rdgthree

xAI had a lot of negotiating power here because Anthropic had ~0 comparable options and ultimately desperately needed the compute now. So, it wouldn't surprise me if data sharing was an explicit part of the agreement

Nothing to add to this conversation in particular, but just wanted to say - truly amazing paper. Well done!

Should have added one sooner! Here you go: https://1393.xyz/rss.xml

I appreciate that you feel this way, but the mechanisms behind exactly which neural circuits are activated by TMS are simply not yet fully understood.

From 2024:

> Transcranial magnetic stimulation (TMS) is a non-invasive, FDA-cleared treatment for neuropsychiatric disorders with broad potential for new applications, but the neural circuits that are engaged during TMS are still poorly understood.

[0]https://journals.plos.org/ploscompbiol/article?id=10.1371%2F...

Of course!

But also:

> Although the biology of why TMS works isn't completely understood, the stimulation appears to affect how the brain is working.

https://www.mayoclinic.org/tests-procedures/transcranial-mag...

I think it's reasonable to assume there's room to sharpen our understanding of it quite a bit.

The whole argument hinges on the idea of tuned resonance: https://www.nature.com/articles/s41598-025-07988-2

If you're interested in my personal chain-of-thought on the subject:

This was where I started pulling this thread (October 2025): https://1393.xyz/writing/could-the-root-cause-of-alzheimers-...

And this is an even further ancestor of the ideas (December 2023): https://1393.xyz/writing/are-we-only-conscious-while-were-le...

I'm operating off of my own subjective experience, and this idea lines up tightly with System 1 and System 2 in cognitive psychology.

It seems that many jump to "AI psychosis" when one mentions magnetic fields, but the evolutionary tree is very straightforward:

1. Nature evolves magnetoreception for navigation

2. Eventually, a brain in nature with magnetoreception accidentally "hears" its own magnetic field with with resonance

3. That lossy global summary of the brains ends up being an evolutionarily advantageous "higher-order sense"

4. Evolution sharpens the blade for many years

On first principles, that seems perfectly viable and even likely given that magnetoreception was such a boon for survival for all life.

Just glad others are finding it interesting!

Not necessarily - I think it works like Daniel Kahneman's System 1 and System 2. Your conscious system is System 2 - when it's not working correctly, you just fall back to System 1.

Independently, since the whole idea relies on resonance, it may be the case that an fMRI doesn't actually interfere with the "stochastic resonance" mechanic quite like TMS (transcranial magnetic simulation) seems to.

If you model the brain this way, dementia looks like a clear breakdown of System 2, which is an interesting thought experiment even if the mechanics aren't perfect: https://1393.xyz/writing/alzheimers-is-the-symptom-not-the-p...

In cognitive psychology there's all sorts of evidence that we have two distinct processes, but I don't think anyone has really mapped it to a physical system yet.

Modeling two physical systems is pretty interesting though because dementia ends up looking like a clear failure of System 2. Really neat idea generator even if imperfect.

I think Daniel Kahneman's System 1 (habits, unconscious) and System 2 (learning, "error correction", conscious) are physical systems, and System 2 takes a LOT more energy to run.

So, when you get tired, System 2 leans more and more on the much more energy efficient System 1. So you get behaviors that look like unrestrained habits: poor impulse control, lowered emotional regulation, etc

Edit: I wrote more about this idea if anyone is curious: https://1393.xyz/writing/alzheimers-is-the-symptom-not-the-p...

Did you read it?

These have always felt like symptoms of the problem to me, so perhaps just downstream effects! Definitely not sure, obviously plenty of complexity to this one.

Interesting. Yeah, fear being a component of these things makes sense to me from an evolutionary perspective. Self-harm makes sense if your arm is pinned under a boulder. I think the psychology of these conditions is closer to that than anything else.

In my experience, I've found most deeply embedded self-destructive ideas to be rooted in fear.

The brain and body are effectively locked in a paranoid "fight" state - "I know this behavior is unhealthy, but if I don't behave this way, then [insert irrational fear] will happen."

If you can figure out what she is afraid of, you might be able to help her walk back the behaviors from there. But trying to correct the behaviors on their own is very hard. Since the fear is usually irrational, the behaviors may be a relatively rational response to said fear. They may hold the fear itself as table stakes - not understanding that they are afraid of something that isn't real (i.e. I will never find love if I'm not as skinny as the other girls on Instagram).

Again IME - the feeling from inside the self-destructive brain is "I shouldn't be doing this self-destructive behavior, but I don't know another way to achieve X goal, and I absolutely need to achieve X goal." Asking the question "Why do you need so badly to achieve X goal?" is usually a good start to defanging the fear.

FWIW the comment you are responding to was authored by AI.

You aren't too old. I've been out of school for ~10 years and AI makes me feel exactly the way you've described it. I'm trying to navigate it and find an exciting space to carve out for myself, but the future feels unclear in a way it never has before. I'm confident in my own ability to build great things, but to your many points, AI has the unique potential to do everything better than me.

I don't want to say this makes me feel hopeless, because it genuinely doesn't. I'm working on a new startup, I'm starting a family, and I'm doing well.

I just want to build things that matter, that people find useful, and I've been having a hard time figuring out how I'll do that in a world with sufficiently advanced AI. Hobbies have never been enough for me, and never will be. It needs to be more than fun to energize me - I need to feel like I'm moving a needle.

For now, all I know is that I don't know. It feels nearly impossible to guess what the future holds.

Doing this is pretty critical for SEO in a situation like this, and these businesses tend to live and die by search traffic.

An interesting question that's still TBD is how well companies can operate with a significant in-office team and remote team. I think this past 18 months has proven that fully remote can work, but it may be the case that a 50/50 remote/office split is significantly harder to manage.

Anecdotally, my company had an office team and a remote team split about 50/50 before and unrelated to COVID, and it was tricky. We don't have a lot of meetings, but when we did, it was hard to have a physical group in a conference room with a remote group on a screen. We solved that by effectively going fully remote for meetings that involved remote employees - everyone who was physically in the office would connect to a video call at their individual desks. I was personally fine with it, but some people definitely prefer (and work better in) that physical conference room setting.

All of that to say, the way one person works might be a forcing factor for others, even if they don't intend for it to be. Tricky problem.

(Caretaker cofounder.)

I've been rejected/failed/banned by an emotionless machine a handful of times, and I don't know that anything else has made me feel quite so hopeless. I like to think a similar knot in my stomach keeps me honest.

Thankfully, one of our fundamental incentives as a property management business is to get more quality tenants approved for more apartments. If there are high quality renters qualified to sign a lease and fill a vacancy, mistakenly rejecting them directly impacts a landlord's bottom line. So we're motivated beyond altruism to get this right, which is important.

Along these lines (and perhaps surprisingly, relative to the automation in the post), our income verification product is decidely not fully automated. Non-salaried income reporting can be extremely tricky, and we've run into a number of renters with reliable income on a monthly basis that doesn't fit neatly onto a bi-weekly paystub. In those circumstances, we work with them manually to sort out how we can best present it to landlords on an application.

Well said. It was frustrating how little he talked about the actual problems with the codebase, instead choosing to handwave and blame it all on React. The dig on the React team in particular was bizarre:

> React’s team enjoy experimenting with new ideas, but this is killing the ecosystem! They should be brave and take the blame for it!

Have to make a lot of assumptions with this sort of thing, but I can't imagine that attitude results in high quality engineering work.

Is this just an assumption? It's an interesting thought, I'm curious if there's some sort of data to back it up.

Can confirm this is from Kapps, and I believe it was added recently. Not sure how he does it, but he streams on Twitch[0] and you can probably just ask him in chat if you catch him when he's live. He's live right now actually (7:15AM EST).

[0]https://www.twitch.tv/kutu182

Neat, iRacing on HN is a collision of worlds for me. Outside of software, I'm a huge racing fan and spend a fair bit of time on iRacing (mostly oval). I've been wanting to build something related but haven't come up with any ideas that I love yet.

Cool to see I'm not the only one!

Though I'm sure you know this, it's worth noting for those unfamiliar that it started from a NASCAR game - NASCAR Racing 2003 Season[0]. A lot of the quirks are related to oval-racing based assumptions for that reason. A lot of the severely lacking (or particularly rough) areas are things that exclusively exist in road racing, like non-solo-driver events.

[0]https://en.wikipedia.org/wiki/NASCAR_Racing_2003_Season

The title is technically accurate, but I think the omission of scale and any sense of how far along these plans are leaves quite a sensationalist tidbit. "Google workers announce plans to unionize" translated to "the necessary amount of employees at Google to form a union are unionizing" on my first pass. I clicked immediately because that seemed significant. I'm sure this was the intention, though I can only speculate. A clear and less sensationalist (though less click-worthy) title could have been as simple as: "230 Google workers announce plans to unionize"

That being said, I'm not contesting the validity of the movement - it's certainly possible that thousands of Googlers will sign on in support now that the movement is public, and more power to them!

It just seems like the reporting on this should be making it more clear where this effort stands and just how much needs to happen before it's legally viable. Arguably, more honest reporting in that regard would help make clear to potential allies that their support is needed, and this is not a sure thing.

Yeah, they definitely need to gather support of some sort, I'm just curious about the scale. 40,000 people from within the 120,000+ organization seems huge. If that's really the case, the coverage so far would seem fairly sensationalist - 0.25% of employees signing on to unionize is a drop in the bucket next to a required minimum 30% of the workforce signed on.

It seems incredibly significant that only 230 people are officially involved with these plans to unionize. As of 2019, Google had nearly 120,000 employees.[0] That seems quite small, relatively speaking.

Legally (I think, just learning about this now), to form a union, a majority of workers must show their willingness to form a union. Alternatively, to choose an existing union to join, an election with 30% of the workers support is required.[1]

So with that said, am I reading this right? Does this group of 230 people need to find at least ~40,000 more people for this to be a valid effort to form/join a legal union?

[0]https://www.cnbc.com/2020/01/02/google-employee-growth-2001-...

[1]https://www.legalmatch.com/law-library/article/how-to-form-a...

This is an interesting topic for me because I hold a handful of domain names (10-20) that I'm not using and have often had this internal debate.

I have them because they're ideal names for a few side projects I've been considering building now or in the future, and I hold on to them because I want to make sure I have them if inspiration hits for any of the given projects.

I'm sure many of you own some "aspirational" domain names like this, and while I realize it's a much different scale, why is it functionally different than the OP? Is there a reasonable quantity of unused domain names for an individual to own before it becomes unethical?

I wouldn't want to hold a domain from someone else who wanted it in good faith and was going to use it (like I hope to, eventually), but if I want a good domain name for a product I'm building in the future, I have to play the game a bit ahead of time.

I think the argument here is essentially that of the prisoner's dilemma. A system or game can put two rational (or we can say ethical for the purposes of this argument) actors in a position where the most rational/ethical decision isn't the most ideal for all parties.

You can argue that the individual acting in a self-interested way is the problem, or you can argue that the system itself is the problem.

I'd have to agree with the parent comment - it's complicated.

This is an argument that I've found myself making fairly regularly, and often frustratingly after attempting to hunt down a bug through 10 levels of abstraction that didn't need to be there. Add "debugging complexity" when it's warranted, but don't add it by default. More files and functions mean more jumping around and more potential blind spots.

That being said, John Carmack makes the very important point that this decision (whether to inline or not) should be made case-by-case. Someone in these comments used the flow of a novel as an example and I think it's actually the ideal way to describe the balance you want to strike with code as well. Using "The Tortoise and the Hare" as an example:

Inlined too much: "The race began. The Tortoise took a step. The Hare took a step. The Hare took another step. The Hare took another step. The Hare took another step..."

Inlined too little: "The race began. The Tortoise won!"

It's interesting because some authors actually make mistakes in this realm - overdoing it on details, or leaving the reader confused without enough context. Thinking about the flow of code like the flow of a novel is probably a good idea - both should delicately balance complexity with readability. The best examples of both often describe complex and nuanced concepts while remaining surprisingly straightforward.

If you think this example is contrived and it's always obvious where to abstract and where to inline, it's likely you're over-abstracting. The question of "where to cut" varies greatly from one bit of code to the next and often needs some thought to get right. John's list under "To sum up" here is a great set of guidelines to answer that question.