githubnext/ace2Public

Ace: a collaborative coding environment on the team's own machines

Design: tailor explanations to concepts a participant marks as known

IA

**Design proposal with open questions. This is not a settled implementation.**

Follow-on to the explainer pane (#100). Let a participant explicitly mark concepts they already understand. A participant-controlled knowledge profile would then tailor future explanations, written logs, and rich context, so that Ace stops re-teaching things the participant already knows.

## Proposal

- Mark a concept in the explainer pane as understood (or not understood) explicitly. Ace never infers this silently. - A knowledge profile records those marks, and the participant can view, edit, and reset it at any time. - Future explanations, plain-language progress accounts (#99), and rich context use the profile to skip or shorten familiar material and expand unfamiliar material. - Personalization changes only how much is explained. Progress, outcomes, blockers, and the audit trail stay visible no matter what the profile says.

## Constraints

- This is not an account or access-control system. Ace has no accounts, and team membership comes only from the tailnet (see docs/terms.md). A participant's profile is a personal preference store and grants no access. - The participant can see and control everything that is stored, with no hidden memory.

## Open questions

- Where does the profile live (per host, per participant identity, or per project) and how does it sync across a participant's hosts? - How are concepts identified and deduplicated across channels and projects (e.g., free text, tags, links)? - Should marks expire or decay, and can a participant say "explain this again" just once? - Does a profile ever affect what other participants see in a shared channel? Default proposal: no. Tailoring applies only to the person viewing it. - How is the effect visible ("skipped: you marked X as known"), so that missing explanations are never confusing? - Does the agent's own context use the profile, and how is that kept within the participant's control?

## Acceptance (once the design is settled)

- Mark a concept as known, then see a later explanation that omits or shortens it, with a visible note saying so. Unmark or reset it and see the full explanation return. - Progress, outcomes, blockers, and the audit trail are identical with and without a profile.

Depends on #100 (and #99). Part of #98. Product feedback from Nate.

  1. iamnbutlerreferenced this in [Meta] Help participants follow and understand agent work #98last month on GitHub

  2. iamnbutlerreferenced this in Add an optional pane that explains what the agent built and why #100last month on GitHub

  3. iamnbutlerparent-issue-addedlast month on GitHub