VikRubenfeld

Surprisingly, there's no category name yet for these things. "AI Agents" -- tons of things are agents that aren't in the category of OpenClaw, Hermes, Grok Bot, etc. "AI Bots" -- everything's a bot. Can we think of a good category name for them?

I was using Claude yesterday and the "advice" it was giving me quickly became confused and irrelevant to the prompt, even though there was not much text in the context window. Speculation: since Claude re-reads the entire chat at every turn, the "minimal" text revisions required by watermarking quickly compound such that even Claude can't follow the discussion.

In many cases, isn't the time-consuming part of experiments irreducible? E.g. breeding plants?

Or to take another example, Make Solar Energy Economical

How does Discovery Loop make this go faster in a way that a different group of scientists, also using frontier models, will proceed?

I'm sure Discovery Loop has considered this and has good answers to this question. I'd be interested in hearing more about this.

On the one hand, Darius argues:

> ...banning the use of these models by US businesses does nothing to address this risk, because bad actors are unlikely to be legitimate US businesses. It would protect US AI companies from competition, but that has never been my goal.

But on the other, he argues:

> All sufficiently capable models, open and closed, should go through mandatory safety testing.

Models that don't pass safety testing would be banned. Darius does not appear to be against banning models. He wants the government to have a regulatory body that has the ability to ban models. Then Anthropic can do regulatory capture of that agency and control what models are permitted to be released.

Also, during this mandatory safety testing, models would be blocked from use, and by the time the testing was done (probably years) the models would be obsolete.

"everyone knows the security threat is a pretext." On what planet? Anthropic itself made a big stink about Mythos being able to hack every app out there, and very dangerous as a result. Many reports have confirmed this.

The article says "A senior C.I.A. official was arrested last week after investigators found hundreds of gold bars worth over $40 million stashed in his Virginia residence, a small fortune that he apparently brought home from work, according to court papers."

The same article also says:

"When the C.I.A. conducted a review of where the gold and currency were stashed, the agency was “unable to locate the gold bars or significant amounts of the foreign currency,” according to court papers."

The NY Times fact checkers don't seem to have seen this article.

"That's the self-doubt that it's feeding on and what I need to fight back."

Yes -- now let's talk about the correct form of fighting back.

It is not "I don't want to feel self-doubt so I will suppress that feeling."

It is, "The self-doubt is valuable -- it's pushing me to improve."

The AI is never going to be able to say what you really mean. But it may inspire you to push harder to improve your ability to do that.

These LLMs are great for now, but they have to go by their training materials. And if people stop creating new ways to code, new languages, new coding patterns, etc, then the code LLMs produce will be stuck in 2026 forever.

Is there something free like Codes or AntiGravity that can run open-source LLM models?

What I don't think people are talking enough about yet, is that AI doesn't invent new ways of doing things -- it just predicts the next word based on the materials it was trained on. That means that if a company lets all its coding be done by AI -- that company will be permanently stuck in, say, 2026 -- while other companies will be continuously improving.

We've all seen that if you are interacting with an AI over a lengthy chat, eventually it loses the plot. It gets confused. It appears to me that it's necessary, when coding an AI, to keep its task very limited in terms of the amount of information it needs to complete the task. Even then you still have to check the output very carefully. If it seems to be losing focus, I start a new task to reduce the context window, and focus on something that still needs to be fixed in the previous task.