Local AI helps teachers grade work, break down knowledge points and generate traceable study plans. Parents stay in control — every observation enters a child's growth archive only after you confirm it.
Fragmented school-home ties, one-size-fits-all teaching, inefficient communication — Hemiaoji takes these apart with real collaboration and personalized learning.
School and home stay separate; a child's growth is split in two and no one sees the whole.
Teacher, parent and child collaborate on one growth archive — school and home truly unified, without overreach.
The same homework and pace for everyone, ignoring each child's strengths and gaps.
Local AI breaks down knowledge points and pinpoints weak spots, building a per-child plan — teaching truly suited to aptitude.
Group-chat bombardment, repeated relays and untraceable verbal feedback breed misunderstanding.
Reviewable, traceable memory proposals sync to families on demand — communication is evidence-based and never repeated.
Data scattered across platforms; children casually judged and labeled.
Local-first and permission-controlled, only gentle reminders — never diagnoses or personality labels.
Teachers log the classroom, parents add daily life, children take part in confirming — all observations form one growth archive, with clear roles and no overreach.
Within authorized subjects and classes, logs classroom performance, grades work and breaks down knowledge points — submitting “memory proposals” as a restricted external role.
Guardian of the archive: reviews every external input, adds daily-life observations and interests, and configures each teacher's scope.
The protagonist of growth: completes tasks and submits receipts, adds self-made questions under knowledge points, marks “I've got it,” and joins in confirming and self-assessing.
Every external input lands in an inbox as a “proposal” with source, confidence and evidence. Only after a parent approves does it enter the child's archive.
Learning, talents and wellbeing capture a child in the round. Wellbeing is sensitive — only gentle reminders, never diagnoses or personality labels.
Local AI builds the weekly plan from confirmed memory events; every item traces back to a source event_id — auditable and explainable.
Memories and files stay on-device by default; the cloud is only an encrypted relay. Teachers are restricted roles whose visibility parents tune by subject and class.
Hemiaoji's AI never sends a child's data to the cloud for training. Grading, knowledge breakdown, study plans and Q&A all run on local models — explainable, traceable, and the AI never oversteps.
AI grades work, breaks down knowledge points and pinpoints weak spots — all on-device, files and data never leaving. “Teacher-graded” or plugin grading are also options.
AI builds the weekly plan from confirmed memory events; each item traces to a source event_id — explainable and auditable, not a black box.
Ask the AI assistant about a child anytime — it answers only from confirmed memories in the archive, giving teachers and parents grounded advice.
AI output is also a “proposal” that must pass parental review before entering the archive. It makes no diagnoses, attaches no personality labels — judgment stays with people.
Want a complete, continuous picture of their child without scattering data or risking privacy.
Want grading, knowledge points and feedback done in one pass, then safely synced home.
Want to take part in their own growth record — finishing tasks, self-assessing, tracking progress.
For the first time a tool stitches classroom and home observations into one whole child — and nothing enters the archive without my nod. Reassuring.
After grading I break down knowledge points, generate feedback, then selectively sync to parents — far less repetitive communication, with clear permission boundaries.
I can see my own weekly study plan, finish tasks and self-assess — watching my progress add up bit by bit feels really rewarding.
A collaborative memory system for education. Local-first and permission-controlled, so every step of growth is recorded — steadily.