01
Explore before you commit
Start with interests, read what others have tried, and compare promising directions. A testable ML question comes after the landscape begins to make sense.
Independent research workbench · Grades 10–11
Find a question worth turning into research.
Abstract
Explore a field, find what others have done, then shape a question around public data and ML. The workspace keeps reading, experiments, methodology, writing, and revision in one research record.
- Research map
- Flexible
- Areas
- Unlocked
- Record
- Continuous
Method
A mentor’s process, not a substitute author.
02
Keep the reasoning yours
The mentor teaches methods and critiques what you record. You choose the question, sources, design, interpretation, and manuscript language.
03
Leave an evidence trail
Papers, daybook entries, data versions, experiments, tables, figures, and revisions remain together as one inspectable research record.
Research map
A living map, shaped by the research.
Move among these areas as the project develops. Reading can change the question; results can change the method; writing can reveal another experiment.
Topic exploration
Return anytimeRead broadly before deciding what the project is.
The Question
Return anytimeTurn a raw curiosity into a sharp, contestable question.
Literature
Return anytimeFind the existing work your project must answer to.
Steelman & Novelty
Return anytimeDefend the existing work before you're allowed to contest it.
Evidence
Return anytimeGet evidence into usable form — often the longest, grindiest stage.
Preservation
Return anytimeKnow what to keep before you run — so you never re-run.
Method
Return anytimePlan the experiment you can understand and defend.
Validation
Return anytimeDoes my result survive scrutiny?
Write-up
Return anytimeFeedback, feedback, feedback.
Venue
Return anytimeWhere to submit — a place AI is genuinely useful.
Review Loop
Return anytimeConcede, defend — and do the real new work when a critique is right.
Working agreement
You stay responsible for the work.
- 01Methods and ML concepts are taught when they become useful.
- 02You find, open, verify, and save every paper you use.
- 03Generated code is inspected through its inputs, assumptions, and outputs.
- 04Real progress means evidence on the record—not completed-looking boxes.
- 05The paper and experiments can send the project back into an earlier loop.
Founder’s field note
“My mentor did not need to know my field. He trusted me to investigate the subject. What he contributed was the experience of conducting ML research: how to test, preserve, question, and revise the work.”
Begin with an interest
You do not need a topic yet.
Start by collecting fields, questions, and real work that catches your attention. If you already have a topic idea, begin by shaping it into a measurable ML question. No statistical vocabulary is required at the start.
Start a project →