skybrian

This is getting causation backwards. Effective Altruism was a relatively small scene that attracted some people who understood the importance of AI early. It’s not all that surprising that some of those people went on to start AI labs later.

You can make a conspiracy out of it, but it’s sort of like thinking people who met in school had it all planned out from the beginning.

They buy a relatively small number of eggs at a high price, but the price of the transaction is an input into an index that the larger market uses.

The numerator on that is “Real Value-Added Output for All Workers” [1], which is based on revenue (subtracting some costs), and the denominator is “Hours Worked for All Workers” [2].

Anthropic’s revenue has seen enormous growth [3]. They have few employees. Obviously, as these things are measured, they are highly productive! It might be hard to reproduce that sort of productivity outside the tech industry, though. We can’t all be Anthropic.

[1] https://fred.stlouisfed.org/series/OUTNFB

[2] https://fred.stlouisfed.org/series/HOANBS

[3] https://pages.skybrian.com/2026/anthropic-run-rates/

A report doesn’t need to be discontinued if it’s automated. It’s just another graph on a dashboard. The cron job might put a little more load on the system, or the code might be a bit of a software maintenance burden, but cleaning it up can be put off for years. It’s there if you want it.

Number of CVEs fixed at an all-time high:

https://pages.skybrian.com/2026/kernel-cve-fixes/

It seems fairer to say that the Waymo fleet has >4,000 driverless cars [1]. There is still a long way to go, but it's rising.

[1] https://pages.skybrian.com/2026/waymo-robotaxi-fleet/

That analogy doesn't make sense. It's more like you drew up the plans for house, someone else built it, but you're refusing to pay the developer for the part they did, because somehow, only drawing up the plans counts.

Academic researchers don't get a vaccine approved and manufacture it at scale. They do experiments and publish papers.

For me, the visual design usually has problems the first time, and then I ask it to fix it. But the edits are still done by the AI.

Based on my experience using AI to build software for myself, I think this mostly depends on what you ask for, not whether you use AI to implement it.

How much electric sports car owners care about increasing range beyond 300 miles or so? It seems hard to believe that such a car would be obsolete.

The headline, at least, is making an comparison between shareholder benefits and public costs.

Why that particular combination? You could just as easily compare private spending and public benefits, and it would make about as much sense.

Maybe better to include both public and private costs and benefits?

How would this study’s methodology account for the public benefit from a new vaccine?

Edit: according to ChatGPT, it doesn’t:

> the analysis considers resources such as research funding, scientists' labor, data collection, analysis, and manuscript preparation. It does not assign an economic value to lives saved, illness prevented, or economic activity preserved by a vaccine

This doesn’t seem like a real problem for companies since they could start up internship programs if they need to.

Automating away the tedious parts of your work is something sysadmins and programmers have been doing for a long time. For example see:

https://sre.google/sre-book/eliminating-toil/

It seems like the sort of thing that system designers could benefit from understanding, but the ordinary users of a system shouldn't need to know about it.

I don't think holding gold does much, but in theory, a central bank could prop up its currency by selling the gold. It's like selling any other asset.

> Since 911’s inception in 1968, and even before, we have always defaulted to the police for things that really have nothing to do with crime.

I'm not sure why that should be surprising. For example, I'd expect the highway patrol to show up first at an accident. Maybe an ambulance or the fire department would be there too.

Why bring up the singularity? It hasn't happened and might not happen the way anyone thinks. For example, the timelines in different fields might not all go vertical at the same time. So, you could get something like an Internet singularity without a corresponding robotics singularity. Also, war is different from civilian applications. The use of drones to bomb data centers doesn't imply that drones will be useful for civilian applications like delivering pizza.

There's no particular reason to believe that solving math theorems generalizes to engineering or robotics without a whole lot of additional work.

That seems wrong unless you have an unusual definition of "understanding":

* Lots of people use math for calculations without having proven any theorems.

* You can understand a mathematical statement without understanding or even having a proof. There are proofs built on the axiom of choice and conjectured results.

No, it's not an "everyone" problem yet. Each field is likely on its own timeline. Knowing how well AI is at proving theorems is of only limited use for predicting what will happen and when in other fields.

Maybe math is more like a science now, though? It's as if ancient astronomers saw the results from a modern telescope. "Why even look at the sky anymore," they might ask. "And we don't even own the telescope!"

Maybe the answer is to look into how Big Science works? There's still plenty of work to be done figuring out what to do with the outputs from these new math-observing instruments.

You use the word "already" as if it already happened, but this is confusing speculation with reality. Market prices are a summary of speculation. Plenty of things are "priced in" that haven't happened yet, and maybe won't if the market is wrong.

Anticipating that something will happen is not the same as evidence that it's happened.

I don’t see anything vague or misleading about either tweet. I suppose it might be misleading if you don’t know what “bots” means in this context and don’t bother to find out?

How often are you doing calculations though? Using the same names as everyone else seems more important.

I think you’d be better off with something compatible with an existing system. Perhaps start with nashville numbering and add accidentals?

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

Or maybe it’s the Iranians this time? Whoever it is, getting good at cybersecurity seems more useful than going to war over it.

Okay, paraphrasing text describing an idea isn’t necessarily a copyright violation.

I think the issue is that it’s case-specific, but there are plenty of people claiming copyright violations in circumstances where there’s no specific evidence for it, and a few other people claiming that LLM inference necessarily means it’s transformative use.

It can have good uses, but we are also going to have AI botnets soon.

Ideas can’t be copyrighted. Paraphrasing some text explaining an idea isn’t a copyright violation. Having trained on something you wrote doesn’t necessarily imply a copyright violation. It depends on what the LLM outputs.

If Big-O lets you run every possible algorithm in "parallel" while still being in some sense "optimal," that seems like yet another exploit?