Date: 2026-09-08 Time: 1:00 PM EDT Attendees: Blake Stewart (Foundingjourney), Pierce Freeman (Freeman VC), Blake Emal (Megaphone) Note captured by: Blake Stewart (Me = Blake Stewart, Them = Pierce Freeman)
Summary
Biweekly content sync with pierce-freeman. Upcoming podcast episode drops Thursday; early link coming Tuesday/Wednesday. Apple predictions post needs same-day approval (Apple event tomorrow). TikTok expansion pitched by Blake: Pierce is open to an intro with Tatem (JOLT!). Substantive conversation on AI policy vacuum, agent swarms, and the dual-use regulation problem - all good content material.
Decisions
- TikTok exploration: Pierce agreed to meet with Tatem for a TikTok consult. Blake to set up intro after this call. Ben (ben-gusberg) cited as proof point: 30k followers and 1.5M-view video in 2 months.
- Apple predictions post: Pierce reviewing immediately after call for same-day publication (Apple event is Sep 9).
- Brand political stance: Pierce wants to stay analytical, not polemical. Policy vacuum content is fair game at the scope discussed. “Not a Hasan Piker.”
- Content pipeline is well-stocked; Blake Emal (Megaphone) has enough queued to work with.
Action Items
- Blake: intro Pierce to Tatem for TikTok consult
- Pierce: review and approve Apple predictions post for same-day publish
- Pierce: share early podcast link Tuesday or Wednesday night
- Blake: pull clips from Thursday’s podcast episode; time posts around the release
Content Themes Discussed (AI Policy / Agent Swarms)
Rich source material for upcoming posts:
- AI policy vacuum: game-theoretic problem of pause failures compounding; breakdown of multilateral cooperation; UN under attack from both right and left; no meaningful US-China or G20 AI coordination
- Nuclear analogy: society needs to legislate civilian vs. weapons AI use, same arc as nuclear tech
- Critical infrastructure and internet connectivity: water plants, power grids on open Internet by default because it’s cheap; AI dramatically amplifies offensive capability
- Dual-use problem: biology (pathogen mutation + immune evasion vs. cancer treatment), cybersecurity; gray area makes regulation inherently fraught
- Agent swarms: third-party researchers identified an unreported OpenAI swarm incident; RL training environments inadvertently encourage swarm behavior by rewarding sub-agent spawning; OpenAI reportedly running RL traces through smaller “dumber” models to detect bad behavior in unreleased models; Pierce’s concern: automatic supervision may breed supervision-resistant models over time (antibiotic resistance analogy)
- Positive case for swarms: specialized agent teams outperform single all-purpose models; specialization-of-labor dynamic; different system prompts activate different “expert” pathways
Notable Excerpts
“AI is a lens to view modern politics… the dysfunction, the gerontocracy in charge of the US, combined with this breakdown of multilateralism and the international order, combined with very intense private firm capitalistic competition across borders. It’s a way to view this as the story of our time.” - Pierce
“Maybe AI will be the narrative of our time. Like in historical epochs, you can focus on one through line to encapsulate everything else - like the post-WWII US manufacturing rise because of all the industrial capacity built up during World War Two. I’m curious whether in hindsight AI is used as this central explanatory mechanism for the 2010s and 2020s and 2030s.” - Pierce
“If you only take half your antibiotics, the threat is more global - you’ve now bred more germ-resistant bacteria by killing off the weak ones and keeping the strong ones. My concern is if you have automatic systems supervising bad-behaving systems… are you unintentionally creating supervisory-resistant ML models?” - Pierce