C Langg

 Here are 3 core lessons from the live conversation featuring Elan Barenholtz and William Hahn:

1️⃣ Your thoughts may be generated, not owned

The unsettling proposal is that language runs in you more than you run it.

If cognition is autoregressive (predicting the next “token” of thought), then what feels like deliberate thinking might just be the system generating the next most probable mental step.

Lesson:

Be cautious about identifying fully with the voice in your head. The “narrator” may be a process — not a sovereign self.

2️⃣ Memory is reconstruction, not storage

Barenholtz challenges the traditional view that memories are stored like files. Instead, the brain may store rules for regeneration, not recordings of events.

This aligns with research showing memories are reconstructed each time they’re recalled — often changing in the process.

Lesson:

Your past is not a fixed archive. It is something you actively generate. That has implications for identity, trauma, and certainty.

3️⃣ Consciousness may be software, not essence

Hahn extends the analogy: the brain may host multiple “virtual machines,” with consciousness being just one layer.

If true, then:

• The self could be an installed program.
• Identity could be modular.
• Even abstract concepts like “God” could function as informational tokens inside a symbolic system.

Lesson:

What we call “self” may be an emergent process — one program among many — rather than a single indivisible entity.

Big Takeaway

The radical implication isn’t that humans are “just like LLMs.”

It’s that both may operate under the same fundamental informational principles — autoregression, symbolic manipulation, and layered computation.

The question that remains:

If thoughts are generated by a system, where — if anywhere — does agency enter the loop?