Try the learning layer.
No account.
Chat with a small Adaptive Learning Model under the Advanced Epoch Wrapper. Every turn shows its runtime, routing, memory state, curriculum context, and evaluation trace, the same surfaces a private deployment exposes.
- accounts required
- 0
- in-browser option
- 100%
- runtime tiers
- 3
- curriculum domains
- 6
What to watch
Every turn is traced
through the wrapper.
The console on this page runs each message through these layers. In a private deployment the same trace is exported for review.
- 01
A learner asks a question.
- 02
The wrapper verifies identity and session.
- 03
The relevant curriculum context is attached.
- 04
Memory permissions are checked (off · session · vault).
- 05
Retrieval scope is selected: curriculum-scoped, not the open web.
- 06
Safety and curriculum guardrails are applied before generation.
- 07
The router picks the lowest-power runtime that meets latency: on-device, edge, or eco server.
- 08
The smallest capable model responds.
- 09
An evaluation trace scores clarity, correctness, and pedagogical fit.
- 10
A turn that falls below threshold is flagged for human review.
The lowest-power runtime
that still meets latency.
Eco Foundry runs on a hybrid of on-device models and energy-efficient eco servers. The same router shown in the console above chooses the least costly path for every turn, which is how the platform stays free for under-served learners.
Lowest-power routing
The router you watch in the console picks the smallest runtime that still meets latency: on-device for light work, eco servers for heavy lifting.
GPU power-capping
Heavy inference runs under power caps, trading a little speed for a large drop in energy on the jobs that need it least.
Carbon-aware scheduling
Non-urgent batch work shifts toward cleaner grid windows, so the same output costs less carbon.
Scale down when idle
Idle capacity scales toward zero instead of burning power while it waits for traffic.
Map understanding. Estimate uncertainty.
Route to the smallest capable model.
Pick a domain. The network maps the learner across the curriculum graph, finds the frontier concept whose prerequisites are met, estimates uncertainty, and selects the runtime to teach it. This is a product preview over sample state.
Public without sign-up.
Private through a cohort.
We communicate the gate up front: everything visible here is public. Persistent and institutional features unlock when your cohort is accepted into the Free Compute Program.
Public · no sign-up
AVAILABLE NOW- Run a real small model in your browser via WebGPU
- Or chat with the deterministic, simulated tutor
- Laplace Learning Network curriculum-graph explorer
- Live runtime detection that reads your browser's real capabilities
- Advanced Epoch Wrapper trace on every turn
- Sample curriculum graphs across six domains
Private · cohort required
COHORT ACCESS- Eco-server inference at classroom and cohort scale
- Persistent student memory vaults
- Private cohorts and educator dashboards
- Custom curriculum uploads and model distillation
- Institutional policies and compliance workflows
- Research telemetry exports