Cognis is the reasoning engine inside InteliG: evidence-backed verdicts that connect code, people, strategy, and cost. This release puts that reasoning on the surfaces you open every day, gives the platform a home that looks like what it does, and tightens how code links to strategy.

Cognis Overview Verdicts

Every pillar overview now leads with a verdict from Cognis, not just a wall of numbers.

A verdict per pillar. Cognis reads your execution over the selected window and states how that part of the organization is running, with the evidence behind it.
Quiet when it should be. When the evidence does not support a strong call, Cognis abstains. A verdict appears when it is earned.
Aware of your team size. Verdicts are calibrated to the people actually contributing, so a small team is read as a small team.
Scoped to your window. Overviews default to the last 30 days and share one date filter. Verdict, evidence, and numbers all move together when you change the period.

Cognis concern radar

Ask Cognis what your team is worried about, and it answers from what was actually said.

One question, the whole org. Cognis pulls the concerns raised across your meetings, attributes each one to the person who raised it, links the source meeting, and groups them by theme.
Evidence and inference stay separate. Anything Cognis reads from a transcript is marked pending review and is never stated as fact until a human confirms it. Confirmed and unverified sit in their own columns.

The Brain

Your home is now a 3D Brain: the map of how your organization executes, and the way you move through the platform.

A home that represents the work. It sits at the top of navigation as your post sign-in home, each region mapped to a part of how the organization runs.
Click a region to go there. Select a region and the Brain takes you to the matching pillar.
Activity at a glance. Regions light up where work is happening. Native InteliG data shows as memory, connected tools as live perception.

Cognis reaches further

Cognis already searched commits, pull requests, decisions, and sprints. Now it answers across the rest of your surface: contributors by name and activity, action items by keyword, branches by staleness, per-repository stats, and themes rolled up to initiatives.

Cognis memory and people

Cognis now keeps a memory of how your organization executes, and grounds its answers in it.

A spine of execution facts. Cognis records what happened as durable facts: who reviewed, who merged, who attended, who owns what.
People resolved into one roster. Identities across your connected sources resolve into a single set of people, with derived role and relationship, so contributor answers and team size rest on the same roster.
One definition of team size. Team size is drawn from the humans actually contributing, and the overview verdicts read against it.
It keeps itself current. A nightly sweep refreshes the memory, and a newer statement supersedes an older one when they conflict.

Sharper linking

Proposals are verified before they stand. Cognis link proposals pass a grounded verifier, and a re-verify sweep holds the existing backlog to the same bar.
Your explicit tags win. Tag a commit with its initiative ID, such as (INI-57), and that tag takes precedence over inferred ownership.
A coverage funnel. Your work breaks into Linked, Suggested, Unsuggested, and Needs-initiative, with a remediation tab to link the unlinked majority in place.
Linking that calibrates itself. Links are gated on a measured confidence bar, abstain below it, and the bar recomputes monthly as the system learns from your corrections.

Sharper discovery

Discovered initiatives own their commits. Confirm a discovered initiative and it derives its ownership patterns, so future commits attribute to it without manual linking.
Reconciliation with calibrated confidence. Discovery reconciles candidates against your declared initiatives, and the cluster engine reads repository-qualified paths and scope tiers for tighter boundaries.

Also in this release

CTO Briefing groundedness guard. A fail-closed check holds back any brief that the data does not support.
Findings that earn their place. Findings pass a materiality and de-duplication gate at the point they are written, and an empty Cognis search result says so plainly instead of reading as an absence of data.

Why it matters

The product states a verdict, not just a chart. Every overview opens with a Cognis reading scoped to your window and team size, and stays quiet when the evidence is thin.
The home looks like what the platform does. The Brain turns the home screen into a map of execution.
Linking gets more defensible. A verification pass on every proposal and explicit-tag precedence make the strategy model harder to argue with.
Cognis remembers. A memory of who did what, resolved to one roster of people, grounds the answers and verdicts Cognis gives.
It knows what it heard versus what it knows. The concern radar surfaces what your team is worried about and marks anything read from a transcript as pending review until a human confirms it. That line between evidence and inference is the difference between an AI that is useful and one you can trust.

Release confidence

Scope. Backend and frontend, main branch, deployed June 2, covering work since the May 26 release.
Evidence. More than sixty merged pull requests across Cognis overview verdicts, the concern radar and execution memory, the Brain, the new search tools, initiative linking, and discovery.
Validation. Green in CI on the shipped build, with the release and tenant-isolation audit both passing. Schema changes applied additively.
Confidence. High for overview verdicts, the concern radar, the search tools, and linking. The concern radar holds interpretive facts as pending review until confirmed. The Brain is the newest surface, and we are continuing to deepen what each region reflects.

Cognis Overview Verdicts, the Brain, and Sharper Linking