Part Four: Outside Input — Cases and Readings
Cases and articles I’ve read, distilled, with source links.
17. The neo-feudal layering of the AI era (one person’s take: pretty framework, dubious predictions) (2026-10-15)
From rick awsb’s long post. He argues AI society will be neither Hayekian utopia nor digital Leviathan, but an inverted pyramid of “neo-feudal layering”: compute and energy monopolized by “digital lords” at the top, middle-class knowledge intermediaries collapsing into AI-leveraged micro-teams in the middle, and the bottom degrading into tribal society on zero-cost AI content. Source: https://x.com/rickawsb/status/2110464793467584996
Three things genuinely worth keeping:
- “Prosperity without pricing power”: AI-amplified individuals and micro-teams see efficiency explode, but the surplus flows back to the top as API fees and cloud rent — the amplifier itself charges “digital rent.”
- The middle-collapse list (legal compliance, junior finance analysis, routine software outsourcing, middle management): the macro version of point 14’s “grunt work gets skipped” — what’s being skipped isn’t just a learning process, it’s an entire employment class.
- Cognitive environment decay at the bottom: zero-cost hyper-realistic content dissolves the shared factual base — the societal-scale version of point 16’s “AI lies.”
Two doubts:
- “Digital lords” may overstate how inevitable concentration is — open source, distillation, and local small models keep tearing down the wall (open-memex’s local-first is exactly that direction). The mainframe era also predicted centralized computing; then came the PC revolution.
- Straight-line extrapolations like “high walls in 1–3 years, hard landing in 3–7” have a reliably terrible track record.
18. Ambition in four measures: Rieseberg’s driver’s manual (2026-10-15)
From Felix Rieseberg (Anthropic, works on Claude), “How to be more ambitious: A tutorial for humans with models.” Source: https://felixrieseberg.com/how-to-be-more-ambitious-a-tutorial-for-humans-with-models/
He sorts ambition with AI into four measures: Speed (doing the same work faster — skipped), Altitude (abstraction level: from “fix this crash” up to “keep the app crash-free” — ask “what’s the goal beyond the goal”), Depth (field-expert depth: use the tools and methods of the top practitioners, until even the experts learn something), Range (cross-field: assemble a crew — analyst + journalist + editor + a red team of short sellers and forensic accountants — and have the model do what they do, not play them).
How it connects to these notes:
- The piece is essentially a “driver’s manual” — Altitude maps to point 3’s “intent and direction,” Depth to “keep your craft sharp,” Range to think tank review. My notes are the Chinese practitioner’s edition; his is the English methodology edition.
- The Depth chapter says it outright: “Use an outside judge — don’t let the model judge itself.” Points 4 and 5’s “use another AI to review” is industry-consensus methodology by now.
- “Insist on the same instruments used by the best”: asked to “act like an expert,” models often change vocabulary without changing methods — talks like a data scientist, hands you a markdown table. Echoes point 5: pretty form isn’t real method. The person reading the board (point 15) has to check what pieces are being played with.
- The closing psychology bit, self-efficacy: people who believe they can do it really do more; “watching someone do it first” (the four-minute mile was matched 46 days later), “not knowing it’s impossible” (Dantzig turned unsolved problems into homework). The counterpoint to point 14: the learning process really does get skipped — but AI also smashed the threshold for “daring to think big.”
Two doubts: the author works on Claude, so “models keep reaching higher abstraction levels” is partly selling his own product; the examples all assume data access and engineering infrastructure, a high bar for ordinary people. And Speed gets skipped — yet in reality 90% of people only ever use Speed. Ambition itself is pyramid-distributed.