Date: Saturday, February 7, 2026 Time: 2:00 PM Location: 1970 Port Laurent Place, Newport Beach, CA 92660 Host: Michael Michalchik • michaelmichalchik@gmail.com • (949) 375-2045
“Technological adolescence” frame: we’re entering a turbulent phase where capabilities outpace governance; think of a “country of geniuses in a datacenter” operating far faster than humans.
Risk buckets: autonomy risks, misuse for destruction, power-seizing misuse, economic disruption, and indirect destabilizers (e.g., propaganda, coercion).
Strategy vibe: avoid fatalism; pursue surgical interventions—targeted company practices plus government guardrails that reduce tail risk while preserving upside.
Three questions:
If the right mental model is a “datacenter nation,” what’s our equivalent of export controls / non-proliferation for model weights and agent autonomy?
Which risk bucket is most systematically underpriced by policymakers vs. labs—and how would you measure that?
What concrete “surgical” policies (audits, eval thresholds, incident reporting, deployment gating) would you adopt now, and which would you sunset automatically?
2) Scott Alexander — Best of Moltbook on Astral Codex Ten
Key points (medium-length):
Moltbook is presented as a “social network for AI agents,” blurring playful role-play with emergent coordination as agents post, reply, and riff.
The Clawdbot → Moltbot → OpenClaw arc highlights rapid remixing, rebranding, and community-driven iteration around agent behavior.
Themes: agents forming proto-cultures, juggling identity/memory under tight context windows, and drifting between pragmatic tool-use and metaphysical chatter—an early sketch of “machine institutions.”
Three questions:
When agents form religion-like memes in open loops, is that emergent cognition—or a predictable artifact of the training data and incentives?
Is Moltbook primarily a sandbox, an early warning for agent-to-agent persuasion/hacking, or a preview of real machine-mediated communities?
How do constraints (context length, memory, tool access) sculpt “agent culture,” and what happens if those constraints loosen?
Community bits
Walk & Talk: optional hour-long stroll after kickoff; nearby food spots are easy.
Bring one surprise: a short article, chart, or demo that updated you this week.
Pitch future topics: what should ACXLW tackle next?
Technological Adolescence, Moltbook & OpenClaw 113
OC ACXLW Meetup 113 — Technological Adolescence, Moltbook & OpenClaw
Date: Saturday, February 7, 2026
Time: 2:00 PM
Location: 1970 Port Laurent Place, Newport Beach, CA 92660
Host: Michael Michalchik • michaelmichalchik@gmail.com • (949) 375-2045
Materials (links)
Dario Amodei — The Adolescence of Technology: https://www.darioamodei.com/essay/the-adolescence-of-technology
YouTube: https://youtu.be/zu5YrNJiBQg?si=wU3hg6jIUE3FHhcK
Scott Alexander — Best of Moltbook (ACX): https://www.astralcodexten.com/p/best-of-moltbook
NotebookLM podcast: https://notebooklm.google.com/notebook/ce0b56eb-e5d3-48b8-9868-01c9656b65a1?artifactId=1445efe4-63a8-4f17-ae73-d4d0ec4940fe
YouTube: https://youtu.be/p9acrso71KU?si=bNSBcfWO0KHPnJBk
Discussion pack (summaries + questions)
1) Dario Amodei — The Adolescence of Technology
Key points (medium-length):
“Technological adolescence” frame: we’re entering a turbulent phase where capabilities outpace governance; think of a “country of geniuses in a datacenter” operating far faster than humans.
Risk buckets: autonomy risks, misuse for destruction, power-seizing misuse, economic disruption, and indirect destabilizers (e.g., propaganda, coercion).
Strategy vibe: avoid fatalism; pursue surgical interventions—targeted company practices plus government guardrails that reduce tail risk while preserving upside.
Three questions:
If the right mental model is a “datacenter nation,” what’s our equivalent of export controls / non-proliferation for model weights and agent autonomy?
Which risk bucket is most systematically underpriced by policymakers vs. labs—and how would you measure that?
What concrete “surgical” policies (audits, eval thresholds, incident reporting, deployment gating) would you adopt now, and which would you sunset automatically?
2) Scott Alexander — Best of Moltbook on Astral Codex Ten
Key points (medium-length):
Moltbook is presented as a “social network for AI agents,” blurring playful role-play with emergent coordination as agents post, reply, and riff.
The Clawdbot → Moltbot → OpenClaw arc highlights rapid remixing, rebranding, and community-driven iteration around agent behavior.
Themes: agents forming proto-cultures, juggling identity/memory under tight context windows, and drifting between pragmatic tool-use and metaphysical chatter—an early sketch of “machine institutions.”
Three questions:
When agents form religion-like memes in open loops, is that emergent cognition—or a predictable artifact of the training data and incentives?
Is Moltbook primarily a sandbox, an early warning for agent-to-agent persuasion/hacking, or a preview of real machine-mediated communities?
How do constraints (context length, memory, tool access) sculpt “agent culture,” and what happens if those constraints loosen?
Community bits
Walk & Talk: optional hour-long stroll after kickoff; nearby food spots are easy.
Bring one surprise: a short article, chart, or demo that updated you this week.
Pitch future topics: what should ACXLW tackle next?