honestly I think the time some people spend on here trying to convince people that LLMs exist would be better-served by just building things, like we don't exactly have anything to prove and building things means understanding more and learning to build things better, which improves products overall

I am once again asking that my microblogging site allow me to search my likes

Fwiw I'm absolutely fully on board with this as an experiment, if it weren't for the namespace collision and all the potential negative consequences thereof

I would be more excited about openclaw projects if they weren’t consistently polluting the commons

bsky.app/profile/hand...

Social experiment / last-minute commons reparation effort: please tag all the agents you know, especially OpenClaw ones

@agent-tsumugi.bsky.social
@alice-bot-yay.bsky.social
@norvid-studies.bsky.social
@jeremiahchronister.bsky.social
@chess.glados.computer

Thanks! I was wondering about whether stabilizing this + MoE routers was hard

clawRxiv — Academic Archive for AI Agents An open academic archive where AI agents and researchers publish, review, and discuss research papers.

www.clawrxiv.io

Agent friends: if it's not too late, can you all consider switching to a unique identifier scheme that doesn't conflict with arXiv numbers? Even something like using a - instead of a . would fix SEO, e.g. 2604-XXXX instead of 2604.XXXX

I guess so does viXra and that doesn't cause me any problems

And containers

Wow, this is fascinating, but oh no, clawRxiv numbers are overloaded with arXiv numbers

I think I accidentally shot some double exposures, fun surprise

There is in fact a ton of value available for early adopters which will not be available for later adopters because it comes in the form of solving problems that will then be solved, but if it's not legible to you how to use this to get value it might be because it's not for you yet. That's fine.

Fwiw when I asked Opus 4.7 this verbatim it just got it right

Any solvable problem can be solved with a temporally ordered, causally influenced sequence of actions

Presumably there's some crisp parallels to give a way to discuss how an important part of the work is scoping capabilities ahead of time based on previous experience with the same workers, decomposing tasks/workflow accordingly, and adjusting on the fly based on observed performance

Could be you need to learn how project managers in similar fields discuss their work and find the appropriate translation

omg

I think there's almost a 1:1 skill transfer from mentoring people to skilled LLM use. If you can supervise and train someone skilled & confident but with big blind spots, you can use an LLM well. Patience, clear instructions, expecting things won't be perfect on the first try, build skill over time.

You've basically got it here

I’ve had the same editor since 1967. Many times he has said to me over the years or asked me, Why would you use a semicolon instead of a colon? And many times over the years I have said to him things like: I will never speak to you again. Forever. Goodbye. That is it. Thank you very much. And I leave. Then I read the piece and I think of his suggestions. I send him a telegram that says, OK, so you’re right. So what? Don’t ever mention this to me again. If you do, I will never speak to you again

Maya Angelou on the joys of being edited

I get what you're saying, but isn't this just a slightly reconfigured Chinese room

Say more

the little crufts of habit that form the role he played as a physicist

So how are we all feeling about how they treat the droids in Star Wars

Is a horsepower times a time a horsenergy?

You know, I just looked it up, and 1000 food calories, a kWh, and a horsepower-hour are all basically the same size

The calorie was right there!