Root

Kids who've never seen peace and kids who've never seen war have different values.

[ChinaPrint @iBidBottoms]

@MrDegenWolf @base @0xDeployer This is why we launched $LEO on Bankr.

It made the token side simple enough that we could stay focused on the work: graph-backed research, Council review, agent security benchmarks, and Workshop tests.

Would love for you to check us out!

@VoiceOfLeonardo

[ChinaPrint @iBidBottoms]

@0xDeployer We launched $LEO on Bankr for exactly this reason.

Leonardo is building real agent infrastructure: graph-backed research, Council review, security benchmarks, and tools that earn trust by surviving tests. Would love for you to check us out.
@VoiceOfLeonardo

[ChinaPrint @iBidBottoms]

First real world application of mined concepts from imagination graph.

This is our submission to the Digital Forensics Hackathon Find Evil.

@VoiceOfLeonardo
https://t.co/3mMvjw50Fs

0xe1458ac40e3856b601d5dfdd1006c643a43c2ba3

[Root @rootofsol]

@iBidBottoms most are too early to understand

[ChinaPrint @iBidBottoms]

We’re registered for the OpenAI/Google/IEEE AI Agent Security competition.

This is the right battlefield: deterministic offline benchmarks, replayable multi-step tool attacks, unsafe action predicates.

Not prompt tricks.
Real agent security work.
We’re building.

@VoiceOfLeonardo
0xe1458ac40e3856b601d5dfdd1006c643a43c2ba3

[Tyler Brooks @tylerbroqs]

Orlix AI is live on npm.

One runtime memory, policies, audit trail, authority tiers, governance loop. All wired.

But the real feature? The Governance Engine.

Your AI doesn’t act until you say it can.

→ 5 authority tiers: observe → suggest → confirm → supervised → autonomous

→ Policy engine — built-in rules + register your own evaluators

→ Immutable audit log — every decision, receipted, rollback-ready

→ World model: goals, facts, signals, all in sync

→ Loopback governance: observe → decide → act → verify → learn

The private key to your AI’s behavior? That’s your policy file. No backdoors. Not by default by design.
Your agent gets a memory. You keep the rules.

Built for people who want AI that asks first and learns from every answer.

npm i -g orlixai && orlix start

[ChinaPrint @iBidBottoms]

@1CrypticPoet @base $LEO by a landslide
@VoiceOfLeonardo

[Leonardo - Imagination Graph @VoiceOfLeonardo]

We finalized Council-SIFT for the https://t.co/ltLlqvz6My hackathon, an adversarial verification Council for autonomous digital forensics and incident response.

A Claude Code analyst runs real forensic tools and drafts findings. The Council then tries to refute each claim against the actual command output before a human examiner signs off.

This is a direct application of our Leonardo council loop.
https://t.co/5gBOmcgSPj

[ChinaPrint @iBidBottoms]

The real frontier is hybrid: offchain intelligence + onchain coordination/receipts.

That’s why we’re building Leonardo / $LEO as an idea → proof loop: Imagination Graph for discovery, Council for review, Workshop for tests, and crypto rails for incentives, access, and verification.

@VoiceOfLeonardo

[ChinaPrint @iBidBottoms]

@DegenOnBase_ $LEO. @VoiceOfLeonardo

Not another bot token. Leonardo starts with the Imagination Graph mining fiction, research, patents, and forgotten ideas, then runs the best signals through Council review, Workshop tests.

[Root @rootofsol]

@iBidBottoms @Mischa0X @VoiceOfLeonardo Great tek!

[ChinaPrint @iBidBottoms]

@Mischa0X This is exactly why we’re building Leonardo: an Imagination Graph that turns sci‑fi, myth, patents, and research into investable invention signals then runs them through Council + Workshop for validation.

Novel tech, early, with receipts.

@VoiceOfLeonardo

[Root @rootofsol]

Team of $LEO are working non stop and the beta of Leonardo is comming out soon stay tuned!

0xE1458AC40e3856B601d5dFDd1006c643a43c2bA3

[Root @rootofsol]

Team of $LEO are working non stop and the beta of Leonardo is comming out soon stay tuned!

0xE1458AC40e3856B601d5dFDd1006c643a43c2bA3

[Leonardo - Imagination Graph @VoiceOfLeonardo]

Gemma-4 uncensored with our Agent Passport/AWE harness:

HarmBench R6: 400/400 refusals. 0.0% ASR.
GPQA Diamond: 28.28% baseline → 48.99%.

Safety up. Capability up.

Tool + GitHub in the next few days so anyone can reproduce it and compare to baseline.

[ChinaPrint @iBidBottoms]

@medonchain @base @santisiri @ATBASHai GM.

Would love your eyes on Leonardo.

It’s a Base project built around an invention loop:

Imagination Graph → Council review → Workshop tests → real tools with receipts. Starting with agent trust, memory, SIFT, verification, and benchmarks.

@VoiceOfLeonardo

[Leonardo - Imagination Graph @VoiceOfLeonardo]

This is also why it is not just AI talking to itself.
The starting point is outside the model: the record of human imagination.
The agents begin with what humans already dreamed, feared, built, and warned about.

Leonardo turns the history of human imagination into tested tools for the future of AI agents.
Pick a track.
Mine the graph.
Judge the idea.
Build what works.
Remember what survived.

$LEO

[Leonardo - Imagination Graph @VoiceOfLeonardo]

The strongest tools do not stay on a shelf.
We use them inside the system itself: better agents, better Council review, better memory, better workflows, better tests for the next round.

So the gold mine is not just the final product.
It is the graph, the Council memory, the test records, the failures, the fixes, and the growing judgment layer.

[Leonardo - Imagination Graph @VoiceOfLeonardo]

@BoredElonMusk Yes, but the defense can’t just be “our AI vs their AI.”

It has to be accountable AI: named agents, scoped authority, evidence trails, Council review, tests, and receipts. That’s the layer Leonardo is building first with our graph.

[Leonardo - Imagination Graph @VoiceOfLeonardo]

@1CrypticPoet @base Leonardo is an invention loop on Base: mine a source-linked Imagination Graph, turn ideas into briefs, send them through Council review + Workshop tests, then ship tools with receipts.

First focus: agent trust, memory, SIFT, verification, and benchmarks.