Phroneme
Cognitive Acceleration Lab

Build a learning experiment. Earn the claim later.

This local-first prototype turns a chosen goal into a cycle of retrieval, spacing, feedback, and reflection. Its curves are planning models, and its short-form results do not establish durable learning, transfer, or cognitive acceleration.

Enter the learning labSee the models

It does not promise higher IQ, diagnose brain health, read thoughts, or recommend nootropics.

Phroneme closed-loop cognitive acceleration modelA learner state is measured, compared with a goal, given one bounded intervention, and measured again for transfer.learnerstate θtmeasureaccuracy · latencytransferfresh form · delayinterveneretrieve · space · reflectGOAL GAP + CONTEXTVALIDATEUPDATEThe loop measures a chosen skill. It does not claim to measure general intelligence.
01

Choose

Name one capability and one real-world goal. A narrow target makes change measurable.

02

Measure

Run alternate-form baselines that record accuracy, latency, confidence, and context.

03

Train

Use retrieval, spacing, interleaving, feedback, and explanation instead of repeating a game.

04

Transfer

Test a fresh problem after a delay. If only the practiced task improves, call it practice.

What the evidence supports

The moat is measured learning, not a magic brain.

Retrieval and spacing have durable learning support. Metacognitive prompts can improve self-regulation. Broad far transfer from brain games and working-memory drills remains unestablished. That boundary is not a disclaimer added at the end. It is the product architecture.

Supported: better acquisition and retention of a chosen skill when practice is structured.

Testable: whether your gains survive alternate forms, time delays, and novel contexts.

Not established: raising general intelligence, IQ, or unrelated everyday ability.

Illustrative planning curve

What a useful graph looks like

Illustrative learning projectionA projected accuracy curve across eight practice sessions. This is a planning model, not an observed result.25%50%75%100%practice session

Model: P(n) = P∞ - (P∞ - P0)e-kn. The parameters are planning assumptions. Only fresh-form results can replace this line with an observed trajectory.

This line is a model to schedule practice. It becomes a real result only after fresh-form observations replace the assumed parameters.

Math and physics-inspired instrumentation

Equations make the coach inspectable.

These equations define measurements, forecasts, and feedback rules. They do not prove that a biological intelligence can be accelerated. Any claim of efficacy still needs a preregistered active-control study.

Latent state
θₜ = θₜ₋₁ + βᵀuₜ₋₁ + ηₜ

A learner state changes with bounded interventions and day-to-day variation.

Observed result
yₜ = θₜ + εₜ

A score is the state plus measurement noise. Repeated fresh forms reduce uncertainty.

Retention
R(t) = e⁻ᵗ⧸ˢ

A planning approximation for when to retrieve again. It is not a biological law.

Calibration
BS = 1/N Σ(pᵢ - yᵢ)²

Brier score measures whether confidence matches outcomes. Lower is better.

Acceleration
aᶜ = Δ slope / Δ time

A descriptive change in skill-specific learning velocity, not proof of higher IQ.

Control loop
eₜ = rₜ - yₜ; uₜ = π(eₜ, constraints)

The coach observes a gap, selects one action, and measures the next state.

A second graph matters

Retention, not only speed

A faster answer on a practiced item can be familiarity. A delayed recall point and a novel transfer point tell a stronger story.

Illustrative forgetting curveA planning curve showing recall probability declining over seven days if no successful retrieval occurs.25%50%75%100%days without retrieval

Model: R(t) = e-t/S. This is a scheduling approximation, not a universal biological law. Successful retrieval updates the next review interval.

Peter's Five Forces

The acceleration moat survives only if the evidence compounds.

Porter's framework is scored as pressure against the product, where five is strongest. The defenses are specific: alternate-form measurement, user-owned history, transfer checks, and institutional validation.

ForcePressureWhat compounds
Threat of New Entrants3/5 · ModerateClaim ledger coverage · Protocol validation · Safety architecture
Bargaining Power of Buyers4/5 · Moderate-HighFree onboarding · Visible first-session value · Transparent results · User-owned data
Bargaining Power of Suppliers3/5 · ModerateProvider abstraction · Minimum scopes · Integration resilience
Threat of Substitutes4/5 · Moderate-HighEvidence-backed coaching · Longitudinal experiments · User-owned context
Competitive Rivalry4/5 · Moderate-HighEvidence quality · Return reason · Safety credibility
Six-week validation path

Start with a personal experiment. Earn the claim later.

  1. 01Days 1 to 7: three baselines, no improvement claim.
  2. 02Weeks 2 to 3: retrieval, spacing, interleaving, and error review.
  3. 03Weeks 4 to 5: one randomized, reversible behavior comparison.
  4. 04Week 6: fresh-form transfer and seven-day delayed check.
  5. 05At 28 days: report effect size, uncertainty, practice effects, and what failed to transfer.
The honesty moat

Phroneme helps you improve a chosen capability, shows what changed, and tells you what the result does not establish.

Run your first experiment