matrix87

But if the abuser can be obligated to pay legal fees, there could be incentives for lawyers to seek these kinds of cases out

Corporations could just do the same thing. Henry Ford historically had doubled people's salary to get them to have families. This was contingent on a bunch of paternalistic rules and inspections (like participants weren't allowed to drink for example).

From an organizational perspective, really cunning managers understand this: if your reports have kids, they probably aren't going to job hop for more money or better work. It gives significant leverage over employees.

So I don't see why the public sector would need to do this, either the public or private sector could do it. If you lump the US government in with the corporations as two agents which punch down, life in the US could actually be seen as more oppressive than other places. It really depends. The US could accomplish a relatively totalitarian end result with the same form of government just by creating a couple strong economic incentives for corporations. Generally the US govt in its current form prefers delegation to the private sector

brute force scales a lot better than nuance

> I also have a lot of international friends and family that would happily find me an international bride who would never divorce me, and would be happy to marry somebody with a US passport. In fact this is a somewhat common offer, though of course they never say it in those words. You can decide for yourself whether that is abuse, but at least I wouldn't feel comfortable with it.

I'm pretty sure in this specific case, they get an instant green card if abuse happens. So the incentives point in the opposite direction (pro-divorce)

The shared tools that people actually need like git or ms word already exist (because they're things we actually need). Maybe claude code / codex / opencode is the rare exception of a new thing which will survive long term (the coding agent idea in general, maybe not these three products in particular).

This kind of internal tooling always creates a window for empire building. The few things I've done that actually mattered at the jobs I've been at have mostly been in excel, and I'm a swe to boot. People don't want another internal tool, they want signal that helps them make business decisions.

Decent tooling is nice, but really over the top tooling can indicate organizational problems like misallocation of resources

I wonder if all the cargo cult frontend JS people pivoted to doing this agentic stuff and that's why it's so much of a clusterfuck

Because empire building, why else. They had two other internal tools and now they have three

Apparently they used facebook to incite backlash against the Indian government to bully them into allowing their Internet.org initiative to continue there (essentially a stripped down, anti-net-neutrality version of the internet which only allows people to use facebook and a couple other sites).

This was mentioned in Careless People. It seems like they wanted to use facebook as an electoral tool pretty early on because it gives them leverage over politicians. They might not have been deliberately involved in CA, but this is probably exactly what they wanted

> but it's that Cambridge Analytica wasn't Facebook's fault (users willingly granted access), but it was their problem.

In Careless People, they mentioned that this was a strategy they wanted to employ:

They wanted facebook to become a tool for winning elections so that they would have more leverage over the politicians in power.

For example, they were trying to fight the Indian government over internet.org initiative (essentially shipping out a stripped down version of the internet to impoverished areas where they could only use facebook and a couple other apps, subsidized by telecoms). They used facebook features to incite backlash against the Indian government.

If I cut my leg off with a chainsaw, am I addicted to the chainsaw? I think what matters is the intent of the design. E.g. if they design the product to make you need more of it (while it has an actively detrimental effect)

But a substance is not a specific product or a corporation. Maybe you could call "infinite scroll" the substance, but then you're playing whack-a-mole with whatever new interface they cook up. But then we end up with a long list of "disallowed" software features which might not be inherently addictive

> It's hard to answer in general, but does it have to be answered in general?

Maybe not? Maybe if enough minors experience harm from a platform, that's enough to restrict it for minors.

What I meant to say was, the standard for when we call something addictive is more hazy than that. In general, text based interfaces don't light up the brain in the same way. Psychologically it's a lot easier to walk away from a chatbot vs an infinite feed of short form video... potentially that difference could help find some kind of standard to determine addictiveness

Edit: I wonder if a hypothetical digital experience like the samizdat out of infinite jest is possible? I.e. a digital experience which hits the brain in the perfect way to produce addiction without a chemical basis

Okay but suppose you have an AI chatbot which takes in context from previous chats on what your interests are. So people just keep talking with the chatbot more and more as it gains more context. This would meet the criteria. It's not necessarily addictive though because you can stop talking with the chatbot at any time.

Maybe it's something to do with the text-based interface vs short form video. Short form video is better at activating the lizard brain

This seems really messy. We regulate gambling because it's a specific thing, but how do you differentiate a non-addictive platform vs an addictive platform? (aside from this specific company... I mean in general). How do you "test" whether or not they're inherently addictive?

At least in this specific instance, they can look at the blast radius and make a determination. But this feels legally unsatisfying in the general case

> Heroin is addictive, nicotine is addictive. Light and sound are fundamentally different.

Gambling is also an addiction with a non-chemical basis

A lot of it is just fake work. There was already a bunch of fake work before AI. Now there's even more fake work. Organizations tend to produce more and more of it.

The general impression I get is that the real bottleneck isn't needing to write code faster, it's understanding what the technical problems are and having the freedom to do what is needed (which is sometimes risky I guess).

It doesn't seem like we're headed in this direction (from my n=1 perspective on the ground). It seems like we've just dialed up the fake work and fear culture. People are too busy and afraid to actually look around and figure out what the fuck is going on. There's just more bullshit fake work and it just keeps coming. And it's getting done faster, sure, but it was never tied to any specific economic outcome anyway and solely exists to waste people's time and empire build.

And there's not really an incentive to fix the problem either. In corporate cultures which try to be "lean" and are firing people all the time, management is incentivized to intentionally overstaff teams and keep people pushed away from critical initiatives so that they have sacrificial lambs for the next layoff. In cultures which have more fat, there's no real downward pressure anyway.

You can't force organizations to be efficient without cutting out the middlemen and taking the time to figure out what the fuck is actually happening. At large companies, this is already unsustainable anyway. Maybe this is what McKinsey used to do (and what AI is now going to help with): getting an omniscient view of what's happening organizationally. This would require leadership to do stuff that actually makes sense, which is why it won't happen

> When I’m not severely depressed I’m able to do two four hours blocks separated by a break in a day, and maybe squeeze out an extra hour or two on a fantastic day.

I do the same thing, it's always kind of puzzled me how people say they work 10-12 hour days. Realistically, it's probably nowhere close to 12 consecutive hours of deep focus. Sometimes it's easier for me to just take breaks and work whenever I have the motivation. Maybe it stretches into odd hours. Sometimes it's less than expected. Sometimes it's a waste of time because I didn't take the time to step back and gain more perspective. Hours never seemed like a good measurement.

I do think part of it is that when people are in really "high intensity" workplaces, there's social pressure to give the appearance of working hard. That could be part of it. I also think SWE (in SV business culture) tends to be more deliverable oriented as opposed to caring about the hours worked

Personally I'm pretty excited for the future.

I feel like LLMs make reading papers way easier than what it is before. If I ever run into a roadblock, I can just keep digging and probing until it makes sense to me. If anything, the amount of work that needs to be done has been amplified because it makes learning new things that much easier.

The business culture at F50 companies is several decades behind tech. The business culture in middle america is even further behind. If these people can remain employed, I'm not sure why anyone in SV is worried. The nature of white collar work is that it changes over time and frankly I think getting attached to one job title is nonsensical.

Also, half of the battle with white collar work is information asymmetry. The effectiveness of the models with real systems in the field is limited by the amount of context that people are willing to provide. This problem existed before AI, if teams document literally everything, management could see this as an outsourcing opportunity. If management forces them to document stuff, why would they do a good job? Also, why the hell would management even try to do this in the first place if it by implication eliminates their job as well?

There's a social problem here: how do you get people to give high quality information if there is no incentive? How do you know that they're not telling you things or intentionally explaining something badly (as opposed to the AI just being "bad" at something)? Especially now that the billionaire class has done enough saber rattling over it and gotten everyone thoroughly freaked out for the next decade at least? (This is also a problem in high churn corporate cultures as well, the fear fucks up the signal to noise ratio)

> There's a much more reasonable explanation here.. tech jobs had the most wage growth during COVID and this was a pullback.

More concretely, companies do layoffs, visa workers have to find another job in 60 days. Companies like amazon try to scavenge and lowball the shit out of them. This is happening across the industry.

There's consolidation, acquisitions, and layoffs happening across industry. It's just a big game of musical chairs

> There will be a window where fast adopting AI companies can deliver features faster than peers

The fundamental problem is having people doing the right things. Lines of code does not equal real economic value.

I think AI will have a much different effect: raw headcount will no longer give large companies the same moat that they once had.

If AI makes a 10x engineer into an 100x engineer and unlocks this whole world of productivity, getting them to actually deliver 100x requires the proper culture where they aren't blocked by environmental factors, management, and incentive structure.

10x -> 100x is an extreme example, but the point stands: if AI amplifies our productivity and productivity is determined by motivation + environment + alignment on priorities, in order to realize the advantage from AI, a company needs to both embrace it and have the right culture

We're in a big interconnected social system. If you're landing changes, it's dependent on the internal tooling, internal documentation, and the broader social system that you're a part of. Attachment to individual outcomes is the problem. The productivity comes from the integration.

Maybe the sense of alienation people are experiencing right now is because code is the one concrete thing they drive which is "independent" from this social matrix. Most of the real value I've ever delivered doing this SWE thing wasn't through code, it was through understanding the system, recognizing issues, and working with leadership on addressing them

> as far as I know, all the math breakthroughs via AI that I've heard about have come from Anthropic and OpenAI, not Chinese models

Maybe because dishonesty is more normalized in SV business culture

> Microsoft has to be shaking

Their stock hasn't really seen much upside from this whole AI thing, so I don't really see how they'd get punished that badly. At least they have a public cloud business where they can rent out capacity to reduce the loss.

Maybe if OpenAI goes bankrupt, they might be punished more. Or it could be an opportunity for them to scavenge. Oracle is a lot more fucked if that happens.

I don't think big tech is going to be hurt that much. Or if they hurt, it will be for unrelated business reasons like ad spend going down.

Maybe capacity spend will go down for the time being and the hardware stocks will go down. Maybe a bunch of these startups will go backrupt or get acquired. The defensive play of building out the infrastructure and software stack is probably going to continue. These companies have been playing catch up with nvidia for a while now and won't make that mistake twice.

It's kind of mind blowing how much work remains to be done with these models (e.g. finding ways to integrate them in the rest of the economy, finding more resource efficient ways to run them, ads monetization).

It seems like right now, a lot of stocks like oracle, adobe, IBM, and intuit are getting reevaluated.

the market is like a ponzi scheme, it depends on when you got in

Alphabet's business is in information retrieval, they're probably the only company out of all of them where the defensive AI play is justified

Don't join amazon, especially as a first job. The reason being, there is usually a shit ton of political stuff going on which you won't see as a new grad. The general way companies like amazon work, they try to bring in non-jaded people like you who have no clue what the fuck they're walking into.

For example, if your TL or manager doesn't like you and is deliberately setting you up to fail, you might not catch it early enough to be prepared. They won't tell you in the open either. They'll deliberately try to hide it and trick you into thinking that you're safe. Once you're in that position, they'll try to force you to work as much as possible and leaving is more difficult.

If you go the Cambridge route, you're getting domain knowledge which could let you get specialist roles (and likely a better support structure) at better companies. It's a no-brainer really. Amazon is desperate for practically any warm body they can get because no one with options wants to work there. If you're ever at a point in your career where you have to join amazon, it means that you played the game wrong and got blindsided by a layoff or something. (Ditto for all the other shitty amazon-like sweatshop companies, Elon companies, etc. Don't forget this: company reputation matters)

Imo this will be the future model: a company puts a gun to your head, gives you agents, and tells you to figure out how to maximize the utility output you provide as an employee. They might put a token limit which incentivizes efficient token usage.

There's a lot of "room" here for creativity, i.e. the agents don't know how to milk the agents efficiently, which is essentially your new job. The agents might not know how to structure tasks so that the agents run overnight and run the proper experiment.

The difference between mediocre and 10x will increase. The more aggressive SV performance culture is probably going to become more mainstream.

Frankly, I think this whole model is way more burnout inducing. You're not really doing software engineering anymore. You're running this mental loop in the background of trying to multiplex between agents and mapping unstructured business problems to agentic work.

It's probably going to infect other industries over time. Software is the first big wave. I don't think people with computer skills who are good at computers will be screwed out of work long term. The tech industry has a lot of concentrated technical capital in the form of tooling and automation that other parts of the economy could significantly benefit from. Short term it looks bad, but long term, computers will become more and more dominant over the economy.

I don't think companies have yet figured out how to adapt their hiring, managerial practices, or internal incentive structures to reflect this. Most of what they're doing now is extremely crude and terrible for morale.

My plan is to just stay in the game as long as possible and bag as much money as possible. Delay family formation (gen z) or buying property so that the money is liquid. Job hop as needed because of all of the consolidation and acquisitions that are happening right now.

If the whole industry implodes, I'll probably just go and get another degree or a PhD and try to pivot into an adjacent field.

But for right now, I think that staying in tech makes the most sense. The WSJ and these other bootlicker newspapers want people to be uneducated and unable to move abroad or escape the corporate hellscape this country is turning into. More money is more freedom from these managerial cunts who are stumbling over themselves to fuck up everything. I genuinely think that things will not positively change until the educated people start leaving the US en masse and shit up the whole US economy.

But I'm wondering if anyone here can chime in with smart ways to pivot where you can apply the same type of skillset that leads someone to do well in SWE. Is statistics a good route?