aime™ moves the Planner Agent to a replayable, event-sourced architecture

aime today described the next step in its continuing platform innovation: a replayable planning architecture for the Planner Agent — the component responsible for what a learner is taught, when, in what order and at what pace. Instead of regenerating a complete plan every time an educational signal arrives, the Planner Agent will produce small, well-typed planning decisions, and the current plan will be derived by replaying those decisions over time.

Why the change

The Planner Agent reacts continuously to teacher reflections, assessment outcomes, mastery updates, curriculum changes and administrative interventions. Today, even a small adjustment — inserting a single Deep Dive on fractions, for example — causes the entire plan to be regenerated. As plans grow, that pattern becomes expensive, hard to audit and difficult to extend to multiple human and agent collaborators.

The shift, at a principles level

• The Planner Agent produces educational decisions, not educational documents.
• Each decision is stored as a stable, domain-level event — independent of the model, prompt or provider that produced it.
• The current plan becomes a projection: Plan = Replay(Planning Events).
• Periodic snapshots keep replay fast as planning history grows.
• Human overrides, locks and approvals are first-class events alongside agent decisions.

The architecture extends concepts already proven inside aime's Loom™ Workflow Engine, which treats long-running workflows as deterministic event histories. Applying the same posture to planning gives the Planner Agent the same properties Loom™ gives workflows: replay, audit, explainability and recovery.

“A learning plan is not a document. It is the consequence of a long sequence of educational decisions. Once you treat it that way, you stop regenerating a month of lessons because a student struggled with one topic — and you start being able to explain, audit and replay every change a teacher ever sees.”

What teachers and ministries get

• A complete, queryable history of why a plan looks the way it does.
• Smaller, faster, cheaper planner runs — incremental decisions instead of full regenerations.
• Explainable plan evolution — every change tied to the signal that caused it.
• A foundation for human-in-the-loop governance, branching plan experiments and rollback.

“Continuing innovation, for us, is structural. We are not adding another model — we are changing what the model is asked to emit. Decisions, not documents. That single shift unlocks audit, replay and incremental planning across the platform.”

Continuity with the rest of the platform

The replayable Planner sits naturally alongside aime's Smarter-on-Smaller-Models programme: smaller prompts, narrower responsibilities and deterministic software around the model. The specific event taxonomy, projection engine design, snapshot strategy and conflict-resolution rules are aime's proprietary IP and are shared with partners under NDA.

Availability

The replayable Planner Agent enters phased rollout across aimeCLOUD™ from June 2026, sequenced inside the aime Loom™ Workflow Engine for deterministic, replayable execution.

About aime

aime builds the operating system for educational intelligence — the foundational infrastructure layer that future education systems will run on. aime's stack combines structured curriculum knowledge, pedagogy-aware reasoning, compact education-tuned models, agentic orchestration and offline-capable deployment, and is designed for ministries of education, universities and national school systems.