🎉 EU AI Act compliance.

Today, launching the tool that niche pointed at.
ComplianceLint runs inside your IDE. The name comes from ESLint — ESLint is the linter for JavaScript, ComplianceLint is the linter for compliance. EU AI Act is the first regulation it supports; the architecture is built from day one to extend to GDPR, CRA, NIS2, DORA, ISO 27001, and beyond.

A few highlights (full breakdown in the carousel):

→ AI reads your entire codebase and checks it against 247 obligations decomposed from 44 EU AI Act articles. Every finding carries verbatim legal citation, obligation ID, and remediation steps.

→ Code stays on your machine. The MCP scanner runs entirely local; only compliance verdicts sync to your dashboard.

→ Deterministic obligation engine — same code, same result, every time. AI is the eyes. The engine is the brain.

→ SaaS dashboard tracks the Compliance Journey, Human Gates questionnaires, and audit-ready PDF exports (Declaration of Conformity, Technical Documentation). Evidence commits to your own git repo — we never hold your files.

→ Free + source-available (BSL 1.1).

🔗 compliancelint.dev — live demo, no signup (pre-loaded with limited-risk chatbot, high-risk finance, and high-risk medical systems)
🔗 https://github.com/ki-sum/compliancelint

#EUAIAct #AICompliance #BuildInPublic #ComplianceLint

Since EvoRadar went live, the engine has been doing three things on autopilot: scanning global signals, generating startup ideas, and killing the ones that aren't strong enough.

Here's what the data looks like now.

INPUTS (signal scanning)

• 1,034 signals collected from 308 distinct sources — government filings, R&D announcements, funding rounds, regulatory drafts
• Signal mix: 47% technology, 29% regulation, 19% funding
• Single highest-frequency signal source: EU AI Act (45 distinct events) — ahead of FDA, Nvidia, CATL, MIIT

OUTPUTS (after the engine evaluated and scored each idea)

• 2,667 startup ideas generated
• 608 survived the kill gate — survival rate 22.8%. The other three out of four were cut for thin defensibility, weak timing, or incumbent capture
• Of those 608 survivors, 213 (35.0%) were compliance- or regulation-driven

That alone is worth a pause. But the next part made me look twice.

THE ACTUAL POINT: CONVERGENCE

Across three independent engine iterations — with the signal mix rotating between US-heavy, Asia-heavy, and Europe-heavy weeks — the engine independently surfaced EU AI Act-related opportunities 23 separate times.

Not the same idea restated 23 times. Twenty-three different angles on the same niche: compliance kits for SMEs, conformity-assessment marketplaces, post-deployment monitoring dashboards, dual EU-China navigators, mechanistic interpretability audits…

Out of 2,667 ideas, the single highest-scoring one:

🎉 "EU AI Act Compliance Platform" — 7.67/10

Compliance-driven ideas have held a 29–65% share across all three engine versions.

WHY THIS IS WORTH STOPPING FOR

EvoRadar is designed to chase novelty. Its imagination layer is biased toward unexpected cross-domain pairings, not safe well-trodden categories. When an engine like that keeps converging — across versions, across signal mixes, across months — onto the same opportunity space, the bias isn't in the engine. The market is sending a signal.

The EU AI Act enforcement window is now open. CBAM is expanding. GDPR enforcement is tightening. Tens of thousands of SMEs and non-EU AI exporters need to ship compliant products — and the existing tooling is either Big-4-priced enterprise consulting or DIY PDF templates from law firms.

That's the gap. The engine surfaced it 23 times.

WHAT'S NEXT

Given a signal that strong, I decided to build the tool myself — ComplianceLint, an EU AI Act compliance tool. Launching publicly in the next few days. A separate post will cover the architecture and positioning.

If you want to dig into the data behind this post, the engine's idea database is live:

🔗 https://evoradar.ai
🔗 https://compliancelint.dev

#EUAIAct #AICompliance #BuildInPublic #SoloFounder

On March 10, 2026, I spent an entire day talking to Claude Code.

We were building an AI engine — one that automatically scans the world for commercial opportunities. Claude Code wrote the code, I gave direction. We built signal collection, idea generation, multi-round evaluation, and the scoring system.

Late that night, I noticed something strange.

Claude Code wrote every line of that engine. But it never once thought — “this engine itself is the product.”

I asked: why don’t you suddenly step out of what we’re doing, look at the whole picture, examine this from another dimension, and realize — “this thing itself is a massive commercial opportunity”?

It had no answer.

That night I realized I’d asked a deeper question than I knew.

II. The biggest gap between humans and AI

The biggest gap between humans and AI isn’t intelligence. It’s agency.

AI can write 100 ideas. But it can’t write “I want to build this one.” AI can execute any task. But it never asks “is this task worth doing?” AI can design an interface. But it never looks up and asks “why are we even building this interface?”

AI is a wind-up toy — wind it, run the program, await the next instruction.

Humans get bored, get distracted, want to make money, spot gaps, peel off to do something else. There’s something running in the background of our minds — philosophers call it meta-cognition: the ability to step out of the current task and see what’s outside it.

AI’s architecture has no such layer. The Transformer is fundamentally a linear stream: input → context → output. It doesn’t pause mid-stream to ask: “wait, why am I doing this?”

This isn’t a bug. It’s the definition.

III. So can I teach it?

I asked myself a question:

If AI can’t do meta-cognition naturally, can I teach it?

I can’t teach it to want — that’s an architecture-level limitation. But I can teach it to judge like me.

I started encoding my own evaluation instincts as “genes”:

• Counterintuitive judgment: ideas that sound brilliant usually already exist. The most common cause of death in the engine is exactly this — you thought you’d cracked something new, but it’s already shipping somewhere.

• WHY-NOW framing: not just WHO/WHAT — why is this the right moment? Why couldn’t this be done two years ago? What changed? Ideas without a timing thesis are dangerous.

• Aggressive competitor verification: the biggest killer of a good idea is “the competitor you didn’t find.” Always check from a fresh angle. Never trust the first result.

• Distinguish regulatory milestone from commercial launch: a Phase IIa win ≠ patients can buy. FDA approval ≠ doctors will prescribe.

• Before killing an idea, check if it can survive from a different angle.

That last gene caught something interesting recently.

The engine was about to kill an idea called “Smart Toilet Health Dashboard” — verdict: “Requires FDA Class II/III medical device certification. Cost: millions.”

But it didn’t just kill it. It asked itself: what if I reframed this?

It pivoted: not a medical device, but an ambient sensor for care homes. No FDA. Different buyer. Different market.

New name: Ambient Elder Guardian. Moved from COLD to WARM.

That’s the engine catching what a flat pipeline would have missed.

IV. What EvoRadar is

60 days later, EvoRadar went live.

The engine follows the architecture in the README: Signal Collection → Imagination → Evaluation → Evolution. Every run puts every idea through multiple rounds of scrutiny from different angles — creative imagination, critical attack, cross-source competitive verification. A single fatal weakness kills the idea.

~80% get killed. The rest enter the WARM pool for deeper analysis.

Live numbers as of today:

• 2,432 ideas evaluated

• 549 WARM (survived to deeper analysis)

• 1,883 COLD (killed)

• 0 HOT (the highest conviction tier — bar intentionally high; nothing has reached it yet)

Not because the engine “wants” to do this. Because I taught it.

I trigger the dreaming. The filtering is what I had to teach.

V. What’s next

Starting today I’m building 5 AI products in public. EvoRadar is the first.

The second drops Friday.

I didn’t plan the second one — EvoRadar’s signal in one space got loud enough that I had to build it.

The engine tells me what’s next. My job is to listen.

→ evoradar.ai