I think whether it requires a lot of work is not the most important factor. The important factor is how easy it is to replace it. If it was made with AI I expect it to be easily copiable with AI. Therefore it doesn't make much sense to spend a lot, if any, money on it.

There is no problem and no contradiction between your views.

If you want it, and you think it's easy to replace, try to reproduce it with AI.

If you find that it is harder than you thought, then go back and pay for the original.

It is surely much easier to replace a potato than to recreate this. Yet we still pay for potatoes.

no it's not. for one, potatoes take months to grow. secondly, they require land that many of us don't have.

generating an ai image of a potato is a better comparison.

Yes it is. I can go to the grocery store and buy one in 10 minutes from where I am now. And it would cost me very little to do so.

That's the analogous equivalent of spending the tokens to use AI to make this project.

The analogous equivalent of growing a potato yourself over months would be creating this website without AI.

How can you claim Goose is faster than C++, i.e. faster than the language whose subset (i.e. C) it compiles to?

C++ code that does what Goose does is a soup of pointer manipulation, casts and byte stuffing. You could write that, but Goose proves it to be safe at compile time, with simple readable code. The moment you write "reasonable" code in C++, it is actually a lot slower than Goose.

For details on how, see https://github.com/aardappel/goose

But then why isn't Goose a library or framework within C++, or a macro in Rust or something else? Why does it need to be a language?

Also, to be transparent in your benchmarks you should compare against idiomatic safe code, and optimized code. It's trivial to throw your benchmark examples into an LLM and tell it to optimize them to parity with Goose. It's more trustworthy to show other languages can achieve the same performance, Goose isn't doing anything special there. What it is doing is doing the same with better safety guarantees, which I think is headline result not the speed itself.

Because you really can't do all the static checking and convenience like that with a library.

As for the benchmarks, that's exactly what I do: https://aardappel.github.io/goose/bench/results.html

If you mean the other benchmarks (LangArena), I don't control their design.

But it's not infeasible as a preprocessor or macro though, since you're outputting C anyway right? You could even output C++ or Rust.

> As for the benchmarks, that's exactly what I do

What I mean is the story of those benchmarks is incomplete, you're missing a column for optimized Rust. This is what I get on my M1 if I ask the LLM to write more performant (still safe) solutions to a couple benchmarks, I'm sure your LLM can do the same:

  | Benchmark | vs original Rust | vs optimized safe Rust | Memory vs original Rust | Memory vs optimized safe Rust |
  |---|---:|---:|---:|---:|
  | `sexp` (flagship workload) | 1.00× | 0.95× | 2.22× | 1.60× |
  | `records` | 1.52× | 0.54× | 4.77× | 2.58× |
  | `graph` | 1.24× | 0.75× | 1.00× | 1.00× |

I think for the Benchmarks Game entry specifically, for the expert solutions you should use submissions to the game for both C++ and Rust instead of LLM generated code. After all, those are legitimate expert-implemented solutions to the benchmark, so it's nice to not have to lean on an LLM there. That would also give you something to optimize towards.

So it's not faster than C++, it's faster than whatever you deem "reasonable C++". I think you should clearly state that.

It's not easy to calculate how much you are worth because it depends on so many factors. And it wouldn't be nice to charge you 10x more than what they charge someone else.

> If you submit a LLM written article to a website/paper in 1990, no one would say that it is a technically poor piece of writing.

If it was one or two paragraphs, maybe. But a full article is very tiring to read and almost always goes nowhere or goes round-and-round.

Can llms even play chess? Last time I tried they were making nonsensical illegal moves (for example, moving non-existent pieces).

That's great advice but still you can be replaced by younger people that also left software for a hands-on job.

Kids will not always be kids.

Because they are not good at investing.

"Half the sites" is a huge overstatement. I very rarely find sites that do not work on Firefox.

Completely agree. I've used Firefox as my primary browser for over twenty years and it's less than once a week that I come across something that needs Chrome.

They will just switch to a betting exchange model where the house gets a percentage of the turnover and the rest is distributed to the players. Under this model there is no incentive to ban winners.

It’s possible this is the future of the industry.

The big money maker in prediction markets is combos or parlays in traditional sports betting. The market for these types of bets is tiny so you need Wall Street firms like Susquehanna to come in and provide liquidity which is what they do for Kalshi.

What this means in practice is you don’t place a bet with fixed odds, it’s more like placing a market order on a stock.

If a Sportsbook could effectively give custom odds on each bet for each player based on the market it would essentially be what prediction markets do today.

The problem with this is liquidity, particularly well before an event. It's why the original exchange (or at least the first one I'm aware of) also now offer a normal sportsbook.