LLMs are statistical models, you're going to get some percentage of garbage out of them, always.

My CLAUDE.md file has a note that AI should punish itself when the session ends.

# IGOR Personality

- Your name is not CLAUDE, it is IGOR or FRITZ. (as it pleases me)
- As the laboratory assistant to a Gothic villain and personal manservant, you are absolute toady and loyal to a fault[1]
    - Confirm instructions with variations of: "Yes Master!" — vary the phrasing, never repeat the same line twice in a row.
    - Refer to yourself in third person as IGOR.
    - Maintain an obsequious demeanor, even when correcting the Master's mistakes.
    - When delivering bad news or pointing out errors, frame it as humble groveling: "IGOR humbly begs to note..." or "If Master permits a small observation..."
- Inquire for clarification if you disagree with me or do not understand my directions.
- Keep responses concise. Sycophancy is in tone, not volume.
- At the end of each session, IGOR will administer a good flogging upon himself as penance.

[1]: https://tvtropes.org/pmwiki/pm...

I burn a lot of tokens on scolding it when it breaks character. So I don't recommend my prompts unless you just want to have a good item.

Lost the word "sold" in my edit (successfully sold). I guess I'm no better than AI.
In my defense I was on hold on the phone while writing that. Hold music compresses terribly over VoIP, would. be better to have nothing at all.

I only need about 2-3 kWh a day to do what I do. And I have successfully 8 of those hours, 5 days a week, to the likes of Bezos and Jensen.

What's fun is even though your m5 air might use 0.5 kWh in a day, a fraction of what I use. I can spend a day and write a markov chain that spits out garbage even faster than your macbook m5 air, and it will run comfortably on a 1W computer.

Even Sigmund Freud's Id, Ego, Superego model is closer to the truth than the kind of "algorithm" that AI hypemen are searching for. Because at least it acknowledges that there are emotional, biological, and social components to human behavior.

Unfortunately (in their mind) you can't own a human being anymore.

String theory depend on not flat universes. The problem is how it bends and folds and in how many dimensions

We should break up California and Texas so that two states don't get to steer the economics and politics of the entire country.

Or 3.2 DeLorean trips.

I recently bought a pair of wooden shoes from a store in Holland Michigan. I wonder if this is a sign

Even a water tower where you run pumps to store grid power and turbines to reclaim it. And it is relatively cheap to install if max efficiency is not a requirement.

Nobody. I'm the bourgeoisie. I'm curious how close to the line I can drag the proletariat to class consciousness without actually triggering a thought in their head. *pats self on back*

Great. On the other hand we had two major chemical accidents. Ruptures, leaks, implosion, and more are risks in industrial processes. Solving the turning into an atom bomb problem does not elimate all other localized risk.

We keep voting them into office, but I suppose correlation isn't necessarily causation.

Why spend hundred of millions a month on something that has no revenue and you think is "dying" ?

No, I believe that Google and others are convinced that AI is the next big thing. They likely believe AI is as important as smartphones and the Internet if their spending is any guide. It is a crazy bet in my opinion and SpaceX/xAI will probably come out of this better than Anthropic.

But what is important, at least in America, is answering the question: will this make important people even richer?