Schayfr

📡 AV Tech, Ai Consultant to those who didn't ask - Carries tools and books with equal familiarity 🛠️ 📚

[Schayfr @schayfr]

@AbudBakri Most won’t say “my sample size is small and biased”, they’ll say “the institutions are wrong, captured, or behind”

[Schayfr @schayfr]

@hubermanlab Unsolicited podcast guest recommendation: Joel Greene @realjoelgreene

Immune-centric, anti-aging, longevity, diet and health wizard! Consultant to Ron Penna in the early Quest Nutrition days, authored 2 phenomenal books, runs Veep Nutrition. Guy's pretty neat

[Dr. Cameron Maximus🤴🏻 🥷🏻 🧙🏻‍♂️ 🤵‍♂️ @DrCamRx]

Here’s why you don’t want longevity:

[Schayfr @schayfr]

@EthicalSkeptic The genetic code looks engineered because living matter is capable of engineering itself, long before humans existed, using goal-seeking dynamics we are only beginning to understand.

[Schayfr @schayfr]

8/
A constraints-driven reality means:

• patterns have causal power
• histories matter
• goals matter
• environments shape agents
• structure drives behavior

Once you see it, reductionism never makes sense again.

[Schayfr @schayfr]

1/
Most people still think reality is built “bottom-up”: atoms → molecules → cells → organisms → minds.

That’s smallism — the assumption that smaller = more fundamental.

It’s wrong.

Reality is a constraints-driven hierarchy, and it changes everything.

[Schayfr @schayfr]

7/
Assembly Theory points the same direction:

The history of construction determines what smaller pieces even can do.

Macro-level order → micro-level possibility.

Top-down causation is everywhere.

[Schayfr @schayfr]

6/
Ontological anarchy (“every level is equally real”) misses the point.

It’s not that all scales are equal.

It’s that the scale with the constraints is the one that matters for explanation.

Hierarchy by control, not by size.

[Schayfr @schayfr]

5/
This explains why:

• ecosystems shape evolution
• tissues tell cells what to do
• culture shapes individuals
• shop layout shapes action
• prompts shape LLM behavior
• workflows shape cognition
Constraints beat components.

[Schayfr @schayfr]

4/
This isn’t mysticism or emergence woo.

It’s just how multiscale systems work.

Constraints at higher scales regulate what’s possible at lower scales — like rules, boundaries, and goal-states shaping behavior.

[Schayfr @schayfr]

3/
In biology, Levin’s work makes this impossible to ignore:

Cells don’t “build” a limb.
The target morphology does.
The macro-pattern constrains the micro-actions.
Pattern memory > molecular mechanics.

[Schayfr @schayfr]

2/
A constraints-driven hierarchy means:

The level that explains the system is the one that controls it,
not the one that’s physically smallest.
Causation follows control, not size.

[Schayfr @schayfr]

@Sara_Imari @MikeArdoline While I agree there's no "ontological sheriff" in town, I think reality is ordered by constraints, not scale. There's no "most fundamental"; the explanatory level requires knowledge of the controls of the constraints. It isn't anarchy but a constraint-driven hierarchy

[Schayfr @schayfr]

11/
The debate isn’t:
“Do LLMs really understand?”
The real question is:
Why did we ever think understanding required internal representations at all?

[Schayfr @schayfr]

1/
Large Language Models just quietly overturned 70 years of cognitive science.
They show that meaning without representation isn’t just possible —
it’s the natural result of distributed learning.

This is the biggest philosophical shift since the birth of computing.

[Schayfr @schayfr]

10/
The takeaway:
LLMs are the first large-scale artifact built on the principles of ecological cognition rather than classical computation.

They work because they ignore the old assumptions.

Meaning is emergent.
Cognition is distributed.
Representation was a detour.

[Schayfr @schayfr]

9/
This is why philosophy, neuroscience, and AI all feel like they’re being rewritten in real time.
LLMs didn’t prove they’re conscious.
They proved that our model of cognition was wrong.

Meaning lives between agents, not inside them.

[Schayfr @schayfr]

8/
LLMs aren’t showing us “artificial intelligence.”
They’re showing us what intelligence actually is.
A system doesn’t need inner representations.
It needs structured exposure + compression + reuse.
Everything else is mythology left over from the 20th century.

[Schayfr @schayfr]

7/
The key insight:
Meaning is relational, not internal.
It arises when a system becomes attuned to the patterned constraints of its environment.

This is true for babies, beehives, tissues, and transformers.

[Schayfr @schayfr]

6/
LLMs succeed not because they understand the world, but because they understand the structure of how humans talk about the world.
They model the ecology of expression, not a semantic map.

Meaning emerges from participation, not internal mirrors.

[Schayfr @schayfr]

5/
The classical view assumed:
“Intelligence must use internal representations.”
This made sense in 1960, when computers = symbol manipulators.
But nature doesn’t work like that.
Brains don’t behave like logic machines.
Language doesn’t behave like formal grammar.