arbuge

> I gleefully await the "find out" phase of this little experiment.

I wouldn't hold your breath. US manufacturers and their unions are very well taken care of in Washington, DC. Been that way for decades and will likely be that way for several more, unless something entirely unforeseen happens.

> Her identifying information is basically found at a crime scene.

This means nothing in reality. These days half the USA has their personal information stolen in a data breach every week.

In the US, the burden of proof is on the criminal justice system, which is required to prove guilt beyond any reasonable doubt. The defendant's goal is therefore to introduce that reasonable doubt, not necessarily prove innocence. They are presumed innocent until proven guilty.

> All you have to do is line the walls of your ship with a shit ton of water, which can be mined in space.

But that water needs to be encased in something, and your problem now becomes protecting the encasement.

You need to use search grounding.

> it’s tiny compared to all human payroll

The benefits of replacing all human payroll will go to AI companies as a whole. AI companies as a whole != Anthropic.

As the article points out, the greatest threat to closed models like Anthropic's comes from cheap open source models. Not from the possibility that AI won't be successful at replacing payroll.

..."a mere $150 billion"

I mean, compared to the desired IPO valuation in the trillions, it certainly deserves the "mere" description, but still...

> Many grocery stores have in house peanut butter grinders where the ingredient is nothing but peanuts

Are you in the USA? Never seen one of these to the best of my recollection.

> A harness is storage plus the machinery needed to run one or more conversations with large language models in parallel. It provides the tools those models call, and the execution environments the tools run in.

I don't want to quibble on definitions and semantics here but by this definition, there wouldn't be a single harness out there that I can see except those that include paid cloud storage like Dots.

I think what they're really getting hooked on is the lowest cost provider.

Which makes the Muse 1.3 launch this week particularly interesting, although to get the low cost version you do need to agree to share data with Meta.

> Writing professionally was basically gambling in the 19th century.

Plus ça change...

They approached Nvidia to start the conversation. I guess sometimes great companies are sold, not just bought.