Read the week in plain language, refresh the foundations, follow each challenge step by step, and check what evidence to keep. Includes worked examples, answer keys and prompts for visual study resources.
Estimate PCA Hotelling T2 and SPE control limits using only training/calibration normal batches; choose alarm persistence before evaluating faults. Acceptance: Limit estimator, sample support and nominal alarm policy are reproducible.
alarm-examples.csv Build
Implement batch replay with timestamped alarms and compare normal/fault cases selected independently of results. Acceptance: Event timing and undetected cases are retained; no cherry-picked pass rule.
Check leakage, missing channels and fake-data reproducibility; prepare a one-command monitoring report. Acceptance: The smoke run works without network or credentials and numerical invariants are checked.
Checklist, learning assessment, gate status, shortfalls, next actions and carryover
Notes on artifact states
These filenames are outputs to prepare during this week's execution; this scaffold does not claim they already exist. Drafts and measured results have different states. Technical code/data stay in the canonical project repository and are referenced from these coordinating files.
During preparation, future measured tables get column definitions, input requirements and run instructions with PENDING status. During execution, replace pending values only with measured outputs. Do not create realistic-looking fake evidence.
▦
600 MINUTES MAXIMUM
Seven-day plan
Monday90 min· 2 cards, 60 min
Target. 🚨 Draw the alarm line. Hook: what should count as abnormal? Do: compute Hotelling T² and SPE limits from normal training batches only, with the NIST formula cited above the code; choose the alarm rule before looking at any fault. Done when: the limits reproduce from one command.
Target. 🌱 The growth curve, for real. Hook: put your week 1 plot next to a textbook curve. Learn (45 min): OpenStax Microbiology ch. 9 growth-curve section only, then draw the curve by hand with lag, log, stationary and death. Done when: you can show which phase Monod explains and which it does not. Then stop studying and continue Monday's limits.
Target. 🎯 Replay and catch the fault. Hook: will your alarm fire on the broken batch? Do: replay batches with timestamped alarms, and compare your limits with the erdogant/pca README. Done when: alarm times and missed cases are saved, with no cherry-picking.
Target. 🧭 What PCA really found. Hook: 3Blue1Brown eigenvectors (17 min). Learn: StatQuest bias-variance (7) and linear regression (13), then fit one regression on two IndPenSim variables. Done when: you can explain bias vs variance and read a regression line aloud. ISLP is parked. Last 30 min: founder paper.
Target. 🛡 Make it trustworthy. Hook: try to break your own pipeline. Do: check leakage and missing channels; one command produces the monitoring report offline. Done when: the smoke run works without network or credentials and the numbers check out.
Target. Catch up, then reward. First: finish the earliest unfinished mission (only its Done when). Then the rabbit-hole slot: anything from PARKING or the Deeper list, guilt-free. Done when: no mission from this week is half-open, or it is logged as moved.
Target. Show it. 20 min: post the alarm chart catching the fault with three lines (what, how, one honest limitation). 5 min: tick evidence. 20 min: write what was hard and why. 15 min: write next Monday's first action, startable in 2 minutes. Done when: the post is drafted and Monday's action is written.
Block ritual: read your next action → start the 90-min timer → open one file. Close: write tomorrow's action and one log line.
Feels pointless? Does today's Done when need it? No → skip it and log it. Yes → let AI do the boring part, you review. Still stuck → the smallest version in 25 minutes. Pulled into a rabbit hole? Write it in Parked thoughts and save it for Saturday.
The model must turn these into exact physical actions using the observed files and available inputs. Each 90-minute block includes start/stop rituals. All calls, posts, debugging, reading and exercises consume this same capacity.
Each day is one mission: the Hook gets you in, Done when tells you to stop. Study stops at the understanding check. If a piece feels pointless, ask whether Done when needs it: no → skip and log; yes → let AI do the boring part and review it; still stuck → smallest version in 25 minutes.
◎
WHAT TO WATCH, READ AND RUN
Materials for this week
Every plant manager has a "golden batch": the run that went perfectly and that everyone secretly compares against. The problem is that a single golden batch is an anecdote. This week you replace the anecdote with an envelope: what 80 normal batches look like together, with statistical limits instead of someone's gut thresholds. The analogy from your world is the portfolio dashboard. A single project running 5% late …
In the plan's running order: each card says which block it belongs to and what has to come
before it. The seven-day plan lists the same cards under the day that owns them.
Kourti & MacGregor (1995), "Process analysis, monitoring and diagnosis, using multivariate proje…
Saturday sandbox
· after Nomikos & MacGregor (1995), "Multivariate SPC charts for monitoring batch processes"
Kourti & MacGregor (1995), "Process analysis, monitoring and diagnosis, using multivariate projection methods", Chemometrics and Intelligent Laboratory Systems 28:3-21 (verify)
James, Witten, Hastie, Tibshirani & Taylor, An Introduction to Statistical Learning with Applica…
Saturday sandbox
· after Crash Course Statistics
James, Witten, Hastie, Tibshirani & Taylor, An Introduction to Statistical Learning with Applications in Python — free PDF at statlearning.com — Ch. 2 (whole) and 3.1-3.3 only. Moved out of core on 2026-09-29
Stanford Online · Statistical Learning with Python
Saturday sandbox
· after James, Witten, Hastie, Tibshirani & Taylor, An Introduction to Statistical Learning with Applica…
Stanford Online · Statistical Learning with Python (Hastie, Tibshirani et al.) — Playlist: youtube.com/playlist?list=PLoROMvodv4rPP6braWoRt5UCXYZ71GZIQ · course page: online.stanford.edu/courses/sohs-ystatslearningp-statistical-learning-python. Moved out of core on 2026-09-29
Nomikos P. & MacGregor J.F., "Multivariate SPC Charts for Monitoring Batch Processes", Technomet…
Saturday sandbox
· after Stanford Online · Statistical Learning with Python
Nomikos P. & MacGregor J.F., "Multivariate SPC Charts for Monitoring Batch Processes", Technometrics 37(1):41-59, 1995. DOI 10.1080/00401706.1995.10485888 · author PDF mirror: three-mode.leidenuniv.nl/pdf/n/nomikosmacgregor1995tmet.pdf. The founding paper of batch MSPC — abstract and figures are enough. Cite it in week 08's report
Kourti T. & MacGregor J.F., "Process analysis, monitoring and diagnosis, using multivariate proj…
Saturday sandbox
· after Nomikos P. & MacGregor J.F., "Multivariate SPC Charts for Monitoring Batch Processes", Technomet…
Kourti T. & MacGregor J.F., "Process analysis, monitoring and diagnosis, using multivariate projection methods", Chemometrics and Intelligent Laboratory Systems 28:3-21, 1995. (verify)
Its README shows Hotelling T² and SPE outlier detection on top of PCA; run the example, then compare its limits with yours (never ship its numbers without understanding them)
This week delivers the three concrete artifacts below and the supporting learning/branding/opportunity pack. The scientific or commercial outcome follows the preserved source brief, while the acceptance checks make its evidence testable. The career objective is to turn this work into something a scientist, hiring manager or buyer can inspect.
Scientific/build priority: Estimate PCA Hotelling T2 and SPE control limits using only training/calibration normal batches; choose alarm persistence before evaluating faults.
Public/profile priority: Draft a golden-batch envelope explanation with measured examples and visible false-alarm limitations.
Career/relationship/commercial priority: Prepare a useful technical comment for one founder/peer and keep the connection-ladder stage in the private tracker.
02
READ FIRST
Deeper walkthrough (v6)
Read v6 · Week 3 · Golden batch, with statistics before Monday. It carries this week's v5 brief unchanged and adds the Welcome analogy, Learn it (▶ core and ＋ deeper sources), Run it (repos with a 20-40 minute exercise), the NotebookLM kit and the mentor check below. Standing v6 rules: resource rules, NotebookLM protocol, chapter anatomy. v6 adds depth only; it does not change a goal, gate, output or the 600-minute cap, so this specification still governs execution.
Verified links, exact scopes and durations for this week's reading and watching: materials.md.
Act I · Threads: B, M, BL · Next gate: 4
Where this week sits.
- ← From: week 2 PCA on normal batches; week 1 injected fault.
- → Feeds: week 4 (hours-to-detection and false-alarm rate), week 8 (technical report "golden batch" section), week 17 (hours-to-detection per model in Paper 2), week 33-35 (batchlens alarms trigger labdocs deviation drafts).
- ⟳ Returns in: every Loop Audit (week 29): "how early would you have known?"
Goals.
- 🎯 This week: T² and SPE limits computed from data; batchlens monitor --batch 87 alarms on a faulty batch and not on a normal one; CI green.
- 🏁 Next milestone: Gate 4.
- 🌟 Final: the early-warning headline number that opens founder conversations in archetype A.
Mentor check."I can explain to a plant manager why their golden batch is not a control limit, and what T² and SPE each catch."
03
OPEN WHEN NEEDED
Required inputs and dependency check
Week 02: inspect its actual evidence and review; source/code artifacts must exist before the dependent task begins.
Inspect the actual canonical repository state, available data, existing figures and required human/external inputs. Never assume a previous week succeeded from its folder name.
Verify current resource links, versions, dataset rights and any live company/venue information before relying on them. Use source catalog; explicitly record inaccessible resources.
04
READ FIRST
Learning specification
Biology/domain unit
Source unit: B3 · Microbial growth
Material: OpenStax Microbiology ch. 9, growth-curve section only — openstax.org/details/books/microbiology
Produce: Notes + the growth curve drawn by hand with lag, log, stationary, death
Understanding check: You can connect the log phase to the Monod equation in your simulator
During preparation, write one worked example connected to the week's actual task, one exercise with a separate answer key and five questions. Peter supplies answers; score explanation quality and flag misconceptions in QUESTIONS. Numeric/scientific conclusions require actual data or explicitly synthetic teaching values.
AI/scientific-method unit
Source unit: M3 · Statistical learning I
Material: 3Blue1Brown "Eigenvectors and eigenvalues" + StatQuest bias-variance and linear regression (trimmed 2026-09-29; ISLP is Deeper)
Produce: one regression fitted on two IndPenSim variables
Understanding check: You know bias-variance and can fit and read a regression
Use the prompt pack: P1 explanation ladder, P2 concept-to-project bridge and P3 quiz. Coding practice belongs within the allocated study/build block. Where the source asks for a favorable scientific result, substitute an honest measured comparison as explained below.
Founder/scientist paper
Select one verified founder/scientific-lead paper relevant to this week's domain. Use P11; save the question, method, numerical result with context, data workflow, three author questions and one connection sentence in founder-paper.md. Keep abstract-only access limitations explicit. On source weeks with no study, prioritize the commercial/submission task and log any skipped digest.
05
OPEN WHEN NEEDED
Branding and opportunity execution brief
Branding: Draft a golden-batch envelope explanation with measured examples and visible false-alarm limitations.
Opportunity action: Prepare a useful technical comment for one founder/peer and keep the connection-ladder stage in the private tracker.
Use career asset contracts. Publish at most the agreed cadence; one figure and one supported number or admitted failure can support multiple drafts. In long-form weeks, prepare a 1,500–2,500-word article from existing evidence within the build/buffer allocation. In release/demo weeks, prepare a 60-second script: problem, input, command/action, result, limitation and proof link.
Verify individuals, jobs and affiliations at execution time. The GridX list and v4 company map are research seeds, not current facts. Do not reveal named company weaknesses or employer/client confidential information. Keep drafted/sent/replied/booked/held statuses separate. The planning request alone does not authorize external messages or publication; honor explicit authorization in the execution session.
06
OPEN WHEN NEEDED
Quality risks and week-specific constraints
Unfolding a whole batch can leak the future into early warning. Require causal feature availability for every alarm timestamp.
Apply the root scientific evidence contract: provenance; meaningful independent splits; no future-data leakage; baseline comparison; actual denominators/uncertainty; failed cases; human review where required. State whether data are simulated, public experimental, synthetic or partner-provided.
Keep source gates intact, and log scope/time conflicts in review. Missing people, money, access or results must not be replaced with invented evidence. Reduce optional polish first. A specific essential external dependency stays pending while independent preparation continues.
07
READ FIRST
Completion and gate review
This is an enabling week. Its deliverables feed the next scheduled gate: 4. Complete the acceptance checklist below; do not promote reputation from a draft.
Monday artifact satisfies: Limit estimator, sample support and nominal alarm policy are reproducible.
Wednesday artifact satisfies: Event timing and undetected cases are retained; no cherry-picked pass rule.
Friday artifact satisfies: The smoke run works without network or credentials and numerical invariants are checked.
Study pack is prepared; Peter's actual learning/quiz result is recorded separately.
Branding and opportunity drafts use supported claims and the correct external-action state.
Reproduction or document checks are recorded with file/version references.
This week's special gate, if any, has real evidence or remains explicitly pending.
Review, root progress/log and the next physical action are updated.
08
READ FIRST
Minimum viable week and recovery
Source-preserving minimum: T² limits computed.
Operational first step if stalled: open monitoring-protocol.md, write the required inputs and produce the smallest checkable example in ten minutes. If a more urgent missing gate dependency controls the week, use the source minimum above.
If the full week does not fit, use P14: retain one gate-critical deliverable and one directly supporting study exercise; record what is skipped versus moved. Do not mark the original full gate passed merely because the minimum was completed. After a missed week, reforecast from actual evidence and avoid a second unexamined lost week.
09
OPEN WHEN NEEDED
Handoff to the next model
In review, state exactly what exists, what was verified, what Peter can now explain, which external events remain pending, where the canonical code/data live, and the next file/function or draft paragraph to touch. Link each fact to evidence. Carry scientific assumptions, split definitions and rights constraints forward explicitly.
10
OPEN WHEN NEEDED
Ready-to-copy model prompt
Prepare Week 03 of this LABLOOP execution plan.
Read execution-plan/specs.md, execution-plan/DECISIONS.md,
execution-plan/GATES.md and
execution-plan/weeks/week-03-statistical-batch-monitoring/specs.md, commitment.md, evidence.md and review.md.
Read prior prerequisite reviews and inspect actual repository/data state.
Use v5 as the governing sequence, the v6 Week 03 chapter
(LABLOOP-Master-Plan-v6.md) as the deeper walkthrough, and v4 only for
retained details.
Produce the full local preparation pack and the exact week-specific files named
in the artifact table. Give me a bounded lesson with exercises/answer key,
daily physical actions, technical runbook or draft, branding drafts and
career/commercial actions. Keep the total to 600 minutes.
Separate prepared artifacts from measured results and my learning completion.
Verify sources and do not invent metrics, citations, papers, people or outcomes.
Do not execute external communications, publication or spending unless the
current session already authorizes the specific action.
Finish by updating evidence/review and TODAY with the next physical action.
11
OPEN WHEN NEEDED
Preserved original v5 weekly brief
v6 reproduces this brief unchanged inside v6 · Week 3 · Golden batch, with statistics.
The following is source planning text, not a claim of completion. Apply the explicit evidence corrections in this specification and DECISIONS.md.
Week 3 · Golden batch, with statistics
Goal: control limits that are statistics, not thresholds.
Tue (B3): Microbiology ch. 9. Draw the growth curve.
Thu (M3): ISLP ch. 2-3. Redo Lab 3 on IndPenSim.
Build: Mon: Hotelling T² and SPE with limits from the normal batches (NIST ch. 6.5 as reference). Done when limits are computed, not typed. Wed: batchlens monitor --batch 87 replays a batch and prints alarms with time stamps. Done when a faulty batch alarms and a normal one does not. Fri: tests for the two commands; make report FAKE=1 works on simulator data. Done when CI is green.
Sun: L-Sun: "Golden batch: what 80 normal fermentations look like as one envelope". L-Wed.
State: UNCONFIRMED. Owner: Peter Ciancaglini. Specification
Start date: not set. End date: not set. Available hours: confirm (planned ceiling 10).
Primary outcome: statistical batch monitoring.
Minimum that keeps the next gate alive: T² limits computed.
Deliverables to commit
☐ monitoring-protocol.md: Limit estimator, sample support and nominal alarm policy are reproducible.
☐ alarm-examples.csv: Event timing and undetected cases are retained; no cherry-picked pass rule.
☐ verification-report.md: The smoke run works without network or credentials and numerical invariants are checked.
☐ Bounded learning exercise and personal teach-back.
☐ Evidence-backed branding and opportunity pack.
☐ Sunday review and next physical action.
Proposed calendar
Day
Minutes
Physical action / outcome
Actual slot
Done
Mon
90
Estimate PCA Hotelling T2 and SPE control limits using only training/calibration normal batches; choose alarm persistence before evaluating faults.
Not set
☐
Tue
90
Notes + the growth curve drawn by hand with lag, log, stationary, death
Not set
☐
Wed
90
Implement batch replay with timestamped alarms and compare normal/fault cases selected independently of results. Last ten minutes: optional note.
Not set
☐
Thu
90
Lab 3 redone on IndPenSim variables Include relevant founder-paper work within this block.
Not set
☐
Fri
90
Check leakage, missing channels and fake-data reproducibility; prepare a one-command monitoring report.
Not set
☐
Sat
90
Resolve critical spillover; long-form draft if due; sandbox only if done.
Not set
☐
Sun
60
Build log, evidence, gap review, next actions.
Not set
☐
Dependency and capacity check
Missing essential inputs: not yet inspected.
Human/external dependency, owner and expected date: not yet inspected.
Week-specific constraint: Unfolding a whole batch can leak the future into early warning. Require causal feature availability for every alarm timestamp.
Existing session authorization for external actions or spending: not assessed; local planning only at scaffold creation.
Scope to cut first: optional polish/additional examples; preserve scientific validity.
If the forecast exceeds available time: record a concrete reduced scope or revised delivery proposal before committing.
Peter's commitment / adjustments: pending.
Reviewing model and date: pending.
Execution state: NOT_STARTED. Specification
This file is a blank register with planned artifact names; none of the rows prove completion.
Item
Path / URL / version
State
Verification / actual result
Verified by / date
monitoring-protocol.md
Not created
NOT_STARTED
Pending: Limit estimator, sample support and nominal alarm policy are reproducible.
—
alarm-examples.csv
Not created
NOT_STARTED
Pending: Event timing and undetected cases are retained; no cherry-picked pass rule.
—
verification-report.md
Not created
NOT_STARTED
Pending: The smoke run works without network or credentials and numerical invariants are checked.
—
Learning exercise / Peter's quiz
Not created
NOT_STARTED
Peter's actual answers required
—
Founder paper digest
Not created
NOT_STARTED
Source/content and access checked
—
Branding pack
Not created
NOT_STARTED
Claims audited; drafted/public status distinct
—
Opportunity actions
Not created
NOT_STARTED
Drafted/sent/replied/booked/held distinct
—
Reproduction and scientific measurements
Question / metric
Dataset, N and unit
Split / seed / config
Command or verification method
Output path and measured value
Uncertainty / limitation
External events, when authorized and actually performed
Date
Action / counterpart
Receipt or private reference
State
Next action / date
Actual learning and decisions
Peter's teach-back / quiz score: not assessed.
Scientific assumptions and deviations: none recorded yet.
Budget allocated / spent: not assessed / no spend recorded.
Missing evidence and responsible owner: pending inspection.
Do not put API keys, source full texts with restricted rights, client data or invoice personal details in a public version of this register. Keep private evidence references private.
☐ Limit estimator, sample support and nominal alarm policy are reproducible. Evidence: pending.
☐ Event timing and undetected cases are retained; no cherry-picked pass rule. Evidence: pending.
☐ The smoke run works without network or credentials and numerical invariants are checked. Evidence: pending.
☐ Learning assessed from Peter's actual answers. Evidence: pending.
☐ Branding/opportunity states recorded correctly. Evidence: pending.
Learning and outcome
What Peter can explain or reproduce now: not assessed.
Quiz / teach-back gaps: not assessed.
Measured result and limitations: pending.
Best evidence link for a scientist/recruiter/buyer: pending.
What failed or changed, and why: pending.
Gate and capacity
Gate requirement: No standalone gate; check this week's acceptance criteria and next gate dependencies.
Gate status: NOT_STARTED.
Original target week: See next gate in specs.
Actual completion date / forecast: not set.
Missing external evidence: pending inspection.
Blocks completed / 7: not recorded.
Days with commits / 7: not recorded.
Sunday build log / optional Wednesday note: not recorded.
Founder conversations cumulative: not recorded.
Earned evidence level and proof: not assessed.
Consecutive lost weeks: not assessed.
Week-specific risk to revisit: Unfolding a whole batch can leak the future into early warning. Require causal feature availability for every alarm timestamp.
Decision and next-model handoff
Continue / recover / blocked: not decided.
Retain as the minimum: T² limits computed.
Skipped permanently with reason: none recorded.
Carryover with owner, dependency and target week: none recorded.
Canonical repository paths and versions: pending.
Data permissions, split definitions or assumptions to preserve: pending.
External action authorization that still applies: not assessed.
Next physical action, exact file/function/paragraph: pending.
Updates to root PROGRESS / LOG / TODAY completed: ☐