spencer-p

> In almost every run, the GPT agents were much more aggressive .. riskier strategy vs. Claude’s more defensive posture that tried to win by kill count.

> A tool bug stopped block-breaking. GPT diagnosed it and shared workarounds; several Claudes concluded breaking was impossible.

I wonder if Claude's defensive strategy would have emerged had it believed breaking the bed was actually possible?

Do you actually come back to a tab 2,000 tabs later?

Lately I've tried using an extension to close any tabs I haven't opened for a few days. I haven't missed them, and if I actually want to come back to something, I make a todo now.

Astronauts can spend many months on the ISS, and it’s very important they don’t become irritable and kill one another before the research is complete.

Weird that the steps are for "Gemma 4 12b", which does not exist, and then switches to 26b midway through.

There's also a step to verify that it doesn't fit on the GPU with ollama ps showing "14%/86% CPU/GPU". Doesn't this mean you'll have really bad performance?

The question in the zeitgeist is whether coding agents will be to software engineers as the quartz revolution was to expert watchmakers.

Commoditized software is here. Will there be a market for high-end, luxury software? Becoming an artificially scarce veblen good is unlikely to work for digital goods the way it has for watches.

I do miss n-gate. I have to assume they are much happier now that they've ended that project, though.

I think they mean to refute the article's suggestion that tiktok and misinformation are the cause by highlighting that they received accommodations at a young age.

>You seem to imply your conditions have not really improved and you need same/similar accommodations

They didn't share the nature of their current accommodations.

The article frames being smart and promising (to a university) as at odds with having a learning disability, which is not necessarily true. It also frames depression and anxiety as learning disabilities, which they are not.

Amazon claimed their energy usage was "matched with 100% renewable energy" in 2023, while Google's goal is to actually "run on 24/7 carbon-free energy".

Google already claims to have matched 100% of their energy with renewables since 2017.

I appreciate how the author highlighted the python domain-specific tricks like dropping imports and rewriting tabs/spaces. It's good to be reminded that even with "large" language models you can get better results with quality over quantity.

The quote you pulled out mentions "distributed execution" being completely contained in a "single-program system" - isn't a single-program system, by definition, not a distributed system?

With the emphasis on distributed systems, I was waiting to see how the approach would help synchronization, replication, network dependencies, etc. But all the examples show calling other services as functions or spawning new processes. At the end orderd starts listening for requests, but I don't see any example in the article where the example program talks to another computer in the system. Perhaps I am missing something?