Wethic

Wethic: AI-native education. Gemini builds workforce-aligned courses, creates personalized micro-lessons to close cognitive gaps, and deploys graduates into real operational roles.

Comment

Inspiration

We were tired of the lie that a "certificate of completion" means someone can actually do the job. Most courses are static: a fast-but-wrong learner and a slow-but-right learner receive the exact same next lesson, because the platform ignores the most valuable signal it already has — how the learner absorbs. Employers told us the same thing: graduates of online platforms are not deployable. They watched, they clicked, but they cannot operate.

So we built a learning system that behaves like a nervous system instead of a library: it feels the Operator in real time, classifies their cognitive state, and reacts with surgical precision — not by rewriting the course, but by minting a learning intervention unique to that person. We named it Wethic, and everyone on it is an Operator, not a student.

What it does

ethic is an AI-native learning ecosystem where education ends in real deployment, not a certificate. It runs on TutorLMS, BuddyPress and Elementor, with a Gemini Enterprise bridge to Google Cloud Vertex AI at its core.

Every quiz becomes a Synaptic Assessment producing two measurements: accuracy, and time deviation against a reading baseline:

$$t_{b} = \left\lceil \frac{W}{200} \cdot 60 \right\rceil \ \text{s}, \qquad \delta = \frac{t - t_{b}}{t_{b}} \times 100\%$$

When an Operator fails — score $s < 70\%$, or a fault quadrant ($\delta > 10\%$ slow, or fast-and-wrong = FALSE_CONFIDENCE) — the Synapse Orchestrator mints an exclusive, temporary Deep Sync Node micro-lesson via Gemini Enterprise, tailored to that failed quadrant, and gates the next core Knowledge Node until the gap is bridged. Courses stay 100% static; personalization happens around them.

The Node Factory generates Knowledge Nodes aligned with labor-market demand; the Synapse Pulse copilot gives Milestone Logs, Proactive Guidance and Gap Analysis; the Synapse Map shows synchronization % per Octagon category; and the Deployment Grid lets Operators initiate real deployments on Host Nodes once their Knowledge Nodes are absorbed. Keystone gates the whole site, and Subscription monetizes it (Stripe: $1 trial → $14.99/mo) with a live MRR/churn dashboard.

How we built it

Ten custom WordPress plugins speaking one dialect (Octagon categories, Absorption States, ISO codes, Nodes), enforced by a shared utility toolkit:

  • Wethic Gemini Bridge — the secure connection layer between the Learning Core and Gemini Enterprise on Google Cloud; one client shared by the ecosystem.
  • Wethic Keystone — full identity gate (/login, instant registration, passkey reset); site 100% closed to unauthenticated users.
  • Wethic Synapse Orchestrator — telemetry layer (server-side reading-time tracking vs. WPM baseline), the 4-quadrant matrix, and the cron-decoupled Deep Sync Node minting pipeline with dedupe, HTML sanitization and a shared-variant cache that reuses identical transmutations instead of paying a fresh Gemini call per Operator.
  • Wethic Node Factory — autonomous Knowledge Node generator aligned with labor-market demand, feeding TutorLMS directly.
  • Wethic Synapse Pulse / Synapse Map / Deployment Grid / Mine / Course Aesthetic — the copilot, the BuddyPress synchronization map, the deployment board, course display, and the dark glassmorphism identity.
  • Wethic Subscription — Stripe checkout, trial-to-monthly conversion, site locking and the revenue metrics admin.

The hardest constraint was latency: Gemini generation can take 60+ seconds, so minting runs decoupled from the request via cron, with caching and pruning to keep per-Operator cost low.

Challenges we ran into

  1. Personalization inside a static LMS. TutorLMS gives you lessons and quizzes, not cognition. We built the telemetry layer ourselves — tracking lesson reading time server-side, computing deviation against the baseline, and fusing it into an absorption state — without touching TutorLMS internals.
  2. Making AI production-safe. Gemini writes real learning interventions, so every payload is deduplicated, sanitized and gated by the failure-quadrant rule before it reaches an Operator. Nodes expire if not bridged, keeping the system honest.
  3. Latency as a product constraint. A 60+ second AI call cannot block a learner's request. We decoupled generation to cron and built a shared-variant cache so identical transmutations are paid for once and reused safely.
  4. Business model before traffic. We shipped monetization on day one — paid trial via Stripe, site locked without a plan — because an unsustainable platform will not scale its impact.
  5. Coherence across ten plugins. One wrong term in one plugin would corrupt the matrix math. A single shared utility toolkit (baselines, thresholds, state registry) keeps every subsystem reasoning with the same vocabulary.

Accomplishments that we're proud of

  • A single physiological number for learning. Absorption Velocity $v = s - \max(0,\delta) \cdot 0.2$ (clamped to $[0,100]$) condenses an entire cognitive state into one 0–100 number that the whole ecosystem reasons with — maps, copilot, deployment grid, admin matrix.
  • Detecting FALSE_CONFIDENCE. A fast, careless pass is treated as a fault quadrant and gated — the metric most platforms never look at.
  • A fully closed, monetized product in production. Paid Stripe trial from day one, live MRR/churn dashboard, no free-tier leak.
  • AI that changes the learning path, not just the content. Deep Sync Nodes are exclusive per Operator, expire, and gate progress until the gap is actually bridged.
  • The Deployment Grid: the first time on our platform that "graduation" became an operational act — deploying on real Host Nodes.

What we learned

Learning is a 4-quadrant problem, not a score. A passing score is not retention — and the quadrant math proved it to us:

$$\text{state}(s,\delta) = \begin{cases} \text{OPTIMAL_SYNC} & \delta \leq 10,\ s \geq 70 \ \text{FALSE_CONFIDENCE} & \delta \leq 10,\ s < 70 \ \text{COGNITIVE_FRICTION} & \delta > 10,\ s \geq 70 \ \text{COGNITIVE_OVERLOAD} & \delta > 10,\ s < 70 \end{cases}$$

We learned that Gemini generates pedagogy best when prompted with structure, not prose: system instructions that define failure modes, forcing functions, recall drills and verification gates, with the failed quadrant ISO appended to every generation. And we learned that an AI-native business means AI executing decisions in production — minting nodes, gating progress, updating the matrix — while humans curate the system instructions and the standard.

What's next for Wethic

  • Scale the Octagon. Eight categories today (Core Architecture, Strategic Influence, Cognitive Engineering, Hardware Optimization, Neural Interface, Wealth Protocols, Mass Resonance, Systems Intelligence) → deeper node graphs and specialization tracks per category.
  • Open the Deployment Grid. Real Host Node partnerships with employers and operators: absorbed Knowledge Nodes map directly to deployable roles with live demand data from the Node Factory.
  • Multi-agent Gemini orchestration. The Bridge evolves into an agent fleet — one agent curating nodes, one diagnosing quadrants, one coaching in the Pulse — supervised by a single orchestration loop.
  • Multimodal learning signals. Extend telemetry beyond time: attention, typing cadence, and adaptive quiz design driven by the matrix.
  • White-label the ecosystem. The ten plugins as a licensable "learning OS" for academies, cohorts and enterprise teams, each running their own Octagon.
  • Research-grade validation. Publish the Absorption Velocity metric and quadrant model with retention data — we believe it belongs in the standard metrics of online learning.

Built With

Share this project:

Updates