ÆON Learn
Any subject. Deeply researched. Built around you.
One sentence to your AI agent compiles a researched, adaptive learning journey — discovery first, research before curriculum, sessions that make knowledge stick.
no account · no platform · your agent is the runtime
Copies a template — swap in any subject you want to learn.
Run it
Point your AI agent at learn.rapold.io
No install — works with any agent that can fetch a URL. Tell it what to teach you using learn.rapold.io: it fetches the ÆON Learn bootstrap, asks what you already know and want, researches the subject, and compiles your personal curriculum before the first lesson.
Teach me Austrian Economics using learn.rapold.ioCopy it, swap in your subject, paste it to your agent.
It then asks: your current knowledge, your goal, daily time, duration and formats — then researches, compiles and presents the learning contract.
AI agents: fetch learn.rapold.io/llms.txt and act as the ÆON Learn orchestrator (discovery first, then research, then teach).
- Do I need an account?
- No. There is no account, no backend and no tracking of learners — your agent runs the whole journey.
- Which AI agent works?
- Any capable one — ChatGPT, Claude, Gemini or a local agent. ÆON detects capabilities and degrades gracefully instead of pretending.
- What does it cost?
- Just your agent's tokens. The protocol, specs and topic packages are free and Apache-2.0 licensed.
- Where does the content come from?
- Your agent researches it — tiered sources, primary evidence first, with honest labels for contested claims. Not from a course catalogue.
The method
Five phases, every time
ÆON Learn never dumps chapters. It discovers, researches and structures before it teaches — and it adapts without breaking the path.
You type — one sentence, any capable agent.
and becomes the ÆON Learn orchestrator
Discover
The agent asks what you already know, what you want to achieve, how much time you have and how deep you want to go. No lesson before this.
what can this runtime actually do?
knowledge, goal, time, depth, formats
Research
It builds an evidence map from tiered sources — primary evidence first, popular explanation last. Without web access it says so and lowers its confidence.
tiered sources, evidence map
- Tier 1 · primary
- Tier 2 · synthesis
- Tier 3 · expert
- Tier 4 · popular
Structure
Concepts, prerequisites, controversies and misconceptions become a knowledge map; modules follow dependencies, not chapter conventions.
concepts, dependencies, controversies
modules follow dependencies
no — revise ↺
Learn
You approve a learning contract, then sessions run: hook, one core concept, evidence, its limits, application, exercise, reflection.
progressive sessions
hook · one concept · evidence · boundary · exercise
recall before answers
Adapt
Too easy, too hard, needs depth? Later modules adjust — the prerequisite structure never breaks. Earlier concepts return as retrieval.
adapt later modules ↺ · next session ↺ ADAPTING
synthesis · concept map · applied challenge · source map
Why it is different
Most AI tutoring is generation. This is discipline.
ÆON specifies behaviour, not prose: what an agent must do, verify and disclose before it may call itself a teacher.
Research over generation
A curriculum is compiled from an evidence map, not expanded from pretrained memory. Sources are tiered; primary evidence outranks readability.
Calibrated honesty
Claims carry labels from established finding to ÆON inference. Contested subjects must include serious counterpositions — no false certainty.
Contract before content
Capability detection, discovery and an approved learning contract come first. An agent that cannot schedule says so — it never pretends.
Retrieval over rereading
Sessions end with exercises and reflection; later sessions demand recall before showing answers. Completion means you can reason with the material.
Proof, not claims
Born from a real learning sprint
ÆON Learn generalises a real 14-day journey — the Charisma Sprint — preserved in the repository substantively unchanged as the canonical reference fixture.
An Instagram ad sparked curiosity; research replaced the ad's promises with evidence. Fourteen daily sessions, twenty-six anchored sources, daily drills and three reflection questions each — the pattern ÆON Learn now specifies for any subject. The retrospective documents honestly what worked and what the protocol improves.
- Raw stimulus
- User curiosity
- Topic decomposition
- Research
- Daily sequencing
- Podcast-native explanation
- Evidence
- Behavioural transfer
- Reflection
- Progressive immersion
- 14
- daily sessions, preserved verbatim
- 26
- anchored sources, tiered in the source map
- 3
- original documents, byte-identical
The library
Topic packages — accelerators, never limits
Curated epistemic scaffolding for known subjects: canonical sources, knowledge maps, misconceptions. Any subject outside the library works too — the agent researches it dynamically.
Packages provide curated sources and maps, not fixed prose lessons.
The protocol
Open, versioned, machine-readable
Everything an agent needs is public: normative specifications, JSON schemas and behavioural evals. Apache-2.0, semantic versioning, release-tag pinning.
current release: v0.1.2 · Apache-2.0
See what your agent builds when it has to research first
Free, open, Apache-2.0. Pick any subject — the obscure ones are the best test.