Pete Kaminski

Key Insight: Agentic AI for Document Synthesis

Pete has been using agentic AI for 6-8+ months for software development and is now exploring its application to large text corpora.

Main Points

What is Agentic AI?

AI that can read and write many files
Creates and manages its own todo lists
Works through tasks autonomously like a human developer
"Works with me about at the level of a human software developer"

Tools Mentioned:

Claude Code (both terminal and web versions)
Droid (from Factory Eye company)
Web versions work on mobile
$250 credit promotion for Claude Code Web (expires Nov 18, 2025)

Current Applications:

Complex software development (primary use for 6-8 months)
Serial fiction - understanding large story arcs, writing new episodes
Legal documents - working with document sets that need to fit together
Research studies - analyzing and synthesizing large corpora
"Field of Sheaves" project (see below)

Field of Sheaves Project

Pete is building a "meta-collection" of smashed-together frameworks:

Concept: Taking frameworks that look at the same part of the world but don't talk to each other, and using AI to synthesize them into cohesive 10-20 page websites.

Examples Created:

Collective efficacy
Social cyclic theories
Intention-action gap
Strategy and management frameworks
Government governance principles
Commons stewardship by communities

Challenges:

Half amazing, half embarrassing because "mostly built by AI"
Working on framing what's good vs bad about these syntheses
Too embarrassed to release publicly yet
Seeking collaborators interested in this approach

Each synthesis is a "sheaf" - hence Field of Sheaves

Observations on Adoption

Excitement vs Reality:

Hard to explain why this is cool to others
Watched non-technical people struggle ("oven mitts" analogy)
High overhead to get started
"Getting there" but not quite ready for mass adoption
"Everybody's gonna be doing this kind of stuff in 18 months, 24 months"

OGM Integration:

Often inspired by OGM call mentions
Example: Stacey Druss mentioned building community protection → Pete connected it to broken windows theory → created synthesis

Chat Contributions

Shared Notes:

From Scott: Kevin MacLeod and remixability
From Gil: transcript anonymization/de-identification
From Klaus: farmers, food crop PLUS environmental crop → secondary revenue stream
Defined CSA (Community Supported Agriculture)
From Jerry: Dan Brown's latest novel, non-local consciousness, GABA, psilocybin
From Mike: IETF identity systems, AI preferences standards, Nelson's Law
From Doug: tab groups

On Curiosity: "I'm a very curious person by nature, and observationally, I see a lot of people who are incurious. I'm torn between whether it's a typical human development as people grow up, or our educational system that trains passion and agency out of people, or sort of a social contagion to be incurious as a way to deal with too much future."

"The place I see the curious/incurious split happen is when someone plays with an AI. Curious people use it quickly as a bicycle (or jet plane) for the mind; incurious people do the equivalent of trying to use the bicycle as a hammer or a chair, and discard it as terrible at the small uses they can think of."

On AI Bubble: "We've been in a technology 'bubble' since, say, the PC era. Sure, there have been ups and downs, but we're stuck on tech 'inflation' (both capability and economic).

AI feels like a kind of endgame; either the whole thing (technology, and modern living) crashes pretty hard, or AI enables massive growth in productivity and even things like social justice, perhaps with high costs around giving up sovereignty or agency (or heaven forbid) humanity.

What do you 'invest' in financially when everything will crash, or everything turns wonderful?"

Related Concepts

Back to README

Site Navigation