01 · What Ms. Sonoma is
A learning facilitator, not a chatbot beside a lesson
Ms. Sonoma actively participates in the instructional session. The application supplies lesson material, keeps track of the current step and progression, and preserves session state. AI performs meaningful teaching interactions using the current instructional prompt and the learner's input.
02 · What happens during learning
A structured path from conversation to assessment
A full session can include the stages below. The exact starting point and path can vary by lesson, resume state, and session configuration; stages may be skipped or combined when the experience calls for it.
- 01
Discussion
Open the topic, surface prior knowledge, and give the learner a reason to engage.
- 02
Teaching
Explain ideas, vocabulary, and examples in language suited to the lesson.
- 03
Comprehension
Ask questions, listen to the learner, and respond to signs of confusion.
- 04
Exercise
Move from explanation into guided practice with feedback and another attempt.
- 05
Worksheet
Give the learner room for more independent practice while preserving the session record.
- 06
Test
Collect assessment responses that the educator can review alongside the rest of the session.
- 07
Completion
Close the session, recognize the work, and make the resulting evidence available for review.
Progression is organized around explanation, comprehension, practice, and evidence—not merely elapsed time.
03 · Why AI
Capabilities worth testing, not outcomes to assume
AI may be well suited to parts of facilitation that benefit from responsiveness and repetition. Those capabilities motivate the product; they do not prove that AI teaches better than a person or that one session creates durable mastery.
Patient repetition
An explanation can be repeated or approached another way without social pressure or frustration.
Responsive interaction
The next instructional turn can respond to what the learner just said instead of following a fixed recording.
One-to-one pacing
A session can pause, revisit, or continue around one learner rather than the pace of a whole group.
Useful documentation
The application can preserve the exchanges and responses an educator may want to examine later.
04 · What the educator controls
Educational authorship stays with the educator
Educators decide what should be taught, what material is acceptable, who receives a lesson, and what to do with the evidence afterward. AI can carry more of the instructional workload without becoming the educational authority.
- Choose from built-in lessons or supply lesson material.
- Generate lesson drafts, then review, revise, and approve them before use.
- Manage learner settings and the optional AI features available to a learner.
- Assign or schedule lessons and decide when a learner should use them.
- Review available transcripts, responses, scores, progress, and notes.
- Intervene, adapt the next lesson, or make a different educational decision.
05 · How the product handles records
Preserve what happened, then interpret it carefully
Session snapshots
The application can save the current phase, completed phases, phase data, timer state, and timestamps so a session can resume.
Transcripts and responses
What Ms. Sonoma and the learner said, along with available question responses, can be retained for educator readback.
Local and account-backed persistence
Snapshots are saved locally first and can also be persisted through authenticated account storage when those services are configured.
Transcripts, responses, scores, timing, and facilitator notes can show what happened during a session. They do not by themselves prove long-term retention, durable mastery, or superiority to human teaching.
06 · Safety, boundaries, and privacy
Practical safeguards, described without guarantees
Ms. Sonoma uses application and prompt-level boundaries intended to keep learner interactions tied to an age-appropriate learning experience. We continue to test and improve those safeguards.
Current safeguards
- Checks on learner input supplied to the main Sonoma instructional route.
- Prompt guidance intended to keep responses age-appropriate and lesson-focused.
- Lightweight checks on generated text before a response is returned.
- Authenticated, owner-scoped routes for supported account records and lesson actions.
Safeguards vary by feature and route. They reduce risk; they do not make AI output infallible or remove the need for adult judgment and supervision.
Optional learner features
The instructional experience itself includes active AI-guided turns. Depending on lesson and learner settings, optional features such as Ask, Poem, Story, and Fill-in-Fun may also be available. Educators can configure access to those optional features for a learner.
Providers and model memory
The main Sonoma route can use Anthropic or OpenAI, depending on deployment configuration. It sends the content supplied for the current turn rather than maintaining a hidden, durable model conversation in that route.
When AI is used, the request data needed to generate a response is sent to the configured provider. Provider and deployment terms govern that processing; this page does not make a categorical model-training promise.
Application records are different
A stateless model request does not mean the learning application retains nothing. Ms. Sonoma may keep local snapshots and, for authenticated accounts where services are configured, persist progress, phase data, transcripts, and related records for resume and educator readback.
The standard we are choosing
Evidence should outrank the story we hope is true.
Ms. Sonoma is an experiment in giving AI more responsibility for the work of teaching while increasing, not diminishing, educator authority. Calm supports that work by making correction, repetition, and another attempt easier. Calm is the method; learning is the mission.