includes some contortions to ensure this outcome. It has always been surprising to me that the original position argument had so much impact when it seems like a very odd starting point to me. But I like probability and statistics, so maybe I'm weird.

This is a really important problem, and a lot of software attacks it. Pretty much every big HR/workforce management system includes options for it, e.g. https://www.workday.com/en-us/products/workforce-management/... and https://www.oracle.com/human-capital-management/workforce-ma..., plus a lot of dedicated providers.

As some other comment mentioned, these systems have been controversial, because some of them are used in ways that don't take into account normal human needs. E.g. some schedule back-to-back shifts, change with little notice, and can't take into account sorts of real life things (like child care) that a human manager might be able to.

There's a phenomenon in which inflation-related posts generate lots of comments or concerns that the CPI methodology is manipulated in some way. If you are interested in learning more about CPI, the methodology is very complex but transparent. You can find documentation here: https://www.bls.gov/cpi/factsheets/ and raw datafiles here: https://www.bls.gov/cpi/data.htm

If you would like to publish your own dataset based purely on changes in price level of chocolate cupcakes, per pound, in cities, you can do so by extracting series APU0000702411 from the raw datafiles.

There is also a short summary on "why averages and an individual's experience of inflation may differ" here: https://www.bls.gov/cpi/factsheets/averages-and-individual-e...

Although it was not my main subfield, I spent a bit of time in grad school with people who are deep in the theory and practice of measuring inflation. It's really really hard, and it's the sort of place where there is a great public desire for a single number, but it's not clear that a single number is meaningful to cover all the most common use cases for inflation. (Use cases include: managing cost of living adjustments, understanding money supply, making investment decisions, making historical comparisons of other time series.) The many variant measures of inflation (core, non-core, etc.) exist to try to serve different use cases more effectively.

For folks with GPUs, note that HDBscan is very optimized in cuML (https://docs.rapids.ai/api/cuml/stable/api/#clustering / https://developer.nvidia.com/blog/faster-hdbscan-soft-cluste...).

Man, I want to appreciate a nice new hardware approach, but they say such BS that it is hard to read about them:

> “There might need to be a new term, because by the end of next year we’re going to deploy enough LPUs that compute-wise, it’s going to be the equivalent of all the hyperscalers combined,” he said. “We already have a non-trivial portion of that.”

Really? Does anyone seriously believe they are going to be the equivalent of all hyperscalers in compute next year? (Where Meta alone is at 1 million H100 equivalents.) In the same article where they say it's too hard for them to sell chips? And when they literally don't have a setup to even accept a credit card today?

Hercule Poirot (in "The Cornish Mystery"): "Hein, the pancreas is nothing. Of the digestive organs, the liver is the key. Look after the liver and life will take care of itself."

(I have cut my alcohol to 2 drinks per week to look after my liver - still looking forward to life taking care of itself!)

If your company makes $1bln per year now and pays a CEO $1 mln, and then someone comes along who has much better ideas, better work ethic, and more experience and you (the board) are convinced they can add 5% of revenue as pure profit to your company per year. They will add $50 mln in value per year. If that person asks $5 mln as compensation would it be responsible to say no?

Or to take a more concrete example, if Elon Musk decided to quit Tesla in a huff and focus entirely on X and stop dealing with the annoyance of public markets, the stock's would probably lose $75 bln in value overnight. If your legal goal is to maximize the company's value and it takes $500 mln pay per year to keep that from loss happening, it seems like it's probably a good deal to just pay the $500. As another example, Steve Jobs received hundreds of millions in stock options, and that was probably an excellent deal for Apple, though they probably could've hired someone cheaper.

I agree there are many cases where a more expensive CEO is actually worse. But I think it is very very easy to imagine cases where it makes a ton of sense to pay a large amount for an expensive CEO if they are the right person for the job and will add 10x more value than their pay.