← the studio
A MoreSalamander · StudioLabs production

my-AI-stro

A local-first, self-improving knowledge system that turns school lessons into a queryable Source of Truth — engineered as notes for an open-book exam where LLMs are disallowed. The AI does the capture weeks early; the cold notes are the only aid you carry in.

The maestro — it conducts your knowledge.
▶ Open the live app — the in-depth About explainer ↗ live demo · best on desktop

What it is

Notes you can trust, because a judge keeps them grounded.

A study tool that hallucinates is worse than none — so my-AI-stro makes verification a system property: every lesson is captured into a structured SOT, and a deterministic (non-LLM) judge scores grounding at every write. The crowning jewel of the MoreSalamander methodology — the reference implementation a stranger can read to understand what the thesis means in practice.

Watch it run

A model summarizes a lesson; a judge refuses anything ungrounded.
my-AI-stro · ingestion pipelineattempt 1
graph_entry
retrieval
summarize
validation
judge ●
memory_write
grounding
— /100 · need ≥ 75
The validation judge ● is deterministic, formula-based — not an LLM, because the grader can't be the thing it grades. Ungrounded entries are rejected and re-summarized before anything reaches the SOT. (Two more pipelines — advisor and Classroom — run on the same event vocabulary.)

The three moves

Explain · Synthesize · Verify — at every persistence boundary.
01 · Explain

The lesson, the SOT schema, and the grounding criteria the judge will apply — written down first. Trust isolation is part of the doctrine: the model that owns the SOT never handles ungrounded chat.

02 · Synthesize

A local model (llama3:8b) summarizes the lesson into structured SOT entries; retrieval pulls related prior entries so the new entry fits what's already known.

03 · Verify

A deterministic, formula-based judge scores grounding against the source and rejects ungrounded entries. A background audit loop keeps rotating canonical entries toward more-grounded versions over time.

Best at — and the honest limit

Your knowledge, grounded — and queryable.

my-AI-stro runs a full curation chain (Ingest → SOT → Chat → Notebook → Classroom), each boundary Python-verified, with three coexisting pipelines on one shared NDJSON event vocabulary and a gradebook that extends verification into learning analytics. Everything local; the notes are yours and stay on your machine.

What it can't do: the judge verifies that an entry is grounded in your source material — not that the source was right, and not that it's a general-purpose oracle. It's your notes, kept honest — deliberately not a chatbot that answers from thin air.