The Titan Protocol · Revision 7T · Full Tutorial Edition
Start small. Measure honestly.
Titan maps real human superpowers; this guide helps you enter that map through one meaningful outcome without turning the result into a judgement about your worth. It keeps a non-scoring current-state profile separate from task-specific evidence, and neither becomes a global rank, bottleneck verdict or automatic programme recommendation.
The public system
Read the full research edition. Use the public tool.
The designed 440-section browser reader contains the published Revision 7T Public Research Edition, including the preserved Capability Standard, Candidate Lab, Research Observatory, Personal Science, all 32 Power families, the 26 Gold pathways, the 256-record individual learning directory and 255 dedicated full-tutorial sections. Revision 7K is an archived predecessor retained for provenance, PWR-078 remains status-only and external review remains pending.
Public research edition
Full browser reader
Search and read every section of Revision 7T online with its persistent header and contents navigation. No full-book PDF or DOCX download is published.
Full-tutorial architecture
Individual Powers and Gold pathways
Pick any of 256 canonical Power pages, open one of 255 full individual tutorials, or follow one of 26 Gold pathways built from those same units. Thirty-five governed source lessons are preserved and each has a complete ready-to-use Power-specific practice pack; 220 additional lessons teach the named Power in plain English with 6–12 numbered steps, a worked example, right-versus-wrong comparisons, mistakes and fixes, guided practice and progress checks. PWR-078 remains status-only.
New governed method
Personal Science
Run only low-risk, reversible personal comparisons, keep null and adverse findings, and separate a private observation from scientific proof.
Planned next
Capability Commons
The planned governed collaboration environment will eventually support public corrections, evidence leads and bounded proposals. MLCA v0.2 is one candidate framework inside it.
Private candidate
46-sheet Titan Engine
The Revision 7L candidate workbook implements the same non-scoring rules, but it is not currently a public tool or download.
First 15 minutes
One safe setup, one real next action.
- Export first. If you used an earlier edition, open Data & privacy in the Cockpit and export a backup before changing anything.
- Run the Safety Screen. Apply clearance, stop conditions and accessible substitutions before performance work.
- Record only the relevant current state. Use the 24-domain profile without rating or ranking it. Reviewed, declined, unavailable, inaccessible, contraindicated, interrupted, invalid, not required and not assessed remain distinct states.
- Use the public Cockpit. It is the current route for private, browser-based records, quick logs and retest reminders. The separate Titan Engine workbook remains private and unavailable.
- Compare the available routes. Review prerequisites, evidence, likely burden and boundaries side by side. Titan will not rank the routes or choose one for you.
- Lock the declaration. Write the direct criterion, delayed retention check, novel transfer test, evidence plan, burden and stop rule before observation; lock it to receive a declaration ID and local digest.
- Run the smallest protocol. Append what happened against the declaration ID, including support, friction and any structured safety event. Never overwrite an earlier assessment.
Maintenance, declarations and safety history
Keep the level. Add the lifecycle.
Maintenance
Not a sixth level
Maintained requires a linked, in-window, materially comparable retest that still meets the declared criterion. No retest means review due or historical.
Declaration lock
Criteria before results
A local SHA-256 digest makes post-lock changes visible inside the record. It is not a trusted timestamp; external receipts remain disabled.
Safety review
See patterns across modules
Burden, adverse events, near misses, stop actions and resolutions are append-only and reviewed descriptively without diagnosis.
Review brief generator
Select before sharing
Create a time-bounded, question-scoped coach or clinician brief with the exact book location. It is educational, consent-driven, preserves withdrawal as valid and is never clearance or certification.
Fifteen current-actionable modules · one conceptual/read-only identity
Current routes exclude the held sensory-discrimination identity.
Status correction · not a runnable example
PWR-078 and MOD-SENSORY-DISCRIMINATION
PWR-078 remains held from explanatory publication and current operational use. No training, navigation, mobility, assessment, session, retest or handoff route is authorised. Read the correction.
Spatial Simulation & Cognitive Maps
Sketch-map reconstruction and controlled novel route
Landmark, route and orientation representations can be practised and scored on defined maps and controlled routes; success does not establish safe independent navigation.
Sensorimotor Intelligence & Fall Resilience
Screen-based driving hazard perception
Validated video tasks can measure latency to anticipate developing traffic hazards, and brief video-based training improved experienced drivers' test performance with the benefit still evident after at least one week. This does not establish safer on-road driving, lower crash risk or fitness to drive.
Learning, Memory & Skill Transfer
Task-specific motor learning
A visible motor sequence can be acquired and measured without claiming broad cognitive enhancement.
Metacognitive Calibration & Error Detection
Confidence-tagged quizzing
Recording confidence before feedback makes task-specific calibration and recurring error patterns measurable; it is not a general self-awareness score.
Creative Cognition & Insight
Constraint-switch prototype sprint
Separating generation from selection and testing makes the diversity, novelty and usefulness of alternatives observable on a bounded problem; it does not establish a general genius mode.
Prospective Memory & External Cues
Cue-anchored daily task
A future intention linked to one observable event cue and one fallback can be measured by timely completion and reminder burden; missed intentions are design data, not a moral score.
Quantitative Intelligence & Forecasting
Task-duration forecasting
Repeated estimates can be calibrated against actual task duration; the direction and size of bias must be measured rather than assumed.
Decision, Causal & Experimental Intelligence
Low-risk N-of-1 time-of-day experiment
A randomized or counterbalanced within-person comparison can test whether performance on one safe task differs between two feasible time windows; it does not identify a universal chronotype or prove a medical cause.
Argumentation & Dialectical Testing
Claim-and-counterargument decision memo
A short falsifiable memo can make evidence, assumptions, strongest opposition, cruxes and confidence updates auditable.
Listening, Teaching & Negotiation
Teach-back with a delayed learner test
A learner’s immediate reconstruction and delayed application can test whether one bounded explanation produced usable understanding; fluency or agreement alone is not evidence of learning.
Collective Intelligence & Team Orchestration
Independent estimates before discussion
Preserving independent judgments before group discussion can expose useful disagreement and reduce premature convergence.
Making, Repair & Tool Intelligence
Dim-sum fold or culinary technique progression
A visible making skill can be improved through repeatable quality criteria, defect review and safe transfer to a related form.
Operational Systems Design
Backup-and-restore fire drill
A backup is operationally useful only when a defined artifact can be restored and verified within the declared recovery target; one successful drill does not prove disaster readiness.
Human–AI Complementarity & Reliance Calibration
Human-only versus AI-assisted brief
A human-AI workflow is useful only when it beats the better safe baseline on a declared quality-and-cost rubric.
Assistive Technology & Person–Task Fit
Two-week assistive-technology fit trial
A time-bounded real-use trial can compare task benefit with setup, symptom, privacy and abandonment burden; availability or novelty alone does not establish person-task fit.
The evidence gates
| Level | What earns it |
|---|---|
| 1 · Baseline | Valid baseline; inconsistent performance or substantial support. |
| 2 · Supported | Improvement on the trained task under narrow or supported conditions. |
| 3 · Functional | The module-specific direct criterion is met reliably and recorded. |
| 4 · Transfer | Delayed retention plus a predeclared novel transfer result. |
| 5 · Robust | Performance holds across defined constraints with safe explanation, teaching, independent review or benchmark verification. |
Five verification modes
Record how the result was checked.
Self-report
Reported experience or outcome without a contemporaneous record, instrument, witness or independent assessment.
Self-recorded
You retained the conditions and result or an artifact. This is the generic minimum for L3 and L4.
Instrumented
A declared device, sensor or software instrument produced the measure; setup and calibration still matter.
Witnessed
An identified observer directly saw the performance and conditions.
Independently assessed
A qualified third party or externally governed assessment judged performance against disclosed criteria. This is the default for L5.
Your operating cadence
Each session
Measure the task
Record the protocol, context, direct result, support, burden and any adverse event. Stop when the guardrail says stop.
Weekly
Review the pattern
Look for reliability, data coverage and cost. Keep, adapt, pause or retire the protocol; do not defend sunk effort.
After delay
Test retention
Repeat after the module’s defined interval without quietly changing the criterion after seeing the result.
Before Level 4
Test something novel
Use a safe, predeclared task or context that was not practised. Transfer must be observed, never assumed.
Worked example
Forecasting without self-deception.
- Write ten questions that resolve within six weeks; define Yes, No and Invalid in advance.
- Record a probability and rationale for each. This is the baseline.
- Practise base rates and decomposition; update only for new evidence.
- Resolve every question and calculate Brier score. Repeat on a second date for reliable direct evidence.
- After a delay, score a fresh question set. For transfer, predeclare a different domain such as delivery estimates.
- Report calibration and misses, not just the most impressive prediction. Forecasts inform decisions; they do not guarantee outcomes.
Migration, privacy and access
- The Cockpit stores records in this browser. Export regularly; clearing browser/site data can erase unexported records.
- Exports may contain sensitive information. Protect them like a private journal.
- Adapt the method while preserving the intended task. Assistive support is recorded, not penalised.
- Readiness is an experimental pacing signal, not a diagnosis, trait or capability score.
- Governance and Research Observatory entries are never averaged with capability.
- The seven Candidate Lab modules are definitions for evaluation, not operational scores. Do not promote them merely because a protocol is interesting.
- Commons contributions will feed the Lab only after intake is explicitly activated; admission will never equal promotion.
What happens next
Run one honest cycle.
Safety → declared outcome → relevant current-state profile → compare routes → human choice → direct result → delayed retention → novel transfer → currentness review → keep, adapt, pause or retire. That is the plan. Everything else is reference material until the loop needs it.
Technical files
Developers and auditors should read the current PWR-078 status authority first. The frozen historical v27 registry and frozen historical candidate registry remain available for provenance and do not override that successor status. Current machine consumers should use the correction-status schema and current Cockpit export schema. Other lineage and method records remain available in the data and schema directories.