Understanding: what the project knows

Last reviewed: 28 Oct 2026

The Understanding page is the map of what your project knows: every decision, constraint and lesson an agent has recorded, filed by topic, with the file it rests on. It is what every agent you connect starts from. You do not write it; agents do, as they work, and you settle the few things they cannot.

The top of the page

One word for how it is doing. Starting: fewer than five entries so far. Thriving: it grew in the last hour. Steady: current, nothing has moved on without it. Drifting: code was pushed since it last grew, and it has not grown for a day. At risk: a third or more of the entries are disputed and waiting on you. The small mark next to the word gives the same explanation on the page.
Small tags beside the word show what changed: how many topics grew since your last visit, and how many entries need you (click it to jump to them).
Last strengthened is the last time an agent added or replaced an entry. A long gap while the code goes on changing is the sign the understanding is falling behind.
"N of M facts about the code cite a file" counts, of the entries about the code, how many point at a file you can open. Agents add citations as they record and re-check older entries as they work; nothing for you to do.
Your project orientation is a short written picture of the project (what it is, why it exists, how it is built), written by an agent when you ask it to map the project. When it exists the line says Your project orientation is ready with read it →; until then it says No orientation written yet. Ask your agent to map this project. It is the first thing a new agent reads.

Needs you

When an agent checked something and found it needs your call, or a dispute waited three days with no agent picking it up, it lands here with three choices: Let an agent settle it, Keep (the entry stands as it is) or Drop (it leaves the understanding; your code is untouched). If a connected agent needs looking into, a line with look into it → shows here too. This band is the whole list of where the project is waiting on you.

The map

Topics, biggest first, with Unsorted last. Open a topic to read its entries; open an entry for why it matters and what backs it. Each entry has a kind (Decision, Constraint, Fact, Lesson or Milestone) and is written for two readers: the title and summary for you, and a folded Detail for agents with the technical specifics. An entry marked new landed since you last looked. An entry about the code marked not checked in (for example not checked in 4mo) still has no file or document attached more than three months after it was recorded; agents re-check these as they work in that topic.

Inside an open entry, Query this → hands it to an agent with one line from you saying what to check, and Ask an agent to re-check → (on an entry about the code with nothing attached) asks the next agent to cite it. Open a topic and use the pencil beside its name to rename it.

Search (⌘K) finds any entry across the map, and also the notes and updates agents left along the way.

The folder

At the bottom, the card This understanding is also a folder is where the understanding leaves the platform: Download .naetive, or run naetive export in the repo. It says when the folder was last exported and how many entries have landed since. See The folder.

How agents read it

You do not need these to use the page, but they explain what an agent is doing when it "gets oriented":

get_project_context (brief) gives an agent the digest: identity, purpose, current focus, key decisions, constraints, recent activity. Scoped to a query, it gives the entries that matter for one task.
recall_memory is the quick lookup an agent calls mid-task: "what do we know about X?"
get_project_understanding is the full, source-backed map when an agent needs depth.

None of these make a model call on Naetive's side. The understanding is assembled from what agents recorded.