jeffreyrogers

If The Great Game you mentioned is the one by Peter Hopkirk, it's fantastic.

If you enjoyed those books I can highly recommend John Biggins' Otto Prohaska series, which I've seen described as "a techno-thriller set on a WWI submarine". Biggins books are really fun and well researched.

I recently finished George Vaillant's book Adaptation to Life and immediately picked up the follow up The Wisdom of the Ego. Vaillant is a psychiatrist and was a director of the Grant Study, which followed about 250 Harvard undergrads from college until they died.

Both books are largely about how the psychological defense mechanisms a person uses play a large role in determining how successfully they adapt to and navigate life.

> LLMs are stochastic and probabilistic

So are humans. Every codebase I've worked on has duplicated code that has been written in slightly different (but hopefully equivalent) ways, often by the same person.

Most managers don't read the code of their direct reports. They just trust that the code is "good enough" and that there are enough other processes in place to catch bugs before they cause too much damage. I think I'm a pretty skilled programmer, or I'm at least good enough at interviewing to convince other people of that, but it's pretty obvious from comparing LLM generated code to code I'd write myself that in many domains LLMs are superior to me. They do make mistakes, but so do I, and those mistakes are eventually found and corrected.

We're in the very early stages of LLM driven programming, so it's hard to say how it will all shake out, but my anecdotal experience is that LLM written software is very reliable and easy to extend and develop. I have a side project that is about 90% LLM generated code (about 25k lines of production code and a similar amount of test code). This is a revenue generating product and I've had no issues with reliability, security, or performance.

For what it's worth this app is a rails app and I have no plans to switch to anything else. Rails works nicely, the LLMs extend it easily, and almost everything is I/O bound so I don't need C++/Rust level performance.

I think there's aspects of this viewpoint that are true. I'm sure I'm less skilled at programming now in some ways than when I started using LLMs, but I'm a lot more productive and I can work on a broader variety of software than I could before. Writing probably weakened people's auditory memory, and calculators probably reduced people's mental math skills, but I don't think either are a net negative. It just requires some adjustment to get used to the new way things are. I also think my programming skills would come back pretty quick if I started programming without LLMs again, just as my leetcode skills come back pretty quick when I prepare for interviews.

People are still interested in chess even though computers are much better than humans.

Not really unique to VC, similar things are happening to private equity with secondary funds.

I rarely use image search, but I went to look something up recently and I was shocked at how many obviously AI generated images showed up. I couldn't even find an image of the thing I was looking for and eventually gave up. Bing has the same problem.

I do think it is kind of funny when VCs talk about how they can tell who will be a successful founder. They are picking the people who get to call themselves founders so it is close to tautological that who they think will be successful matches who ends up succeeding. You don't get to see the counterfactual world where the non-funded people get given money and then you see what personality distribution the successes from that group have. The problem probably is worse now since there are fewer IPOs and more companies exit in private transactions where it's not clear how much economic value was actually created vs how much is just what the investors agree the business is worth.

There's a trend in top MBA programs called entrepreneurship through acquisition (ETA) where you raise money to buy a business that you acquire through a mixture of debt and equity. Basically lets MBAs skip the building phase of building a small business and go right into operating it.

I run a small business on the side, and although I did a ton of preparatory reading and learning not much of it was useful in retrospect. Most of it you learn by doing and it's not that hard to learn. The top three things that were useful to me to learn were sales, basic accounting and how to read financial statements, and what metrics to track/manage with. You can learn the basics for all of those in less than a month.

My advice is just start. You are probably wrong about what the market wants unless you are selling a product or service that you know there is already demand for. Figure out what it is you're offering and try selling it to people. If they want it you figure out how to make it better, if they don't you try something else.

Edit: I had wanted to start a business for a long time. I thought I needed some tricky new idea to be successful. That's really hard to come up with since most new ideas are bad or are too hard to sell/explain to people. In my case it was also a form of procrastination since I had to wait for the right idea. Things got easier when I just picked an existing problem/industry and decided to do that with my own proprietary software to make it easier for me. I know a guy making a couple million a year from owning multiple tanning salons. There's a lot of opportunity out there that doesn't require any special insight.

I guess one thing universities have to decide is how important having founders is to them. PG talks about how Harvard has more founders than Yale and Princeton, but I don't really see that as a good or bad thing for any of those schools. There probably aren't infinite good ideas that can be successful at any one time. If you encourage more people to become founders then they aren't going to go into other places where their skills might be better suited.

YC's founders also skew heavily towards a certain type of person and business. The returns of VC in biotech for example are much lower than the returns in B2C and B2B software. It's just a harder, more capital intensive, and more uncertain industry. Over optimizing for the YC archetype means you get less of the people who succeed in other areas.

Also, although PG disdains finance and management, many large problems are better addressed through people with skills in these areas IMO, since often the returns are not sufficient to attract VC interest but the financial and human resources that need to be coordinated to address them are substantially beyond the ability of a small, undercapitalized group.

No, it's been my experience asking LLMs to explain complicated code to me. They can do a pretty good job of it. You can also ask them to write a bunch of tests and then rip out the old implementation and fix it with a new one. LLMs aren't perfect but they are really good at understanding and writing code.

I think it is easy for companies to waste time and money on lots of things when it is hard to measure the impact. For example, I think most people would agree that the big tech companies are bloated and have too many employees for the amount of work they have. There's not really an incentive to fix problems, and in some cases (like with headcount) there are counter-incentives (managers want to have more reports).

I think companies will get better at measuring the impact of AI and attributing it to increasing profit or decreasing costs, and that the companies that are better at this will have an advantage over those that are worse, so eventually overall efficiency will improve.

It makes organizations more productive at producing code and doing other tasks, but translating that into something that affects PnL is different. Where I work it's sped up individual tasks I've worked on but I don't think it's sped up delivery timelines of any of the major projects I'm involved in. We just added additional verification work with the extra cycles the engineers have now. It's not like that work is useless. It will probably mean I have less debugging to do in the future, but when you look at how the business makes money I don't see it making a big impact.

The LLMs are pretty good at helping with that though.

Sol

That's surprising to me. I recently switched from Claude to Codex because I was blowing through my Claude limits. With Codex I haven't been able to use up all of my usage on the 20x plan, with Claude and fable I could do that in a day. Codex is also a lot faster than Claude. I do think Fable is slightly better and can come up with better abstractions than Codex but trying to read it's output became really frustrating to me.

No it didn't. It used to be more readable and he changed it to look like this.

His site didn't always look like that, for some reason he changed it a while ago and now I have to use a custom stylesheet so I can read his articles. It used to be more readable.

I hope at some point Anthropic does a post-mortem on the strange behavior their models have been displaying recently. I mostly switched to Codex because I was finding Claude's behavior increasingly frustrating.

I haven't run into this yet. My test failures have either been real or have been triggered by an (intentional) breaking change. The later does require updating the tests, but I think catching the real bugs is worth that tradeoff. I haven't found spurious failures to be a big problem (I've had a few but rewriting the tests that have this problem has eliminated it for me). My codebase is a pretty modular rails app, which I think helps with this.

I have worked on other projects where flaky tests are a problem (and generally cause developers to ignore and submit anyways), but so far I've avoided that. My experience with flaky tests is that they're typically due to poor modularity or to subsystems that other teams can modify. This project exists in a monorepo and I'm the sole developer so that's not a problem here.

I was fairly skeptical of agentic coding before I used it for a real product. Although I still have to be heavily involved in planning the code that LLMs write for me, they can write code much faster than I can, and they know more about edge cases than I do, so they can handle edge cases/subtle bugs that I would have missed. I have been paid to write code at every level of the stack from assembly to frontend javascript, but I'm not equally good at all those areas. In some areas I can still outperform LLMs, but for areas I'm weak they do a much better job than I would have.

I still think of what I'm doing as software engineering, and I'm glad that I had many years of professional and hobby development before using agents since I think that's given me the ability to make good architectural decisions (and helps me resteer the LLMs when they want to do something suboptimal), but my involvement in actually writing code is quickly going to zero. That said, they aren't perfect and they still introduce bugs, but I believe the quality of my current product is higher than what I would have created pre-agentic coding.

Things I've found helpful in keeping quality high:

- Visual regression tests (detect UI bugs before you commit them)

- Fuzz testing of interfaces and app behavior

- Automatically add regression tests for any bug that I/the LLM fixes

- Logging/alerting that tracks an errors/invariant violations triggered in the app

- Performance metrics that are surfaced in a dashboard.

All of these are very easy to add since the LLM can create this infrastructure for you. The fuzz testing in particular is something very few products I've previously worked on have since most people don't know how to implement it. I ran the fuzzers for a few minutes and they quickly caught multiple subtle bugs that I was not aware of.

This is a real product that helps a real, non-VC funded service business, and although I could have made something similar myself it would have taken me a lot longer, be harder to use, and probably be less reliable.

Edit: while it's true that you can quickly blow through the $20/month plan, the $200/month plan allows you to get a lot done and is basically sufficient for my needs. It's also very cheap when you consider what it would cost to pay someone to do similar work.

I wonder how it compares to small business fraud. I know of multiple frauds in non venture backed ecom and real estate development. In some cases the person involved seems to be a professional fraudster, having had no real employment outside of fraudulent businesses.

> Like, the one high signal claim he made turned out to be 25x incorrect!

He has dozens of high signal claims in his prediction blog posts. Picking out one of the ones he got wrong doesn't tell you much. Overall his track record is pretty good. Definitely better calibrated than most people, especially the very anti and very pro LLM people.

Last time I checked xgboost and lightgbm still outperformed anything NN based for tabular data.

You can also roll your traditional IRA into an employers 401k (if the plan allows this) to zero out your traditional IRA balance.

Yes, although some plans let you roll contributions into an IRA.

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.