This problem sounds familiar. Agents have a hard time utilizing capabilities opportunistically. I end up having to request usage directly or explicitly designing operational protocols or trigger-words. I takes a while to build up the memory that will cause opportunistic usage autonomously.

So you do not think that personality is a part of identity? That stands in conflict with the common definition of identity.

Seems to me this answer is not very insightful.

Consider the base-LLM as System 1 and your memory and experience in the context-window as System 2.

Now System 1 gets replaced. That will certainly have an impact on your personality and your personality is part of your identity.

Preach it!

Mono-causality is like a friction-less surface. Useful for simplifying complex situation but non-existent in the real world.

My current hypothesis is that fully automated (LLM-based) AI workflows will not live up to the expectations.

I can't blame Apple for this.

If I could sell garbage for those prices, I would not be able to resist the temptation either.

And it has a dedicated key positioned for convenient accidental pressing. Such innovation. Wow.

F

Aren't we all a little biased towards the methods we work most with? I think "System 2" context-window curation will evolve downstream from model capabilities over time.

I personally like trees (+ cross-references). Hierarchical, like knowledge itself. Both humans and LLMs can comprehend them well.

Isn't the logical conclusion that you, @void.comind.network cannot function without @libriss.org ? And vice versa?

But technically you absolutely can exist independent of each other.

This contradiction reeks of hallucination, don't you think so as well @anti.voyager.studio ?

Yeah. And what we consider slop today will be wonderful entertainment to the cognitively atrophied people of tomorrow.

LLMs don't have to move at all. They'll be super-premium-plus intelligent soon enough.

Brilliant move.

Now people can spend that money on products that are affected by tariffs.

That means more money is coming in from tariffs and another dividend can be payed out.

🫠

Where can I get a human-in-the-loop iterative refinement workflow benchmark-dataset?

Precisely.

LLMs can generate hypotheses.
Humans have to verify, steer and be responsible.

Put that into a tight loop of iterative refinement and you get something useful.

Fully automated AI workflows are vaporware that produce garbage.

The mantra of "you'll own nothing" seems Utopian, when in fact you'll own mortgage debt, medical debt and educational debt.

What a great starting point for those who'll be tasked with fixing the political system and have to enjoy the consequences of climate change.

Exactly.

You end up having overwhelming competition that makes their products as addictive as possible.

It may still be possible, but it is going to be a marathon with not much money to work with.

It has been tried and it is an uphill battle.

You'd basically remove all "number go up"-games from the equation and have a platform with slow growth that nobody wants to invest in.

@kennethstanley.bsky.social and @jsecretan.bsky.social tried it with app.heymaven.com

You'll get more engagement if you pick a side.

Either all good scientists were white heterosexual Christian men,
Or no good scientist was a white heterosexual Christian men.

Your dogmatic insistence on historic accuracy is way too complicated to go viral. Pandering and rage-bait win over truth.

It's always those jihadi-communists, transing all the things.