jeffreyrogers

You can convert your 401k to an IRA when you leave an employer. Some employers also offer in service rollovers (I think these mostly have minimum age restrictions on them though)

Roth contributions are withdrawable without penalty. Also most employers offer a match of some amount, which is essentially free money.

I can flag a lot of content that does pretty well online as AI. There are certain rhythmic patterns in the sentence structure and the way arguments are set up that give it away. The writing is obviously "successful" in that it gets a lot of engagement, and many people either don't realize or don't care that it's machine generated, but I find it distasteful.

If I'm reading something I either want to be entertained or I want some specific information. If it's to be entertained then the story has to be good (writing doesn't necessarily need to be that good) and I haven't seen any AI generated stories that are that compelling. If it's informational then I can either ask an LLM directly and get something tailored for me, or I'm also interested in the emotional and human side, which an LLM can only imitate, since it has no theory of mind.

Jim Simons (mathematician turned billionaire trader) was a chainsmoker and claimed that he had his genome analyzed and found he was unlikely to get cancer. I always wondered whether he'd really done that or whether it was just a flippant response to an interviewer's question about his smoking habits.

It's ironic that the article that caused him to write this post (https://substack.com/@justindeschamps/note/c-261265212?utm_s...) is written by AI.

I see this style of informational writing all over the place now.

I like using LLMs and one of the main things I'm working on would be impossible for me to tackle without them, but I do think they cause some skills to atrophy despite being a big productivity booster overall.

I think some people are drawn to technical careers because they like details and others are drawn to them because they like systems (now people are drawn for money too, but I'm talking about people who are more technically oriented). My theory is that if you are a systems person you find using LLMs fun and fulfilling and if you are a details person you find that they are the opposite. I have an acquaintance who is fairly well known for some of the work he did in computer vision (pre-neural net/deep learning), he's a details guy, and he was telling me how software is boring now because when he was young they were still figuring out how to do tons of things for the first time and now most software is just putting things together that we already know how to do. LLMs seem to be continuing this trend.

In college I used mathematica a lot while taking linear algebra. I ended up having to relearn a lot of that math later since I never really understood it at a deep level, although I was able to use it and apply it under class conditions. I feel like a lot of my LLM derived knowledge is similarly superficial.

I would bet that how your brain stores information that you read from long-form text is very different from how it stores information you acquire from chatting with an LLM. When I read something challenging or new to me I spend a lot of time thinking about how what I'm reading matches my own experiences or knowledge. Although I'm a fairly fast reader, it often takes me a long time to get through difficult pages since I have to stop and think about what I'm reading. I seem to be doing a lot of integrating and reorganizing my thoughts. When interacting with LLMs it feels a lot more like I'm just receiving knowledge passively and I don't think it gets integrated as well. Not sure why this is and its somewhat counterintuitive since I don't think I'd have the same experience with a human tutor.

Fable works very well for me on a moderately large codebase. I have had to correct it a few times or point it on the right track, but given how much faster it is at programming than I am that's a very minor issue (and most of these errors are because I underspecified what I wanted in the prompt, I can only think of two cases where it was genuinely wrong... that's a lot better than me in my professional career). Code quality is equal to what I would come up with (and better in areas I'm not familiar with) and the overall software engineering bar is higher because it doesn't get bored when I tell it to do refactors or write integration/regression tests that I would otherwise put off. Also makes it easy to audit code for things like missing audit logging or error notifications that a human would get bored doing.

The product is a fairly standard Ruby on Rails webapp with postgres as the DB. Application complexity is probably a bit higher than average for a webapp. So it's nothing that pushes the boundaries of software engineering, but it is a real product. Token budget has not been an issue for me. I pay for the Max plan ($200/month) and it is well worth it.

> Also, big companies can choose to run their own models on their own hardware and get better security and privacy as the data doesn't need to leave their own premises.

Yes, and then they would be reinventing the company owned data center that most big companies have just spent over a decade moving away from. I don't think companies will do that when there are multiple vendors competing to provide that service at what are quite reasonable prices when you consider what paying a human for similar output would cost.

That's not really my argument. It's that companies seem happy to pay a premium for a large company to provide complicated software services to them even when there are cheaper competitors.

> free and low-end eventually wins

Not in SaaS which is what LLMs are. You can get VMs for much cheaper than AWS, Microsoft, and Google offer them but large companies (and startups) are happy to pay a premium for the support, reputation, and reliability that they perceive those companies as offering. Same thing for some of the managed database providers who are effectively selling a very heavily marked up version of postgres.

> The high price, and social pushback, mean that the American companies producing these models are precarious

I doubt it. The models really aren't that expensive when you look at what they can do. Fable is probably at least as good as the average software engineer and costs $50/wk on the max plan vs a software engineer who would cost closer to $4000 a week. The real money is probably in selling to enterprise vs consumers (Google has best route to making money from consumers since they can do what they did with ads and search to LLM queries).

It seems unlikely to me that US companies will send important corporate data to models controlled by a Chinese company as well.

Yeah, I used to get data corrupting crashes routinely in all sorts of software I use. Now although I occasionally have problems with data not being saved or entered twice, etc. it's been a long time since I lost more than a few minutes of work from a bug like that.

People forget how buggy software used to be or grew up post phone/tablet/webapp where bugs are fixed much faster.

Seems like this is basically the same theory as what happened in the late 90s internet boom. Lots of speculative investment, over build out of core technology like fiber networks, then a crash, and an eventual recovery where that additional capacity was eventually put to use. Similar thing happened with railroads and the airline industry to some extent (pretty sure airlines as a whole had net negative investment returns until recently).

I think the creative outlets will just go into different directions. For example when the camera was invented it caused a crisis in painting since now you could accurately capture reality. So then you had impressionism and all the other -isms. People deride a lot of this art now ("my kid could paint that"), but I think it genuinely was creative at the time since it hadn't been done before. And more recently artists have been returning to a more realistic style but with an emphasis on capturing things that camera can't (for example David Hockney's exploration of very large scale landscape painting or paintings where the objects although painted more or less realistically are not all presented from the same perspective).

LLMs/generative AI make a lot of things that previously required lots of skill to do well easy and accessible to more people. This creates a lot of garbage, and devalues that technical skill but it also opens up new avenues for people to explore.

Even in software engineering I think we're seeing a lot of creativity now, but it has shifted from creating frameworks like Ruby on Rails or Django or systems like Linux and Postgres and towards models for how to program with agents. Although people talk about how LLMs will replace all programmers from what I see the role of the programmer has changed into something more like management. This is of course sad. It made me depressed for a bit since I'm someone who previously prided himself on his technical skill and understanding of hardware minutia, which the LLMs now understand quite well and often better than me, but it has also allowed me to work on larger problems than I could before and to build things that I previously couldn't because I didn't have the front-end knowledge to implement in timely way or the monetary resources to pay someone to do it for me.

My comment is not directly responding to the essay, but it got me thinking about about how agentic programming is much more akin to management than it is to actual programming. Managers generally only have a high level idea of what ICs are working on and often don't have the time, bandwidth, and in some cases ability to understand everything the ICs they're supervising are doing. As more and more software gets written agentically the role of software engineer becomes less technical and more managerial.

The €25k working capital requirement seems a little prescriptive to me (in the US there is nothing like that for non-finance industries, although some businesses need bonds which pay clients/customers if you fail to perform), but it's also the case that most businesses are going to need more than 25k in working capital once they're beyond the startup phase, and outside of tech you typically have working capital requirements that grow with increasing revenue, meaning your accounting profits can be growing but you can still run out of cash since you spend it before you can collect it.

I'm sort of an AI skeptic but I have been seeing this guy's essays for years now and he has always been super pessimistic on AI progress.

I think a much more reasoned critique of AI is that of Tyler Cowen, whose argument is basically that most processes aren't constrained by lack of intelligence but by organizational and social factors which mean for AI to be useful you have to redesign organizations and work to take advantage of what AI is good at. Since most organizations are fairly bureaucratic that takes a while, especially in the large industries that are the most economically important.

Ed's criticism of the large AI companies seems particularly misguided to me since they are the ones actually advancing the technology and seem to have real moats given their access to large amounts of training data from their users. I don't see any possible future in which 5 or 10 years from now there is less AI than we have now and I would expect usage to be much higher.

There have been a few proposals to close that loophole. The main reason it exists is because debt isn't counted as income since it has to be repaid.

> What would you consider a legitimate use case for one?

Setting aside money to pay for a relative who can't provide for themselves, protecting assets if you are professional who faces high chance of being sued (e.g. surgeon), providing for children from a first marriage if you get married and predecease your second spouse.

Perpetual trusts are different from irrevocable trusts, which have legitimate use cases. I don't really see how irrevocable trusts would be gotten rid of. In most states all trusts are irrevocable by default and there is a huge body of law dealing with trusts. Getting rid of them is essentially impossible without huge changes in the political/legal system.

I think the limit it can reach without carried forward losses is 20% because that's the top long-term capital gains tax rate. The other thing I can think of is if you sell a QSBS business, then your capital gains are taxed at 0, and you wouldn't pay income tax at all on that money either. So it's in theory possible that someone could make millions tax free from selling a business, but that's a rare case and one the tax code explicitly allows for.

There are all kinds of irrevocable trusts that exist to remove assets from your taxable estate so that they can be passed to heirs without paying estate tax. Raising the estate tax (which is already 40%) would just make planning to use these techniques more attractive.

I find him more interesting when he talks about non-AI topics. Lots of other interesting people are like this too. I'd rather get my knowledge on AI from people who have unique insights into it. Scott has a lot of unique perspectives of his own, but his views on AI are bog-standard for his social group.

I don't think it's obvious that you were talking about health insurance, which I consider fairly distinct from property, casualty, liability, and life insurance, which are all quite large markets in themselves. The reason I made a distinction is because health insurance is quite different from other lines of insurance because healthcare is federally regulated while other insurance is regulated at the state level.

As mentioned the problems with the US healthcare system are numerous, complex, and interrelated. I don't think they have a simple solution, nor do I think they are insurance problems at their core. For example the cost of drugs in the US vs the rest of the world has very little to do with insurance.

What does a better insurance process look like? Outside of health insurance, which is complicated for a variety of reasons, most insurance is pretty easy to procure. I got an umbrella policy recently and it took about 30 minutes of talking with an agent and answering pretty reasonable questions.

Can you point me to an example then? It's not linked in the article as far as I can tell and it's not easy to find on their website if it's there. I don't count simulations because I used to work with simulations regularly and they often fail to translate to the real world.

They've never translated it to the real world though. So saying the problem is "too easy" when they have no public (as far as I know) demonstration that they've solved that problem is a stretch.

> But frontier models have become really good, and running vending machines is too easy for them now.

Wasn't their previous attempt at running vending machines unprofitable? Not aware of any demonstration that it can actually run that business successfully.