Launches, partnerships and platform milestones from the team building the operating system for educational intelligence.
The industry competes on whose model is largest this quarter. For education that is close to the wrong race. The durable asset is not the model — anyone can call one — it is the accumulated understanding of a learner, a class and a curriculum that deepens with every lesson taught. aime™ is built so that memory belongs to the system that runs it, and to the nation that owns it.
The ambition in most education AI is to need the teacher less. aime is built on the opposite premise — that the teacher's judgment is the highest-value intelligence in the room, and the system's job is to amplify it, never to route around it. Like everything else in the platform, that choice is enforced in the architecture, not asserted in the marketing.
Most "education AI" is a general model with a teaching prompt bolted on top — education-applied. That is a category error, not a product tier. When the pedagogy lives in the prompt, it can be overridden; when it lives in the architecture, it holds. aime™ is built education-native, and that choice is enforced in the system, not asserted in the marketing.
The industry ranks its models on how often they get the answer right. For education that's close to the wrong measure entirely — a model that answers perfectly can teach nothing. aime is built against the scoreboard that actually matters, and that choice is enforced in the architecture, not asserted in the marketing.
Every vendor in education AI claims to be curriculum-aligned. Almost none can show you where a single objective came from. Provenance Gate is the standard aime holds itself to inside Curriculum Foundry™: nothing enters the curriculum graph without a traceable official source, and anything that cannot cite its origin is refused rather than approximated.
Most AI in the classroom optimises for the fastest possible answer. That is precisely the wrong objective. aime is built on the opposite principle — that learning happens through honest struggle — and that principle is not a slogan, it is an engineering constraint enforced by EduRule™.
The app model treats education as a series of features to install. Infrastructure treats it as a system to be reasoned about — education-native, running where the classroom actually is, and owned by the nations that depend on it.
The frontier of education AI isn't the largest model — it's the most reliable one that fits on the hardware a school actually has. A short engineering perspective on the trade-offs we optimise for.
Every AI product that claims to be aligned to the curriculum rests on a curriculum model someone had to build by hand. aime Curriculum Foundry™ replaces that hand-curation with an engine — one that compiles official government curricula into an auditable, pedagogically-sound graph where every node cites the exact source it came from, and nothing ships until an in-country teacher signs off.
Mathematics is not History, Chemistry is not Geography, Programming is not Biology. aime moves past the generic AI tutor pattern and gives every subject its own pedagogy, its own toolset and its own presentation environment — composed at lesson time by EduRule™, ThinkBook™ and the AIME Renderer.
aimeFUSION is the layer that turns aime's device, classroom and cloud tiers into a single coherent system — fusing intelligence vertically across compute, data and sync, and horizontally across a fabric of specialised models, so teaching continues fully offline at good-enough quality and improves when the connection returns.
aime is pursuing its Smarter-on-Smaller-Models programme to push offline educational capability to its maximum on low-end and disconnected hardware — while scaling aimeCLOUD™ to host a premium tier exploring deeper technical depth, simulations, animations, gamification and age-differentiated pedagogy. aime Hub launches over the summer alongside premium aimeCLOUD enhancements.
The Planner Agent evolves from regenerating full plans to producing immutable planning decisions — the current plan becomes a projection of replayable educational events, with full audit, explainability and incremental update.
Following the May rollout, aime reports continued progress on its platform-wide architectural optimization — pushing more work out of the model and into deterministic code, agent by agent, with classroom quality held constant.
aime sets out the engineering principles that let its agent stack run reliably on small-class models — moving work out of the model and into code, with no loss in classroom quality.
aime unveils Kern™, the orchestration engine that lets a compact model produce a teacher's full week of lessons, slides, quizzes and homework in 27 minutes — at a fraction of the cost of frontier models.
A 2-billion-parameter reasoning model built through aime™'s multi-teacher distillation programme and paired with Think Cache and Think Book — a fully offline AI tutoring infrastructure that runs on a laptop CPU, an old Android tablet or a Raspberry Pi.
Calabash turns aime Lesson Studio™ from a lesson reader into a full teaching studio: edit assessments, edit slides, print student-facing sheets and work in English or French — all inside a single offline-capable .aimepack™ workflow.
aime™'s new offline reasoning model, distilled from multiple teacher families, caught the Moses Illusion, held 17×13=221 under social pressure and ran the full benchmark on CPU-only hardware with no internet.
EduRule™ is the layer that turns aime from a generative tool into an adaptive teaching system — built on the research insight that students learn through honest struggle, not through being given the answer.
A novel, aime-built architecture that compares image generation prompts semantically before any image is generated — cutting redundant AI image calls, lowering lesson generation latency and producing a more consistent visual pedagogy across slides and covers.
aime has completed the migration to the aime Loom™ Workflow Engine™ — a novel, aime-built orchestration system for educational AI — delivering deterministic execution, built-in replay, and around 50 concurrent workflows on a single t3-medium instance.
aime unveils the .aimepack format and a lightweight Tauri-based viewer that lets teachers double-click a single file to open a full interactive lesson — with no login, no internet and the same UX as PPTX or PDF.
aimeCLOUD™ brings the full aime operating system online as a managed service, giving ministries, universities and partners a single environment to run curriculum-aligned AI at national scale.
aime™, aimeCLOUD™, aime Lesson Studio™, Baobab™, Calabash™, .aimepack™, Loom™, Loom Workflow Engine™, EduRule™, Kern™, ThinkCache™, ThinkBook™, aime-Reasoner-2B™ and aime-Reasoner-4B™ are trademarks of aime. All products, architectures and engines referenced in this newsroom are proprietary intellectual property of aime.