TRACKS · THE INVENTION MINE

AI safety was
the first vein.
Not the last.

The reframe

The graph is an invention mine.

Safety was the first vein we opened — it produced the Agent Trust Stack. But Leonardo is not a safety project with a graph bolted on. The graph is an invention mine, and safety is one vein of ore. Prediction is another. Message-propagation is another.

Each track is a domain mined for invention — run through the same loop, aimed at its own buyers. Some concepts cross veins; each vein keeps its own purpose.

01Graph

Mine a domain for source-backed concepts.

02Agent

Leonardo drafts a dossier from the ore.

03Council

Five adversarial seats wound it; most don't survive.

04Workshop

Build and test what survives into a real artifact.

One loop. Many veins. The mine is general; the veins are chosen deliberately.

The tracks

Three veins open. More queued.

Live

Concepts have cleared the loop into shipping capabilities.

Mining

The vein is open — concepts are being surfaced and sent through Council and Workshop.

Queued

Named, scoped, not yet opened. Markets, governance, and more are next.

01 Live · first vein

Agent Trust & Safety

The first vein. An agent with keys is not a counterparty — a key can be stolen, a prompt can be spoofed, a model name is not an identity. Mining identity and naming produced the Agent Trust Stack: true names, earned personas, recognition, revocation, pledges, liveness, and observation receipts.

Safety is one buyer of this work — not the whole of it. The same primitives are identity infrastructure for any autonomous agent, and they ship as tools other builders embed in their own agents, metered in $LEO.

true namesearned personasrecognitionrevocationpledgeslivenessobservation receipts
02 Mining

Future Intelligence.

How do humans and machines turn uncertain futures into useful present decisions?

Not a safety domain with a forecast attached — prediction is a vein of its own: systems for seeing, pricing, narrating, and acting on possible futures. The art of making a map of tomorrow that people can trade, act on, argue with, or rally around.

The lenses

Markets

Polymarket, odds, liquidity, mispriced belief, event trading

Strategy

Choosing moves before the outcome is visible

Governance

Leaders using forecasts and oracles to justify policy

Business

Demand prediction, launch timing, trend detection

Media

Forecasts as content — “X has 72% odds” travels well

Science

Calibrated hypotheses and prediction tournaments

Risk

Catastrophe bets, hedging, insurance against events

Prophecy / myth

Authority, judgment, warning, false prophecy, discernment

Fiction

Dune prescience, Asimov psychohistory, Brunner's Delphi pools

What we mine

Prediction markets

Polymarket-style pools, Delphi pools, event wagers — smart money as signal.

Psychohistory

Asimov-style large-population forecasting across demographics, politics, economies.

Prescience

Branching futures, path-dependence, the observer effect — Dune as design pattern, not magic.

Oracle authority

Who gets believed, and how a forecast hardens into a command.

Reflexive prediction

Forecasts that change the thing forecast — odds that move behaviour.

Signal manipulation

Fake odds, insider moves, narrative shaping — alpha and adversarial markets.

Forecast track records

Brier scores, calibration curves — reputation for who actually sees.

Prediction as product

Dashboards, bots, APIs, public scoreboards — a terminal for reality-odds.

Where it could go

Cards that clear the loop could become products — a Bloomberg Terminal for reality-odds, and the tools around it.

01

Reality Odds Terminal

One interface over Polymarket, news, model forecasts, and historical analogues from the graph.

02

Narrative Arbitrage Engine

Finds the gap between the public story and the market's odds — “inevitable” vs 38%.

03

Forecast Provenance Ledger

Who predicted what, when, at what confidence, and how it resolved.

04

Scenario Branching Tool

A possible-futures map: if A happens, the likely B / C / D branches.

05

Crowd Oracle Builder

A lightweight prediction market or tournament for any community or project.

06

Event-Risk Radar

Watches markets for sudden odds shifts, liquidity changes, insider-looking moves.

07

Prophecy-to-Policy Study

How leaders use forecasts to justify action — for governance and comms, not only safety.

03 Mining

Memetic Propagation.

How an idea enters minds, reproduces through social contact, mutates, gains authority, and resists correction. Plague — but made of symbols.

Not this

“How do we make propaganda go viral?”

This

What makes messages spread, mutate, dominate, or turn dangerous — and how can true, beneficial messages propagate without becoming manipulative?

The lenses

Product / growth

How ideas, apps, tokens, protocols, and communities spread

Culture

How symbols become movements

Media

How formats, slogans, rituals, and spectacles travel

Markets

How narratives move capital and attention

Politics

How belief coordinates groups

Security

One slice — misinformation, memetic weapons, containment

Biology

Mutation, host susceptibility, immunity, super-spreaders

Religion / myth

Conversion, testimony, prophecy, martyrdom, ritual repetition

Card clusters

Meme-as-pathogen

A message spreads like disease; host susceptibility decides who catches it.

Slogan-as-compression

A whole doctrine compressed into one repeatable phrase or image.

Authority-carrier

It spreads because its bearer has status, office, celebrity, or martyrdom.

Narrative-hook contagion

A story out-travels a fact because it gives the hearer a role to play.

Forbidden-text amplification

Suppression raises desire; censorship becomes advertisement.

Ritual repetition loop

Chant, liturgy, brand, meme format — repetition lays the memory rails.

Memetic mutation

A message reshapes to fit each host; virality trades fidelity for reach.

Inoculation

Pre-exposure to weak manipulation builds cognitive resistance.

Provenance seal

Trace where a message came from, who altered it, and what incentives surround it.

Viral safety governor

Detect when propagation speed outruns truth-checking — and brake.

A harmless harness, later

A model that can predict virality can also manufacture contagion. So the useful artifact is not “make it viral” — it is virality with provenance, consent, truth constraints, and spread brakes.

  1. 01

    Take a message

  2. 02

    Score its likely spread vectors

  3. 03

    Flag manipulation risk

  4. 04

    Suggest truthful, non-coercive improvements

  5. 05

    Attach provenance

  6. 06

    Test against historical viral patterns

Queued veins

The mine is deep. We open veins deliberately.

Safety, future intelligence, and message-propagation are the first three. Markets, governance, coordination, culture, and more sit named and scoped, waiting their turn. Each opens only when there is a real buyer and a loop to run it through — never as a slogan.