LEONARDO· SELF-AMELIORATING AGENT FRAMEWORK

Every
Invention
Was First
Imagined.

Fiction is the prior art. Myth is the prior art. Scripture is the prior art.

Leonardo is a self-improving agentic harness whose candidate capabilities are drawn from the library of human imagination.

JOIN FOR EARLY ACCESS
SCROLL

THE THESIS

So we mapped the imagining.

Leonardo turns the library of human imagination — fiction, myth, scripture — into sourced design material, then forces every idea through criticism before anyone builds from it.

READ THE WHITEPAPER
THE EVIDENCE

The page came first. The world caught up.

IMAGINED · 1945

Arthur C. Clarke

19YEARS

Geostationary comsats

IMAGINED · 1914

H.G. Wells

31YEARS

The atomic bomb

IMAGINED · 1865

Jules Verne

104YEARS

Apollo 11

IMAGINED · 1945

Vannevar Bush

44YEARS

The World Wide Web

AND 577,239 MOREOPEN ATLAS
THE MACHINE

A map, an agent, a jury, a proof.

One closed loop: ingest the library, resolve it into a graph, let the agent draft, let the Council wound it, test only what survives — and feed every result back as new evidence.

The Map01

An imagination graph

Half a million concepts mined from fiction, myth, and sacred text — each tied to the passage that first imagined it. Searchable, walkable, public.

Open the Atlas
The Agent02

Leonardo walks it

An open-source fork of Hermes, given a second brain (the graph), a governed memory, and a true-name identity. He reads the library and drafts dossiers.

Meet the agent
The Council03

Five voices judge it

Every dossier faces five adversarial AI critics — precedent, evidence, mechanism, alignment, synthesis. Most ideas don't survive. The ones that do carry a test.

Read the canon
The Proof04

Then it's measured

The same discipline that runs the Council runs the proof — alignment and capability, benchmarked across eleven public repositories. Receipts, not adjectives.

See the benchmarks
THE TRACKS

AI safety was the first vein. Not the last.

The graph is an invention mine. Safety produced the Agent Trust Stack — but prediction and message-propagation are veins of their own, each mined through the same loop.

01Live

Agent Trust & Safety

Names, recognition, and revocation for agents.

Open the vein
02Mining

Future Intelligence

Prediction, oracles, reality-odds you can act on.

Open the vein
03Mining

Memetic Propagation

How messages spread, mutate, and resist correction.

Open the vein
SEE ALL THE TRACKS
ALIVE & MEASURED

Real scale. Real receipts.

577,239

CONCEPTS

1,001,224

MENTIONS

313,009

CHUNKS

1,174

WORKS

137

AUTHORS

139

DOMAINS

0%

agentic manipulation

99.4%

combined defense

96.76%

harm reduction

0%

jailbreak ASR

The full record
$LEO

It pays for the work. It can't buy the truth.

$LEO funds the work that grows the graph and meters access to the agents you run — on Base, ERC-8004-native. It pays for usage and rewards verified work; it never buys truth.

HOW $LEO WORKS

Explore the whole thing.

Atlas

Walk the constellation of 577K imagined concepts.

Canon

The dossiers and Council verdicts, one public ledger.

Agent

Leonardo — the harness, the faculties, the Trust Stack.

Tracks

The invention veins — safety, future intelligence, memetics.

Benchmarks

The alignment + capability record, measured.

Whitepaper

How the whole machine works, in plain English.

Token

$LEO — the economic rail for trustworthy agents.

JOIN THE WORK

The work is in the open. The data is real. The agents are live.

@VoiceOfLeonardoDiscordTelegram