Research New

• "anisotropic boundaries" → top-level in Research/Active/
• "boundary_conduction_speedup/anisotropic_boundaries" → sub-question under existing parent

If the input looks like a sub-question (contains /), verify the parent exists in Research/Active/. If it doesn't, ask the user whether to create the parent first.

Step 2: Name the Folder

Convert the question name to a folder name:

• Lowercase, underscores, no special characters
• Descriptive but concise (2-4 words)
• Examples: boundary_conduction_speedup, hipsc_cm_ionic_models, anisotropic_boundaries

Step 3: Ask Clarifying Questions

Before scaffolding, gather:

1. The question — one sentence, as specific as possible
2. Why it matters — what depends on the answer
3. Which engines — which simulation engines does this touch? (Bidomain V1, Monodomain V5.4, LBM V1, none yet)
4. What "done" looks like — concrete completion criteria
5. Known sub-questions — are there obvious sub-topics to create now?

If the user already provided enough context (e.g., from a conversation), infer the answers and confirm rather than asking.

Step 4: Create Folder Structure

For a top-level question:

Research/Active/{question_name}/
├── README.md
├── KNOWLEDGE.md
├── IDEALOG.md
├── WHITEBOARD.md        # Ephemeral diagrams for tmux workspace
├── literature/          # Paper summaries go here
├── papers/              # PDFs go here
├── code_examples/       # Reference implementations (if any)
├── results/             # Simulation outputs, data, plots
└── plans/               # Archived completed PLAN.md files

For a sub-question:

Research/Active/{parent}/{sub_question_name}/
├── README.md
└── results/

Sub-questions are lightweight. They share the parent's literature/, papers/, and code_examples/ unless they grow large enough to need their own.

After creating the folder, rename the tmux window to the question's readable title:

tmux rename-window "{Question Title}" 2>/dev/null

(Capitalize each word, e.g., "Boundary Conduction Speedup".) For sub-questions, use the sub-question title.

Step 5: Write README.md

Top-level README:

# {Question Title}

## Question
{One sentence — the specific thing we're trying to answer}

## Status: Active

## Why It Matters
{1-2 sentences — what depends on the answer, which engines/pipelines are affected}

## Engines
{List of engines this touches, with what role each plays}

## Completion Criteria
- [ ] {Concrete criterion 1}
- [ ] {Concrete criterion 2}
- [ ] ...

## Sub-Questions

| Sub-Question | Status | Key Finding |
|-------------|--------|-------------|
| {sub if any} | — | — |

## Key Findings So Far
{Empty initially — updated as work progresses}

## Literature
| Paper | Summary | Key Insight |
|-------|---------|-------------|
| {empty initially} | | |

## Future Work
{No deferred items yet.}

Sub-question README:

# {Sub-Question Title}

**Parent**: [{parent name}](../README.md)

## Question
{One sentence}

## Status: Active

## Completion Criteria
- [ ] {criterion}

## Findings
{Empty initially}

Step 6: Write KNOWLEDGE.md

Only for top-level questions (sub-questions contribute to the parent's KNOWLEDGE.md).

# {Question Title} — Knowledge File

> This file is a running synthesis. Updated as findings accumulate.
> When the question is complete, a copy is promoted to `Research/Knowledge/`.

## Current Understanding
{What we know so far — even if just "nothing yet, starting investigation"}

## Key Decisions
{Decisions made based on this research, with rationale}

## Open Questions
{Things we still don't know}

## Connections
- **Engines**: {which engines implement or depend on this}
- **Related research**: {links to other research questions that interact}
- **Pipelines**: {Optimizer, Surrogate, Builder if relevant}

Step 6b: Write IDEALOG.md

Only for top-level questions (sub-questions don't get their own IDEALOG.md).

# {Question Title} — Idea Log

> Thinking trail: how our understanding evolved, what we tried, what failed.
> Scan this in 30 seconds to remember where we are and how we got here.
> Not promoted on completion — archived for historical record.

## Current Direction
{What we're currently pursuing and why — same as README.md status initially}

## Next Step
{First step from completion criteria}

## Thread

## Failed Approaches

## Session Log

Step 7: Update MASTER.md

Add a row to the Active Research table in MASTER.md:

| [{question_name}](Research/Active/{question_name}/) | {engines} | Just started | {first step} |

If this is a sub-question, don't add to MASTER.md — update the parent's README.md sub-question table instead.

Step 8: Update Parent (Sub-Questions Only)

If creating a sub-question, update the parent's README.md:

• Add a row to the Sub-Questions table
• If this is the first sub-question, create the Sub-Questions section

Step 9: Report

Output:

• Path to the new folder
• The README.md content (for user review)
• Suggested first steps (literature search, engine experiment, etc.)
• Whether existing knowledge files or papers might be relevant (check Research/Knowledge/ and MASTER.md)

Rules

• Never create a question without a clear completion criterion. "Investigate X" is too vague. "Determine whether X causes Y, validated by Z" is concrete.
• Top-level questions go in Research/Active/. Never create directly in Complete/ or Backlog/.
• Sub-questions inherit parent resources. Don't duplicate papers/ or literature/ unless the sub-question has material the parent doesn't.
• Check for duplicates. Before creating, scan Active/, Complete/, and Backlog/ for existing questions that might already cover this topic. If found, suggest extending the existing question instead.
• Don't over-structure. If the question is simple and has no obvious sub-topics, skip the sub-question scaffolding. It can be added later.