Templates and Transfer Pack

# Current state
Block: [N] [title]
Mode: Normal / Compressed / Paused / Re-entry
Slot: [which of 15]
Blocks remaining: [N] | Slots remaining: [N]

## Open right now
- [the single next action, specific enough to start cold]

## Last command that worked
[paste it]

## Blocked on
- [thing] waiting since [date] chase on [date]

## Done-when test for this block
[copy it here so it stays visible]

---
# Session log
## YYYY-MM-DD
Did:
Learned:
Next:

| Phenomenon | Level | Lens | Measurement | Design | Supported inference | Unsupported claim | Interest (H/M/L) + why |

Name / group:
Institution and location:
Why the match is real (specific, not "does neuroscience"):
One recent work (title, year, one-line summary):
Bounded help I might later ask for:
Contact route:
Status: mapped / contacted [date] / replied [date] / declined / active
Notes:

| Modality | What is physically measured | Spatial scale | Temporal scale | Invasiveness | Supports | Cannot support | Main artefact or confound |

Paper / figure:
1. Question the figure addresses:
2. Preparation and subjects (species, N, nesting):
3. Task or stimulus:
4. What was measured, in physical units:
5. Aggregation (across what, how):
6. What the figure directly shows:
7. One inference it supports:
8. One causal or generalisation claim it does NOT support:
9. What would change your reading:

# Q-[ID] status: curiosity / triaged / audit candidate / killed / parked / selected

## Stage 1 curiosity
Observation or problem:
Where it came from:
Level and domain:
Why it might matter:
What I already know:
What answering it would teach me:

## Stage 2 triage (only if promoted)
Likely population and setting:
Likely data source and its access reality:
Who benefits, or what understanding changes:
Obvious barrier (access / ethics / compute / supervision / time):
Cost of the missing learning:
Smallest test that could kill this:
Fatal risk:

## Stage 3 audit finalist (only if promoted)
Precise estimand:
Comparator:
Primary metric:
Five closest works:
One-sentence differentiator:
Independent-unit count available:
Full feasibility screen: PASS / FAIL by row
Intended-use statement: [link]
Learning contract: [link]

## If killed
Date and reason:

# Overlap memo [question] [date]

## Question, one sentence

## Searches run
| Source | Query | Date | Records | Relevant |
|---|---|---|---|---|
| PubMed | | | | |
| IEEE Xplore | | | | |
| Scopus / WoS | | | | |
| medRxiv / bioRxiv / arXiv | | | | |
| Scholar forward citations from [anchor paper] | | | | |
| PROSPERO / OSF | | | | |
| Conference proceedings | | | | |
| Benchmark leaderboard | | | | |

## The five closest works
| Work | What it did | Data | What it did not do |
|---|---|---|---|

## Differentiator, one sentence

## Verdict
Material contribution / occupied / needs reshaping

## Refresh log
| Date | Trigger | Anything new | Effect |
|---|---|---|---|

# Data card [corpus] [version/release]

Source URL and access date:
Version or release identifier (exact):
Licence and permitted use:
Access route (open / registration / signed agreement / credentialed):

## Contents
Subjects (unique, not files):
Recordings:
Sampling rate:
Channels: are these electrodes or derivations? Any duplicated or reversed?
Reference:
Duration, total and per recording:

## Independence
Unit of independence:
Nesting structure:
True independent N for my estimand:

## Annotations
Who annotated:
Against what definition:
Annotation unit (event / window / record):
Agreement checking reported? What did it show?

## Known limitations

## Integrity
Hash where licence permits:
Local path and retrieval date:

# Learning contract [project] [date]

One scientific concept to deepen:
  Resource: | Exercise: | Demonstrated by:
One method to learn:
  Resource: | Exercise: | Demonstrated by:
One implementation skill:
  Resource: | Exercise: | Demonstrated by:
One thing explicitly OUT of scope:

Prediction before applying the method:
What actually happened:

# Pilot decision [date]
Data used: DEVELOPMENT ONLY confirm: yes / no
Fraction piloted:
Runtime and storage feasible?
Annotation conversion visually correct on sampled events?
Comparator runs?

## Results already seen (must be disclosed at Gate B)
-

Decision: proceed / reshape / stop

# EEG description [record ID] [date]

Technical adequacy:
State (awake / drowsy / asleep / mixed):
Posterior dominant rhythm: frequency | amplitude | symmetry | reactivity
Organisation and continuity:
Symmetry:
State changes observed:
Artefacts (type, channels, likely source):

## Findings, DESCRIBED not named
| Location | Morphology | Frequency | Persistence | Evolution? | After-going slow wave? | Background disruption? |

Uncertainty and what would resolve it:
Requires expert review? yes / no why:

NO DIAGNOSIS. NO CLINICAL RECOMMENDATION.

# [Variant] vs [its closest mimic]

| Field | Variant | Mimic |
|---|---|---|
| Age group | | |
| State | | |
| Topography | | |
| Laterality | | |
| Morphology | | |
| Frequency | | |
| Amplitude | | |
| Duration / rhythmicity | | |
| Evolution | | |
| After-going slow wave | | |
| Background disruption | | |
| Closest confusion | | |
| **Decisive discriminator** | | |

Source and page/figure for every row:

Subject: [specific thing] [your role], Ain Shams

Dear Dr [name],

I am a medical graduate at Ain Shams University working through a
structured programme in neuroscience and computational methods.

I read [exact paper, year]. [One specific, accurate sentence showing
you actually read it, ideally about a method or a limitation.]

I have attached [ONE artifact: figure audit, protocol, or analysis].

My question: [ONE answerable question, not "can you supervise me".]

If this is not a good use of your time, I completely understand.

[Name] | ORCID | [repo link if relevant]

# Gate [Interest / A / B / C / Publication Release] [date]

## Evidence presented
-

## Criteria
| Criterion | Met? | Evidence |
|---|---|---|

## Decision

## Reasoning

## What I am giving up by deciding this

## External input
Who: | When: | What they said: | Peer feedback or expert review?

## If this gate did not pass
Route taken:

# Substitution [date]

What was planned:
What changed, and how I know:
Replacement:
Does it serve the same learning job? How:
What is lost:
Blocks affected:
Calendar effect: none / uses a floating slot / moves the target date

These are fixed objects, not descriptions of objects. Do not rehearse them during instruction. At the completion test: copy the prompt into a dated file, commit your first answer, then open the key and correct yourself in a second section. The point is transfer and correction, not a surprise exam.

import numpy as np, pandas as pd
rng = np.random.default_rng(20260731)
n_patients, n_windows = 30, 40
patient_effect = rng.normal(0, 2.0, n_patients)      # strong per-patient signature
y_patient = rng.integers(0, 2, n_patients)           # label is a PATIENT property
rows = []
for p in range(n_patients):
    for _ in range(n_windows):
        x = rng.normal(patient_effect[p], 1.0, 20)   # 20 features
        x[0] += 0.3 * y_patient[p]                   # weak true signal
        rows.append(dict(patient_id=p, y=y_patient[p],
                         **{f"f{i}": v for i, v in enumerate(x)}))
df = pd.DataFrame(rows)

# X has 40 windows per patient; patient_id repeats across rows
X = SelectKBest(k=20).fit_transform(X, y)
X = StandardScaler().fit_transform(X)
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=7
)
model.fit(X_train, y_train)