alphazard

It seems like the industry is going to be forced to rediscover and implement capability based security models and process isolation by consumer demand.

It is impossible for the average user to reason about a different security model for each app, the only way for anyone to be confident about what a program is not doing is to move to a world where apps don't work by default, and a list of boxes need to be checked which enable network or file access and cause features to work. Apple is the closest to the right answer here, but enabled-by-default and opt-out has to go away.

Entire roles have been created for people to do this, in exchange the headcount makes your manager look more important. In tech alone there exists: product managers, scrum masters, middle managers of all kinds.

If you want to stay in tech, look for roles that can be filled by someone who doesn't know how to build or sell the product. Every business has to deal with supply and demand, the further you are from those things, the more likely the job is bullshit, and not doing it will be unnoticeable.

The thing that always trips me up is the lack of isolation/sandboxing that all of the AI programming tools provide. I want to orchestrate a workforce of agents, but they can't be trusted not to run amok.

Does anyone have a better way to do this other than spinning up a cloud VM to run goose or claude or whatever poorly isolated agent tool?

Programming a GPU in 2025 is complex, that might be because it has been made complicated, but regardless, it is not complexity that this project can control.

The fact that it competes with PyTorch in so few lines speaks to the incredibly low incidental complexity imposed by Tinygrad.

> To fund the operation, we have a computer sales division that makes about $2M revenue a year.

What's the margin on that? Do 5 software engineers really subsist on the spread from moving $2M/yr in hardware?

Yeah this is exactly right, if you can trust the contributors to not code-golf or otherwise Goodhart the LoC metric, then it's a reasonable measure of complexity.

It doesn't work as well when you start mixing languages, or generating code.

> All societies are dysfunctional, great and small, because human beings are dysfunctional.

I don't agree with the 'because' part of this. It misses my point about why organizations are dysfunctional. Even if the organization was made up of perfectly rational, perfectly functional individuals, it would still be dysfunctional. The people running large corporations, are individually, very rational, and very functional. They are among the most capable humans at achieving their goals. Any explanation of organizational dysfunction has to also explain that data.

The dysfunction (which means actions not aligned with the shareholders) comes from the fact that 1. preferences cannot always be aggregated coherently and 2. that the people operating the corporation do not necessarily have incentives aligned with the shareholders. The first is a mathematical impossibility which cannot be fixed, and the second is a failure of mechanism design. That's the "open problem" I mentioned.

I don't know who this kind of cynic vs. idealist vs. optimist thinking works on/for, but it doesn't seem to give me any kind of conceptual edge. Instead of trying to frame things in terms of a mood or a feeling, it's better to try to understand things in terms of what is likely and unlikely to happen.

Large corporations are just groups of people with conflicting incentives, and that means they are basically incapable of performing certain kinds of tasks. It also means that when the incentives do align, some tasks are very likely to be completed, even with other corporations or governments working in opposition.

Some of those tasks might be things you care about, like making a product of a certain quality, or furthering some other goal you have. In all those cases, it is best to to first think about what is most likely to happen and what is unlikely to happen. You have to think of the organization as just another phenomenon that you could exploit if you properly understood it. Unfortunately, how to manipulate complex systems of humans is an open problem, and if anyone had effective, repeatable solutions, then investors would demand that they be implemented.

As it is, most corporations don't act in the interest of the investors a significant amount of the time, even though they are supposed to. The only thing we can reliably bet on is: all organizations tend towards dysfunctional bureaucracies, the longer they live, and the bigger they get.

It depends. Positive real interest rates do not necessarily mean deflation, and deflation isn't necessarily a bad thing.

As an example, you could give a loan for $1 to someone for 5% interest. In a year they pay you back, so now you have $1.05. That dollar could get you exactly the same amount (of real goods or services that you personally want) as last year, or it could get you more, or it could get you less. Inflation and deflation typically refer to the price of a basket of intrinsically valuable goods and services. That is separate from the interest rate which is just what you, the creditor, and the debtor shake hands over. If the dollar gets you the same basket as last year, then you are net better off because now you can buy the basket and you have $0.05 for lending to someone who was able to pay you back.

The missing variable here is the productivity of the rest of the economy, if the economy is growing, then you can see a decrease in dollars per basket (deflation), but that's not necessarily a bad thing. The interest rate is sort of like a best guess for the productivity of the debtor.

People want to be dollar debtors, not dollar creditors. When I said no one wants dollars, I was referring to people's willingness to hold actual dollars or obligations that pay them dollars in the future.

Your other comment mentions the AI bubble, and also makes me think you don't understand what I'm saying, since we seem to agree about what happens to dollars and debt in a bubble. Companies are glad to take dollars now in exchange for owing dollars in the future (something they would be less willing to do if the dollar was strong). They then turn around and spend those dollars on GPUs and electricity. They think they can get more done with a dollar this quarter by trading it to NVIDIA or a power company than by holding T bills.

Fed rates do not track the real demand to be a dollar creditor. That's kind of the point, the Fed is the lender of last resort. If no one wants to give dollars now for more later, then the Fed becomes a creditor to the treasury at an arbitrary rate.

This seems a little pedantic, but sure, no one wants to be owed debt denominated in dollars.

> Devalue against what is the main question though, isn't it? The real longer term issue is that the USD is devaluing against the Euro...

I don't think that FOREX rates are the best way to think about this, but if you work in that world or otherwise have an intuition for it, then go ahead. Most of us only handle 1 currency, and reasoning in terms of 2 isn't exactly an intuition pump.

Instead think about:

1. The dollar valued against itself a year earlier, and a year in the future. That is the interest rate or yield of the asset if held. It should have a positive real yield, but right now it doesn't.

2. How much your personal basket of monthly expenses costs in terms of dollars. Ignore a basket that someone on the news told you to care about, like CPI. I mean your personal basket, all the stuff you personally buy, how much is it in dollars, now, a year in the future, a year earlier.

If you stored value in business or a precious metals in the last year and then converted back, you would probably have more dollars, or be able to buy more stuff, that's all there is to it.

Dollars are currently only in demand for short-term use in transactions. Most of the world still relies on dollars for transactions, because that is what all the banking and payment infrastructure uses.

But no one wants to hold them because they devalue and will continue to do so at an accelerating rate. It's a game of hot potato where everyone is forced to hold equities, commodities or other assets by default in order to preserve their wealth, and then convert to dollars to transact. The days of savings accounts are over, and everyone should think of their checking account as something that they pay negative real interest on for the privilege of being able to transact with the rest of the world.

Meanwhile, the big players in the current financial system are trying to figure out how to continue playing the current game without resetting everyone's progress. They don't want to lose their hard won position to pay for bad decisions by American voters. It's a coordination problem, and the shelling point looks like it is still gold, same as it has been for thousands of years.

It's a fundamentally bad idea to have a single key that applications are supposed to look for in a particular place, and then use to sign things. There is inherent complexity involved in making multi-context key use safe, and it's better to just avoid it architecturally.

Keys (even quantum safe) are small enough that having one per application is not a problem at all. If an application needs multi-context, they can handle it themselves. If they do it badly, the damage is contained to that application. If someone really wants to make an application that just signs keys for other applications to say "this is John Smith's key for git" and "this is John Smith's key for email" then they could do that. Such an application would not need to concern itself with permissions for other applications calling into it. The user could just copy and paste public keys, or fingerprints when they want to attest to their identity in a specific application.

The keyring circus (which is how GPG most commonly intrudes into my life) is crazy too. All these applications insist on connecting to some kind of GPG keyring instead of just writing the secrets to the filesystem in their own local storage. The disk is fully encrypted, and applications should be isolated from one another. Nothing is really being accomplished by requiring the complexity of yet another program to "extra encrypt" things before writing them to disk.

I'm sure these bad ideas come from the busy work invented in corporate "security" circles, which invent complexity to keep people employed without any regard for an actual threat model.

> Why do you say it's rent seeking?

Because that's what economists call it when you get something for nothing, as is the case with any subsidy. I'm not going to argue this point; interested readers can look up how these energy projects are financed. Windmills that are privately funded, including debt and risk show you where it actually makes sense to put a windmill.

> Offshore wind is efficient, turbine blades can safely be much larger giving 3x the output, turbine arrays have unobstructed space giving twice the capacity factor. It's more efficient than onshore.

Not going to argue with any of this, although you left out maintenance costs, and larger blades means more value at risk. I'm not convinced that your efficiency calculation is measured in dollars and not windmill hours.

I would caution any engineer types reading from pressing their nose too close to the details of a particular energy technology. Instead, it's better to focus on the business plan or economic shadow that a particular energy project leaves. Dollars go in and energy comes out. A bunch of money has to go in up front, then trickles of money slowly over time, and occasionally spikes of money have to go in randomly. In exchange there is a modest, predictable flow of money out, which eventually is larger than all the in-flows in the bull case. The question to ask is: how much in and out of dollars and of Joules at each point in time? How does that compare to hamsters on wheels, people on bicycles, and lighting things on fire?

> You appear to be starting from a premise that wind turbines don't generate profits?

This was never a stated premise, and my post starts with the opposite sentiment.

Bringing up a map of wind power deployments tells the story; what you will see is a hot vertical strip in the center of the US. That is where it actually makes sense to deploy windmills, and people will continue to put them there even if subsidies end. It makes sense for the area, the amount of wind, the serviceability of the deployments, etc.

Off shore has always been politically contentious because it's much more dependent on subsidies, it's a battle for/against rent-seeking. One party is in favor of this particular kind of rent-seeking and the other party isn't (they will be in favor of a different kind, no doubt). The subsidies are necessary for these deployments to make financial sense, and if they went away, then it would just be a bad place to put a windmill.

There is no national security issue, there is no real case for energy infrastructure either. This use case needs government money to make sense, and is therefore sensitive to political fluctuations.

> when people trust and genuinely support each other it's not zero sum

This is a clear non-sequitur. Whether or not the situation is zero-sum is dictated by the situation not by the actions of people, or how they feel about each other.

If the situation is zero-sum, and one party doesn't realize, or is too trusting of other parties, then they just lose. And they might not even realize that they lost, or that the situation was zero-sum to begin with.

> basically at the same level in terms of power and influence?

Obviously not true, unless you can truly ignore your manager and still be successful at the company, then they have power over you. If they control more resources than you, or have more face time with people who do control resources, then they probably have more influence than you as well.

This whole comment reads like a sucker who doesn't know they are being scammed.

> but you can establish trust with managers.

This is technically true, but metaphorically false. Most readers, especially the younger crowd, are likely to be too trusting and not correctly recognize fundamentally adversarial relationships, like with their boss, with HR, with lawyers whom they don't pay personally, etc. Especially when those people present the relationship with a faux version of the same naivety that they intend to prey upon.

The only people you can trust at a company, are people who have demonstrated that they value your relationship more than the company's relationship. That's not impossible, but if they really need the salary, then the odds are stacked against you.

It's certainly nice to pretend that everyone is nice and trustworthy. In fact, most people will hold it against you if you don't sufficiently pretend to be trusting. Universal insincerity is part of the human condition.

It's worth mentioning that career conversations with your manager are always bullshit. How could they not be? Your manager's incentives are to keep you around, keep you productive, and keep you inexpensive. It's ridiculous to think that helping you build a successful career, which likely doesn't involve them or your current employer for very long, is something that they would do. Why would you reveal any information about your long term plans to such an adversary?

If they realize that the company cannot provide what you are looking for then they may not want to invest more resources in keeping you around, or keeping you happy. If they realize that you find some sort of intrinsic reward in certain work, then they might put you last for raises, because the money could help them more when spent on other people.

It's best to selectively reveal only your immediate short-term goals, only at the current company, and only as part of an ask. Always make them pay for information, if you are going to reveal what you want, then they need to reveal whether they intend to help you or hinder you in getting it. Slow answers, non-answers, pushing to next quarter, etc. All signal that they intend to hinder. It's rare to get an honest, fast "no".

This exactly. You have to be honest about why you are building something. If the answer is that you actually want to use it, then yes, quality and maintainability are important. It might even be a good idea to use no AI whatsoever.

But if you are building it because doing so is in the long chain of cause and effect that leads to you being fed and having shelter, then you should minimize the amount of your time that is required to produce that end result. Do you get better food, and better shelter if the software is better? It would certainly be nice if that was the case, but it's not.

> Not everything is about money.

Except for your job, which is primarily about money. Making it take less time, means that you have more time to focus on things that really are not about money.

Quality is not rewarded at most companies, it's not going to turn into more money, it might turn into less work later, but in all likelihood, the author won't be around to reap the benefits of less work later because they will have moved onto another company.

On the contrary, since more effort doesn't yield more money, but less effort can yield the same money, the strategy is to contract the time spent on work to the smallest amount, LLMs are currently the best way to do that.

I don't see why this has to be framed as a bad thing. Why should anyone care about the quality of software that they don't use? If you wouldn't work on it unless you were paid to, and you can leave if and when it becomes a problem, then why spend mental energy writing even a single line?

Certainly it's more conspiratorial to assume that his death had something to do with his research, or that he was secretly a some kind of Walter White character?

Being politically outspoken on an issue which is contentious in that area, and which has caused violence before seems like the most plausible explanation that I have heard so far.

"User" generated content on the internet is mostly bots, HN included. Opinions that seem too radical or stupid to be believed are often bots, or NPC humans repeating bot content that they read somewhere else.

IMO the hard problem here is PoS consensus with the private transactions. It seems like the stakers have to come up from the depths of privacy to participate in consensus. Maybe there is a way to do private staking, but that makes the network very difficult to understand and build confidence in. So I don't see upgrading to faster consensus to be a small incremental improvement, it's fundamental.

A separate issue is that both Monero and ZCash are not post-quantum secure, while many of the new zkSTARK VMs are. The ledger lives forever, and state actors will eventually decrypt the transactions if the cryptography can be broken. At that point it seems like it's better just to build the currency product in one of the zk VMs.

I think this is a huge misconception. I certainly don't view this issue as grounded in 'rights' or what governments should or should not do.

It's entirely an issue of what people can and cannot do with this technology. It's a game with sides, and the concern should be that the technology has made playing for one side much easier than playing for the other.

The technology has unlocked total freedom of association, and I don't see a way to reign that in, other than restricting access to computation and the network for the entire population. As long as the average voter wants personal computing to continue, I don't see a way that a government could get the necessary control to shutdown one of these systems.

> Ok, if that's really your thinking then you need to lay out: here's an impossible-to-ignore thing we can do with this, and this is how, and this is why this wouldn't be possible without this thing.

These technologies can be used by people to coordinate amongst themselves whether the outgroup likes it or not.

If you google "network state" you will find things that you might like, and things that you might not like. It's not up to you whether other people create these things. You can only control your own participation.

Cryptography is really the study of incredibly rigged games, games that one side almost always wins, even when both players play perfectly. If human society is a game where humans try to coordinate with other humans to be better off, sometimes at the expense of other humans, then distributed ledgers have totally changed the meta.

> no one has money and everyone cares and is tech savvy and yet no one can be trusted at the same time

lol, you perfectly described a group of college students.

> The problems with the technologies is that the "robust database" guarantee is often highly dependent on the currency mechanisms. They unfortunately go hand-in-hand.

I agree that economic incentives are important for robustness, but I don't agree with this use of the word "currency". The tokens have value in that they can be used to write to the ledger. That is the consumer aspect of the token, you can give it up in exchange for the ability to edit the ledger. The producer aspect of the token is that if you help participate in running the network, you can earn tokens, to edit the ledger yourself later. The tokens have to store value (ledger write access), and there needs to be a market for them, because the producers can't use them all themselves.

Just because something can store value, in this case the specific ability to write to the ledger, doesn't mean that thing is suitable as a currency. Copper ingots let you make cat6 cables, the are objectively valuable, but we don't use them as a currency. And as the world found out over the last decade, the distance from store of value to functioning currency is significant.

The store of value is sufficient to run the networks though, you don't need the tokens to work as a currency, and I think that has been empirically proven. The Ethereum mainnet is unlikely to disappear for a very long time, but Ethereum is also unlikely to ever be widely used as a currency.

The takeaway is that you should have as much of these tokens as you are likely to need for writing to the ledger. If you hold them in Coinbase and never use them to operate the ledger then you are participating in the speculative overvaluing of the tokens.

No one is going to run a database for you unless you pay them. If you want to write data to an existing public ledger, then you are going to have to get some of the token that allows you to write to that ledger. Paying for a token to use the ledger is not speculative, it's pragmatic. Holding the token and not using it to write to the ledger is speculative, and encouraging people to do that in order to make money yourself is a scam.

If you pay AWS to host Postgres for you, are you being scammed? You might be if someone convinces you to invest in Postgres credits that you never use. But if you actually use Postgres, and it's cheap, and you are glad to pay for it, then it's not really a scam.

I think more of the promise of these technologies is in private ledgers, where the traffic is confined and the system could even be run altruistically. For example, a university club or a town could coordinate elections using this technology. They don't have to pay to use a public ledger if enough people in the group can run nodes themselves. Everyone who cares about the result of the election is incentivized to participate in the network just to know the result and ensure its authenticity. They aren't processing transactions for random users, they are coordinating using the latest technology with people who they want to coordinate with.

It's good to know everyone here is weary of crypto scams, but I don't see anyone accurately describing the significance of these technologies.

Bitcoin failed as a currency, and as that became realized, institutional investors pivoted to the "digital gold" scam, to keep people long, while they divest or hedge. The two reasons why it failed as a currency are transaction latency, and lack of fungibility. Transaction privacy is necessary for fungibility. Both of those are just technical problems; I predict that a distributed ledger currency with private transactions like Monero, but a faster consensus algorithm like Avalanche or Hedera will become popular in the next decade. It's likely to be an Ethereum L2.

That is just the currency aspect of distributed ledgers. It's just one use case that we don't yet have the technology to properly address. The exciting thing that distributed ledgers enable is cryptographic institutions. These technologies allow us to solve coordination problems more easily than ever before. Democracies, businesses, communities, projects can all be coordinated better and more honestly using distributed ledgers. It's not an overstatement to say that distributed ledgers are as big of an advancement for human coordination as democracy was.

If you've been soured on these technologies because most of the currencies built with them are scams, I would encourage you to learn about them as if they were just incredibly robust databases that even governments would struggle to take down. Surely you can think of something cool to build with that, which doesn't involve money.