prompt-atlas-ecl

Chapter 1 · Profits with Integrity — Expansion

Canonical prose: PROMPT_ATLAS.md#ch1-profits-with-integrity Prompts (YAML): prompts/ch01.yaml · Part: I — Prosperity and Purpose

Worked Example — Century Dividends

Original prompt (from chapter): “Design a financial instrument that pays profits not just to current investors but to descendants one hundred years from now.”

End-to-end walk-through:

1. Frame the instrument’s term — A 100-year covenant bond with two yield streams: an annual coupon to current holders and a deferred yield paid into a custodial trust whose beneficiaries are determined at year 100.
2. Define the trust’s beneficiary rule — Beneficiaries can be (a) any human born in the issuing region during years 80–100, (b) restoration of an ecological asset specified at issuance (a watershed, a coral reef), or (c) a chartered institution with a public mandate. The choice is locked at issuance to prevent capture.
3. Solve indexation — Coupons indexed to Gross Planetary Well-Being (see glossary) rather than CPI alone, so that an issuer who externalizes harm pays more.
4. Resolve enforcement — Custodial trust governance is split: human trustees + an AI auditor with read-only access to the issuer’s planetary metrics, escalating to humans on anomaly.
5. Stress-test — What happens if the issuer dissolves at year 40? (Insurance pool.) What if the beneficiary class doesn’t exist at year 100? (Cascading rules to ecological asset restoration.)

Prompt Templates

# Antifragile diagnosis
"Treating  as a ,
 identify the three feedback loops that *strengthen* it under disruption,
 the three that weaken it, and one ritual we can adopt this quarter
 to flip a weakening loop into a strengthening one."

# Symbiotic-wealth balance sheet
"Restate 's last annual report as a four-column balance sheet:
 (1) Financial capital, (2) Carbon balance, (3) Biodiversity contribution,
 (4) Community trust. For each row, mark which AI signal would justify the figure
 and where a human auditor must verify."

# Ethical arbitrage screen
"Given a list of , return only those whose exploitation
 also produces a measurable public good (specify which good, who benefits,
 how it is measured)."

Anti-patterns

• Composite-index myopia. Collapsing carbon + biodiversity + equity into a single number invites Goodhart’s Law. Keep the vector.
• “AI says it’s regenerative.” A model’s output is not an audit. Always require an external attestation chain (sensor data, third-party verifier, or smart-contract receipt).
• Greenwashing prompts. “Justify why our product is sustainable” trains a model to rationalize. Prefer “Identify the three strongest reasons our product is not sustainable, ranked by evidence.”
• Time-horizon laundering. A 30-year promise paid in inflated currency is a 1-year promise in disguise. Index to physical or ecological assets.

Try This (5 exercises)

1. Quarterly Symbiosis Review — Once per quarter, run the Symbiotic-wealth balance sheet prompt against your team or business unit. Archive results — the delta across quarters is the actual signal.
2. Antifragile Diagnosis — Pick one biological analog. Map your org. Adopt one ritual change for 90 days.
3. Carbon Inversion Sketch — In one paragraph, describe a transaction your organization currently profits from where a carbon-negative re-design would be more profitable in a 10-year horizon.
4. Royalty Economy Thought Experiment — If users earned a dividend whenever their data trained one of your models, what would change about your product roadmap?
5. Bond a Future — Write the term sheet of a century dividend bond your team would actually issue.

Repo Cross-Links

• Z3 Tester — formal “integrity auditor” pattern with Python fallback (no z3 install required).
• losses_geom.py — Sinkhorn-Wasserstein, MMD, and KL distances are useful when you need a vector distance between distributions instead of a scalar score (avoids composite-index myopia above).
• MACP bus — pattern for agents (e.g., a “regeneration auditor” agent and a “human ratifier” agent) to coordinate without coupling.
• See infra/docker-compose.yml for optional NATS / Qdrant / MLflow.

Guide for AI & Humanity

• Integrity is a vector, not a scalar. Resist the temptation to let AI emit a single “integrity score”. Carry the components separately.
• Audit trails are dignity. Anyone whose data, land, or labor is summarized into a number deserves to see (and contest) the chain that produced it.
• Slow signals matter. Ecological dividends accrue over decades. A model trained only on quarterly rewards will systematically underweight what matters most.
• Human-in-the-loop is non-negotiable for any decision that affects (a) ecological irreversibles, (b) generational beneficiaries, (c) communities not represented at the design table.

Citations & Further Reading

• Nassim N. Taleb, Antifragile (2012) — formal source of the “gain from disorder” concept.
• Kate Raworth, Doughnut Economics (2017) — operationalizes ecological + social floors and ceilings.
• Elinor Ostrom, Governing the Commons (1990) — multi-stakeholder design principles.
• Long Now Foundation — https://longnow.org — for the institutional posture behind century dividends.

This site is open source. Improve this page.