I'm Building Autonomous Intelligence

The source of truth was never Jira. It was always in the code.

Every time I took over a new engineering organization as CTO, the first question was the same one. And it never had a real answer.

Who is actually contributing? Who is driving value? What initiatives are truly moving forward? Where is engineering effort really going?

I’d ask in standup. Rehearsed updates. I’d check Jira. Tickets that didn’t match what was on disk. I’d ask the EM. A summary of a summary, filtered through telephone.

The truth was never in the project management tool. The truth was always buried inside execution itself.

So I started building.

From CodeInteliG to InteliG

The first version was called CodeInteliG. Developer intelligence. A way to read the engineering organization through its commits — who’s writing what, who’s reviewing what, where the work is actually happening.

Useful. Sharp. Not enough.

What I kept hitting was the same problem from a different angle. I could see what was being built. I couldn’t see why. I could read commits, PRs, contributors. I couldn’t read initiatives, strategy, the story the business thought it was telling itself.

So CodeInteliG became InteliG. Code intelligence became Execution Intelligence. The product expanded to read every layer where engineering reality lives — commits, pull requests, decisions, meetings, deployments — and synthesize them into one picture.

That’s when the category got interesting. Because once you can read execution at every layer, you stop needing the business to declare its strategy upfront.

You can derive it.

Autonomous Intelligence

That derivation is what I’m calling Autonomous Intelligence.

The idea is simple. The implementation is what we just shipped.

Traditionally, engineering organizations declare initiatives upfront — in Jira, in roadmaps, in slide decks — and then ask the team to manually map every piece of work back to them through tickets, labels, status updates, and weekly syncs. The process is fragile, incomplete, and entirely dependent on human discipline that nobody maintains.

We took the opposite approach.

Instead of asking the organization to declare its initiatives, Cognis reads the code and tells you what the initiatives are. It groups commits and PRs by branch, by file path, by conventional-commit type, by repository, by contributor cluster, by semantic relationship. It builds a three-level hierarchy — themes, initiatives, workstreams — with evidence for every node. You confirm or correct. The system learns from every correction.

No manual code-to-initiative linking. No dependency on Jira hygiene. No forcing developers to maintain metadata they don’t want to maintain.

The system reads execution and explains it back to you.

What this looks like

Here’s what Cognis just discovered about InteliG itself — without me telling it anything:

Initiative Discovery — 20 commits, 5 PRs Theme → Initiative → Workstream from raw commits. Supporting-commit attribution. Dedup against existing initiatives.

Atomic Code Linking — 10 commits, 7 PRs A 9-strategy heuristic chain replaced with a single AI classifier. Every commit links to the right initiative even without ownership rules.

Cognis Memory — 14 commits, 4 PRs Three layers — thread, user, org — persistent across every conversation. A nightly reflection agent extracts durable facts from real activity.

I didn’t declare any of those initiatives. The system found them. From commits.

What this means for engineering leadership

For decades, the question “what is my team building?” has been answered with rituals — standups, sprint reviews, status decks, OKR check-ins. All of it filtered through humans. All of it lossy. All of it stale by the time it lands on the CTO’s desk.

AI-abundant code production breaks those rituals further. You can’t sprint-plan agents. You can’t standup with copilots. The instrument that survives is the one that reads execution where execution actually lives — and reasons about it.

That’s the bet.

Imagine being a CTO or VP Engineering, signing up for InteliG, and from day one having a living intelligence layer for your organization that built itself.

Not dashboards. Not manual reporting. Not status theater.

A system that discovers, reasons, and explains what is actually happening across your engineering organization. The single greatest source of truth.

That’s what I’m building.

And honestly — I’m obsessed with it.