xw333

Actually I was really happy you included an example because it demonstrates just how hard is it to make a rule so clear that no one has to guess again. It requires adding all the detail and addressing all the nuance and before you know it, you're programming again. The structure of the language is irrelevant.

> "Opus writes better code than me"

Probably true for some definitions of me. :)

I'm not so sure about that in the generic sense.

This sentence from the article really stuck out to me.

> and you write the rule down so clearly that nobody (human or otherwise) has to guess again.

Writing rules, or steps, unambiguously in a high level language and having those statements translated to lower level languages, that ultimately get executed by a machine, is what compilers do.

The commands you gave were not unambiguous at all and although an LLM would produce a result, the result would definitely be nonsense.

"make me a budget, here is 6 months of credit card receipts"

For example, a valid budget requires some knowledge of expected income...

I also found the approach interesting but I've also noticed real players have personalities, i.e. do they push for open or closed positions, and they have tactical strengths and weaknesses. So the probabilities of different events needs weighting.

I also coach kids and none of their moves are random in the mathematical sense. They dont play the best moves but the moves are understandable once you hear their reasoning, it's just their reasoning doesn't properly evaluate the immediate position, future shapes and transformations.

For example looking at my chess.com profile I find basically every hung piece, and I am excellent at taking advantage of pinned piece opportunities and almost never allow an imbalance due to pinned piece, but I have blind spot when it comes to clearance sacrifices both taking advantage of them and allowing them.

So I suspect a good way to approximate a human would be to bork the evaluation based on a profile and then pick from that narrowed set of candidate moves at random. Which is more or less my understanding of how the chess.com bots work

On the other hand, post-release patches made very complex software practical. And very complex software takes a very long time to test and people want the new stuff now.

There is no easy or even single "correct" answer. Just trade-offs.

It's a trade-off. It could be a setting, but if one adds too many settings then the space of possible options gets huge and bloated leading to bugs and complexity for the users and I already barely know how to use iPadOS.

"If you ask a user what percentage of Word features they use, they will tell you 10%. But if you ask multiple users which 10% they use, they will all give you a different 10%."

The moat of the closed source EDA tools is the validation hours that went into them.

I use the open source EDA tools for VLSI everyday and I do like them, especially how hackable they are, but they are very unreliable compared to the closed source tools, and that's not the highest bar.

Run time of the open source EDA tools are also very slow for designs over, say, ~250K gates. And that's an issue when in commercial designs the base unit for measuring the number of gates in a macro is megacells, i.e. 10^6 cells. Much of the effort of the commercial tools has just been keeping up with Moore's law which is not easy when many brute-force form EDA algorithms are like n^3 even with heuristics.