GardenLetter27

But the EU still blocks self driving cars which would make this unnecessary.

How is downloading stuff remotely the bottleneck here?

If you're gonna fine-tune for a closed set classification problem like this, you could just fine-tune BERT and get a faster model with better performance.

Could the harness not check for a failed tool call and pass it to a small model for correction without clogging up the main context?

It's not just the architecture but also the data - the decoder only approach lets you train in parallel over blocks of text (no RNN serial waiting), that allows you train on much, much more data.

What? The UK is doing better economically than Germany right now.

And better than anywhere in Europe for Tech with Deepmind and Ineffable Intelligence there.

It still shares many of the same mistakes as Europe though - e.g. now having to buy Russian oil and gas again instead of using the North Sea oil, not expanding nuclear power, rampant welfare and council housing corruption, etc.

Any chance on port forwarding coming back?

And what will a human do better? Why will the human care? Who will pay the human?

The LLMs just churns out non-idiomatic slop in any language.

It doesn't matter if the 800-line if statement is able to use pattern matching.

There's been a lot of progress on making coding agents able to solve problems when they can easily evaluate in a closed loop, we desperately need something similar for controlling complexity and using relevant abstractions.

What? DeepSeekV3 just came out and is incredible for the price. Mythos is also half-released.

And all the oil companies - CRC and Chevron are huge.

AWS Bedrock supports prompt caching, just note that if you use the Converse API you need to set the cache points manually.

The price of everything will go down. That is the beauty of the free market.

It's not hype, the demand for inference has grown more this year than expected.

Hopefully they put ChatGPT on Bedrock now.

The American models also censor a lot of scientific and political views though.

Conservative thinking isn't responsible.

That's how you end up like Germany still using cash and fax machines for 60+ years.

Yeah, he became American, just like Einstein, Fermi, Von Neumann, etc.

There's a big lesson for Europe there, everyone super productive and able to move to the US does so at the first opportunity.

Mandatory age verification is coming.

Reinforcement Learning changes this though - remember Move 37?

The issue is you need verifiable rewards for that (and a good environment set-up), and it's hard to get rewards that cover everything humans want (security, simplicity, performance, readability, etc.)

Atuin is great. This, fish, LazyGit, and zellij are mandatory for me now.

Damn, the Honda E looked great.

LLM slop, they even left the 403 errors in the original dump, and accepted hallucinations.

The main issue is the requirement for relocation tbh.

But the tests were transformed to the new language, they are not copied as-is.

Software patents would work as you describe, but not copyright.

I love the Ultima renaissance now with Ultima VII: Revisited and the Ultima Online servers.

Anthropic wanted government to have a big role interfering and regulating AI as a matter of national security.

And now they are getting what they wished for.

No, but lobbying the government to prohibit open source / open weight models is evil.

They literally want to use state violence to control what we can do on our own computers.

Anthropic doesn't want us to have the right to run open weight models on our own computers. They were never the good guys.

Automated usage like you described violates Anthropic's terms of service.

It's just a matter of time until they ban your account.