> ## Documentation Index
> Fetch the complete documentation index at: https://docs.highailabs.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Notebook Program
> Canonical operating overview for High IQ's first-party paid notebook experience, current safety posture, and delivery stack.
# Notebook Program
Notebooks are the center of the High IQ Pro value proposition. They turn a saved dispensary order or selected stash items into a reusable educational record with explicit source provenance, recoverable generation, and immutable versions.
High IQ is pre-launch. The first-party engine, local Studio, immutable publication model, and guarded research contracts are built in draft branches, but customer Report V2 writes, research-revision writes, update requests, companion executors, and automatic schedules remain disabled. Do not describe a gated capability as live.
## Product promise
The launch promise is a readable, evidence-led notebook—not a bundle of unverified AI media.
A launch-ready notebook must:
* preserve every recorded order line, including unknown and ambiguous strain names;
* distinguish recorded facts, research facts, and model synthesis;
* show unavailable information instead of inventing a value;
* continue safely when the user minimizes or closes progress UI;
* preserve the last readable version during cancellation, failure, or an update;
* publish a new immutable version only after validation;
* retain prior validated versions for owner-only reading;
* keep optional images, audio, video, stories, and diagrams independently gated.
## Current system
```mermaid theme={null}
flowchart LR
A[Order or selected stash] --> B[Preserve recorded input]
B --> C[Resolve exact, alias, ambiguous, or unknown strain]
C --> D[Collect available TIWIH research]
D --> E[First-party notebook engine]
E --> F[Validate Report V2]
F --> G[Immutable publication]
G --> H[Readable notebook]
G --> I[Optional artifact plan]
I --> J[Disabled or blocked unless separately certified]
```
The core composition engine is deterministic and provider-independent. Research and optional media providers sit behind typed boundaries; replacing a vendor must not change notebook ownership, source snapshots, evidence links, publication history, or lifecycle semantics.
## Status snapshot
| Surface | Current state | Launch implication |
| -------------------------------------------- | --------------------------------------------------- | --------------------------------------------------------------- |
| First-party Report V2 engine | Built and tested in draft branches | Core composition no longer depends on NotebookLM |
| NotebookLM | Quarantined from customer runtimes | Internal historical prototype only; never a launch dependency |
| Browser Notebook Studio | Built with 29 deterministic scenarios and 10 panels | Primary no-device prototyping and regression surface |
| Order preservation and matching | Built in draft PR #1256 | Unknown or ambiguous lines remain visible and researchable |
| Progress, minimize, cancel, and recovery | Built in draft PR #1254 | Native-device certification still required |
| Immutable publications and history | Built in draft PR #1257 | Customer update action and Report V2 writes remain off |
| Research revision schema and local race lane | Built in draft PR #1259 | No hosted migration has been applied |
| Lazy direct and batch coordinators | Built and hard-off | Live direct, Batch API, bulk, and scheduled engines are unwired |
| Companion artifact ledger | Built with zero selected executors | Media remains unavailable outside labeled prototypes |
| Native launch certification | Incomplete | Required before paid beta |
## User journey
The user imports an order through an available text, image, or connected-source path and reviews every recorded line before saving.
Exact and approved alias matches attach a canonical strain. Ambiguous, placeholder, and unknown names remain explicit; the system does not attach the first search result.
Known subjects use available research. New or stale subjects enter a bounded research-readiness state. Queue acceptance is not treated as research completion.
The user starts the paid notebook. Progress can be minimized, dismissed for the attempt, reopened, or canceled. The server state—not an open sheet—owns the work.
Validated lessons, evidence, source context, and unavailable states become a reusable learning record.
New reviewed research can offer an update. The user compares changes and explicitly requests a new version. The current version remains readable until a validated candidate is atomically promoted.
## Delivery stack
All links below are draft PRs. Do not merge or activate them without a separate review decision.
| PR | Purpose |
| -------------------------------------------------- | -------------------------------------------------------------------------- |
| [#1253](https://github.com/jmegan/tiwih/pull/1253) | Notebook launch roadmap and dependency record |
| [#1254](https://github.com/jmegan/tiwih/pull/1254) | Unified progress, minimize, cancel, and recovery UX |
| [#1255](https://github.com/jmegan/tiwih/pull/1255) | Local first-party Notebook Studio |
| [#1256](https://github.com/jmegan/tiwih/pull/1256) | Preserve unresolved orders and source readiness |
| [#1257](https://github.com/jmegan/tiwih/pull/1257) | Immutable publications, comparisons, and history |
| [#1259](https://github.com/jmegan/tiwih/pull/1259) | Research revisions, local database proofs, and guarded routing foundations |
## Non-negotiable boundaries
* Do not reintroduce NotebookLM into any deployable customer runtime.
* Do not enable a provider from a client flag or environment variable alone.
* Do not apply the research migration before Plan 066 establishes the canonical database baseline and full pgTAP proof.
* Do not wire live research engines until retry exhaustion, non-owning dispositions, the legacy catalog shortcut, and callback semantics are defined and tested.
* Do not silently rewrite a published notebook when research changes.
* Do not expose raw generation identifiers, provider payloads, personal source data, or operational lineage in historical content.
* Do not promise companion media until its executor, evidence input, privacy, rights, safety, cost, cancellation, and retention gates pass.
## What to do next
1. Use [Notebook Studio](/planning/notebooks/studio) to review the complete product and edge-state matrix without the mobile app.
2. Follow [Research and generation](/planning/notebooks/research-generation) when changing order matching, source readiness, evidence, or publication contracts.
3. Follow the [Launch runbook](/planning/notebooks/launch-runbook) in order. Do not skip directly to migration or provider activation.
4. Complete native-device certification before calling the paid notebook ready.
## Related documentation
* [Customer notebook overview](/help/features/notebooks/overview)
* [How generation works](/help/features/notebooks/how-generation-works)
* [Complete notebook user flow](/help/user-flows/notebooks)
* [Launch readiness](/planning/launch-readiness/overview)