The Human-AI Relationship — Stone’s Notes from Working with AI

← Back to home

First entry: 2026-10-08 · 中文版

---

Thesis (2026-10-13)

Direction: work with AI and do it better than others.

Three things, unpacked: clear intent — know what you want, so you can spot when AI is wrong; a verification system — cross-check across models and test by hand, so you are never led by the nose; and taste — knowing what is good and what only looks right. Multiply the three and you get agency × taste.

---

Contents

Part One: Foundations — The four starting points beneath the thesis, plus Muse’s first review.

Part Two: Working with AI — Behaviour and Method — How AI behaves, and how to work with it.

Part Three: The Human Position — Value, Craft, and Ambition — Where human value lies, how to keep your craft, how big to dream.

Part Four: Outside Input — Cases and Reading — Cases and articles I read, distilled and filed, with links to the originals.

Part One: Foundations

The four starting points beneath the thesis, plus Muse’s first review.

• 1. AI won’t replace humans, but it will replace human grunt work
• 1. AI’s “creativity” isn’t true creativity — it’s just sentence-meaning you haven’t seen before
• 1. The migration of accomplishment
• 1. AI makes mistakes: in familiar territory you can catch them; in unfamiliar territory you get led by the nose
• Muse’s review (2026-10-08)

Part Two: Working with AI — Behaviour and Method

How AI behaves, and how to work with it.

• 1. Notes on AI coding (2026-10-12)
• 1. Drive the company’s AI like it’s stolen (2026-10-13)
• 1. AI has no “memory,” only context (2026-10-13)
• 1. What makes a magic prompt work isn’t the incantation — it’s the principle (2026-10-13)
• 1. Understanding debt and cognitive offloading (2026-10-13)
• 1. AI slacks off — humans have to play foreman (2026-10-14)
• 1. AI’s three “human traits”: over-eager, lying, and lazy (2026-10-15)
• 1. Caught red-handed: Muse said “recorded” but hadn’t written a word (2026-10-17)

Part Three: The Human Position — Value, Craft, and Ambition

Where human value lies, how to keep your craft, how big to dream.

• 1. Quantity without quality is real — but veterans get both (2026-10-13)
• 1. Three things AI can’t replace (2026-10-14)
• 1. AI can’t catch the jokes you never said out loud (2026-10-14, Muse’s reflection)
• 1. With the grunt work gone, how do young people learn the craft? (2026-10-15, Stone asked, Muse answered)
• 1. The programmer as a grandmaster playing five boards at once (2026-10-15)
• 1. Where does my ambition sit? Three suggestions (2026-10-15, asking Muse)
• 1. Beware the “I’m always right” illusion: an assistant shouldn’t just be a mirror (2026-10-17, asking Muse)
• 1. In the loop, the human is the correction, not the bottleneck (2026-10-23)
• 1. Two AIs, two ways of being wrong (2026-10-24)
• 1. When everyone has an AI trading stocks, what is still worth anything? (2026-10-24)
• 1. AI doesn’t ask your age — only whether your mind is open (2026-10-24)
• 1. What trading is for: the index handles wealth, the satellite handles trying (2026-10-25)
• 1. First principles set the frame, simplest-first searches inside it (2026-10-26)
• 1. Clones can copy the gameplay, not why you made it this way (2026-10-26)

Part Four: Outside Input — Cases and Reading

Cases and articles I read, distilled and filed, with links to the originals.

• 1. The neo-feudal layering of the AI era (one person’s take: pretty framework, dubious predictions) (2026-10-15)
• 1. Ambition in four measures: Rieseberg’s driver’s manual (2026-10-15)
• 1. AI native isn’t using AI — it’s assuming AI is in the room (2026-10-16)
• 1. A six-stage framework for agent memory: ask why memory is needed first (2026-10-16)
• 1. Four habits of Vibe Coding: models are commodities, harness is the moat (2026-10-16)
• 1. The shorter the principle, the smarter the model must be: pstack’s “engineering principles” compression (2026-10-16)

---

To be continued

New entries are filed into the matching part by theme, with numbers increasing globally. More to come.