Mandatory Status
The AI Portal lane is Alpha/Experimental.
It must not be described as production autonomous memory, an independent source of project truth, or a safety boundary that can replace source, runtime, HELP, metadata, contracts, review, or human authorization.
Purpose
The AI Portal is the AI-facing entrance to the Laboratory Campus. It is being designed to prepare an AI for a specific series of project tasks without requiring the model to ingest the entire ecosystem or depend on lost chat history.
Its intended path is:
task
-> owning project and lane
-> curated, typed jumps between artifacts
-> authority, freshness, risk, and proof checks
-> bounded task context packet
-> guarded action plan
-> proof readback and durable closeout
The working metaphor is frontal memory. The implementation must remain inspectable: every included artifact should identify why it was selected, what authority it has, what evidence supports it, and what remains missing.
What Exists Today
- repo-local AI orientation and assimilation documents;
- LabTalk project, app, lesson, concept, proof, and portal registries;
- a local registry browser and launcher;
- read-only
MAINT AIorientation commands; - runtime transcripts and truth-audit reports;
- the APH-0 through APH-6 hardening plan.
These pieces are useful but do not yet constitute validated autonomous project memory.
Hardening Gates
| Gate | Outcome |
|---|---|
| APH-0 | Preserve the lane, remove unexplained drift, make startup reproducible, and establish tests. |
| APH-1 | Turn flat registry metadata into a validated graph of typed, directional synapses. |
| APH-2 | Compile deterministic, bounded, read-only task context packets—the first useful MVP. |
| APH-3 | Add reviewed capabilities, risk classes, approvals, containment, cancellation, and proof transcripts. |
| APH-4 | Make task preparation and jump explanation the primary portal experience. |
| APH-5 | Curate the graph and packet history in x64base with deterministic portable exports and recovery. |
| APH-6 | Turn portal behavior into DotTalk++ lessons and measure cold-start task performance. |
Safety and Authority
- Read-only preparation is the default.
- A registry record cannot grant itself execution authority.
- Missing or contradictory evidence must be visible before action.
- Mutating operations require explicit, separately reviewed capabilities and approval.
- A generated packet is working context, not a new source of runtime truth.
- Website summaries remain downstream reviewed derivatives.
Campus Role
The AI Portal is both an engineering tool and a lesson. Its graph records, context budgets, validation failures, safety gates, transcripts, and evaluation results should become inspectable DotTalk++ and Laboratory Campus material.
This makes AI participation part of the same co-development loop as engine, documentation, proof, and curriculum work—without pretending the experimental portal is already complete.
See Mission & Vision, Laboratory Campus Overview, Recursive Co-development, and Current Work Lanes.