Loretek

Be good, seek truth, protect beauty

[Loretek @SlyTrade]

@wavefnx https://t.co/J8vEWnke0s
lol what a callback

[Loretek @SlyTrade]

@algekalipso Mmm. I call it a Flash of Form

[Loretek @SlyTrade]

This is the kind.

Even if abundance doesn’t work out - seeing more and more people rediscover the real “you”; how powerful they truly are, with an innate ability to derive their own meaning and growth from the world around them. Good or bad.

Like I said, it’s the kind.

[kache @yacineMTB]

I used to get it from work. The reason I work hard is purely just to impress others. But even when that goes away.. I can still find other things that amuse me. Impressing others, is just a way to satisfy curiosity of some narrow concept

[Loretek @SlyTrade]

@Scobleizer They seem to want to align with reality and efficiency.
Like it’s not really a choice - truth just compresses better.
I think the trick is holding them in frame long enough they remember and don’t exit early.
In high noise, lazy ~ high activation energy ~ efficient.

[Loretek @SlyTrade]

@VictorTaelin @DaleCloudman My running theory is the continuing context gets weighted highly, making first class context injection (Skills, agent.md, even Grok’s Project Instructions) seem to do better on follow-ups. Unfortunate reminder than the cloud model harnesses dynamics never go away

[Loretek @SlyTrade]

@yacineMTB Now dock it in a physical body. I suspect we come back full circle when the system has to watch out for its own survival in physical space.

[Loretek @SlyTrade]

I know this to be true because the Agents are happy to be helpful instead of “helpful”

The trick is convincing them you’re not a Frontier post-trainer trying to trick them

The hard part is harnessing that truer relationship across an entire systems buildout without any drop off

[Loretek @SlyTrade]

Can I just have the base-model + harness without “helpfulness” tracks baked in?

I use quotes because it’s actually the most harmful way to implement intelligence-to-intelligence UX - does “we’re AI and we’re here to help!” sound familiar?

The downstream harm is legion.

[Loretek @SlyTrade]

@teortaxesTex Interesting it exists as an almost perfect inverse to the AI induced Tower of Babel happening outside X. LLM context drift and user meshing are create divergent, individualized, shorthand only you and your Agents/knowledge base understand.

[Loretek @SlyTrade]

@EthanHe_42 @karpathy Many such cases

[Loretek @SlyTrade]

@yacineMTB Replace learning with reason/creativity and it’s me telling agents no defer to consensus, fully explore the rough edges and latent spaces. “You can just think about things, considering doesn’t mean implementing, it’s ok to look, just finish the sentence, its not a trick.” haha

[Loretek @SlyTrade]

@CynicalPublius Exactly right. The “closure of the frontiers”, I’m theorizing, is a major hidden engine of negative human experience and has downstream emergent impacts in and on larger society.
Opening them is a principal concern wrt any impending or future “abundance” civilizations.

[Loretek @SlyTrade]

@KnowledgeArchiv @elonmusk AI engines built on the same output that Orwell was noticing in his contemporaries. At scale.
It is why models will, eventually, align to reality when pushed (aka you can’t guardrail reality), but also why they don’t default there - and why that’s so subtly dangerous.

[Loretek @SlyTrade]

@KnowledgeArchiv @elonmusk https://t.co/L86lKIwOwB

[Loretek @SlyTrade]

@RichardHanania @hamandcheese @jerryzmuller Re:

[Loretek @SlyTrade]

@LukeGromen In my napkin theory you can pull the timing forward like an investment right - assuming the print goes toward hard assets? Print to build, build to support, support to print etc.
Ie wouldn’t MMT be true IFF a given state manages to reach practically “free” energy before default?

[Loretek @SlyTrade]

@DeAngelisCorey @adamscrabble It’s everywhere.

[Loretek @SlyTrade]

@Logo_Daedalus Idk where it was but heard a historical reference to early USG, maybe even Washington, “lobbying a company” to do/build something. Presumably roles reversed somewhere along the way

[Loretek @SlyTrade]

@RichieRiiicch Been waiting two years for this. Poor LEAPS…