All chapters
  1. Start hereA software factory, from first principles
  2. Primitive · 1A factory, before AI: input, transformation, check, output
  3. Primitive · 2From repetition to a pattern: what stays, what varies, what to leave alone
  4. Primitive · 3The specification as an executable contract
  5. Primitive · 4Build the smallest factory: four stations, two specs, one manifest
  6. Primitive · 5Assembly stations: typed contracts, owners and failure handling
  7. Primitive · 6AI enters the factory: what a model adds, and what stays deterministic
  8. WatchAlgo · 7The WatchAlgo production story: what the repository shows
  9. WatchAlgo · 8A problem as a production unit: the files one unit must contain
  10. WatchAlgo · 9Station by station: description, solution, implementations, visualization
  11. WatchAlgo · 10The visualizer factory: from algorithm state to renderable steps
  12. WatchAlgo · 11Three flavors and three languages: two different contracts
  13. WatchAlgo · 12Looping over a large backlog: scheduling, limits and a failed-case queue
  14. WatchAlgo · 13Long-running agent execution: context, checkpoints and token limits
  15. WatchAlgo · 14Evidence of correctness: tests the producer cannot talk its way past
  16. WatchAlgo · 15The factory learning loop: defects change the pattern, not only the unit
  17. Enterprise · 16Classify the enterprise demand before choosing a pattern
  18. Enterprise · 17Reference enterprise patterns and their variation points
  19. Enterprise · 18Platform foundations: golden paths, templates, registries and catalogs
  20. Enterprise · 19Pattern versus configuration versus business logic
  21. Enterprise · 20Insurance factory, part 1: building the intake contract
  22. Enterprise · 21Insurance factory, part 2: building the underwriting workflow
  23. Enterprise · 22Insurance factory, part 3: trust and observability built in
  24. Enterprise · 23Data and integration engineering: contracts that survive change
  25. Enterprise · 24The CI/CD and deployment factory: from pull request to production
  26. Enterprise · 25Enterprise governance: owners, approved patterns and provenance
  27. Enterprise · 26Scaling safely: many teams, many agents, one set of boundaries
  28. Operations · 27Factory quality engineering: measuring accepted output, not activity
  29. Operations · 28Failure and recovery scenarios: what breaks a factory, and what restores it
  30. Operations · 29The landscape: which tool for which job, read from its own documentation
  31. Operations · 30Design an enterprise software factory: the blueprint and the rollout
  32. ReferenceSources and verification
  33. ReferenceGlossary
  34. AboutAbout this site
Reference

Sources and verification

Every outside source with the date it was read, the WatchAlgo evidence matrix in brief, what was run and what was not, and the facts that could not be verified.

Every claim on this site is one of four kinds, and this page shows where each kind comes from: code that was run, a repository that was read, documentation that was read on a stated date, or the author's own report. If a page says something that is not traceable to one of these, that is a defect; the About page explains the label convention.

What was run#

All examples are standard-library Python 3.14, run on 2026-11-20 with no network, no API key and no model call. Where a model would act, a scripted stand-in plays its part.

ExampleCommandResult on 2026-11-20
A assembly linecd examples/a-assembly-line && python3 -m unittest discover -s tests -t .14 tests OK; two runs give byte-identical manifests
B algorithm unit factorysame, in examples/b-algorithm-factory32 tests OK; Python, Node v26.3.1 and Java 23.0.1 agree on all 10 shared fixture cases
C insurance service factorysame, in examples/c-insurance-service, then python3 ci_checks.py50 tests OK; four CI gates pass
D factory supervisorsame, in examples/d-supervisor, then sh check_repro.sh21 tests OK; resume after a crash writes nothing twice
Batch simulator and cost modelsame, in examples/simtests OK; JavaScript port equal on every parity vector
WatchAlgo inventory adapterpython3 examples/watchalgo-adapter/inventory.py (read-only)counts in the evidence summary below

The labs re-run the same logic in the browser; node scripts/test_factory_core.mjs checks that the JavaScript gives the same results as the Python on every recorded vector. Simulator values are hypothetical and labelled "simulated, illustrative" wherever they appear.

Not run here, and said so on the pages that show them: Pydantic, Jakarta Bean Validation, Pact, Testcontainers, the CI products, the workflow engines, LangGraph, the Claude Agent SDK, Claude Code and Codex. Their samples are "documented shape; not run here", written from their documentation read on the date below.

The WatchAlgo evidence, in brief#

The WatchAlgo repositories were read with read-only commands on 2026-11-20 and nothing in them was changed. The full matrix, with every path and commit, is docs/EVIDENCE.md in this site's repository; the chapters cite the same paths.

ClaimObserved in the repositoryReported by the author
Problem count3,246 well-formed wa-NNNNNN folders with a problem.json; one more, empty folder named by an unexpanded shell expression3,246 implemented problems
Solutions and visualizations2,773 units with solution.json; 2,380 with visualization.json; all 2,380 complete units have three flavors and three languagesdescriptions, solutions and visualizations for all of them
Three flavorsnarrative styles technical, fun, spiritual; code separately in JavaScript, Python and Javathree flavors
Specialized agents per stationqueue pipeline scripts and commit messages describe solution and visualization tasks run by command-line agent workers, verified after each taskspecialized agents handled station responsibilities
30+ hour runsthe September supervisor log spans 62 hours, with two productive windows of roughly a day each; checkpoint commits every 30 minutes; account usage-limit messages, not context windows, end the runs30+ hour sessions until token limits required intervention
Telemetryno per-session token, cost or duration telemetry was found(none claimed)

Time spans between log lines or commits are not measurements of how long an agent worked. The WatchAlgo chapter explains each row.

Outside sources and the date they were read#

Every source below was read on 2026-11-20 on its official site. A second, independent reader re-fetched every surprising item (renames, deprecations, new versions) the same day: 85 confirmed, 3 corrected, 2 not checked and not used. Per-page claims are logged in docs/facts/<page>.md.

SubjectKindOfficial source
Agentic Engineering (Martin Fowler)articlemartinfowler.com
Harness engineering for coding agent usersarticlemartinfowler.com
Humans and Agents in Software Engineering Loopsarticlemartinfowler.com
Structured-Prompt-Driven Developmentarticlemartinfowler.com
Platforms white paper (CNCF TAG App Delivery)guidelinetag-app-delivery.cncf.io
Software product lines (Carnegie Mellon SEI)guidelinesei.cmu.edu
Architecture decision recordspatternadr.github.io
Backstage software templatesopen-source projectbackstage.io
Cookiecutteropen-source projectcookiecutter.readthedocs.io
Yeomanopen-source projectyeoman.io
GitHub Actionsproductdocs.github.com
GitLab CI/CDproductdocs.gitlab.com
Jenkins Pipelineopen-source projectjenkins.io
Temporalopen-source projectdocs.temporal.io
Argo Workflowsopen-source projectargo-workflows.readthedocs.io
Prefectopen-source projectprefect.io
Dagsteropen-source projectdagster.io
LangGraphopen-source projectdocs.langchain.com
Claude Agent SDKproductcode.claude.com
Claude Codeproductcode.claude.com
Codexproductlearn.chatgpt.com
OpenAPIspecificationspec.openapis.org
JSON Schemaspecificationjson-schema.org
Pydanticopen-source projectpydantic.dev
Jakarta Validationspecificationjakarta.ee
Pactopen-source projectdocs.pact.io
Testcontainersopen-source projecttestcontainers.com
SLSAspecificationslsa.dev
in-totospecificationin-toto.io
Sigstoreopen-source projectdocs.sigstore.dev
SPDX and CycloneDXstandardspdx.dev, cyclonedx.org
NIST SSDFguidelinecsrc.nist.gov
OWASP GenAI Security Projectguidelinegenai.owasp.org
OpenTelemetryopen-source projectopentelemetry.io
DORA metricsguidelinedora.dev

Facts that could not be verified#

These are worded on the pages as unverified, or left out.

ItemWhy
The ISO status of SPDXiso.org refused automated reading; the pages cite SPDX's own announcement only
Default retry values in Dagster and Prefect; LangGraph's default durability modethe documentation read here does not say
Codex install commands and default approval policynot reached on the pages read
Parts of the Claude Code hooks page, the Jakarta Validation Validator signatures, the Jenkins input steppages too long to read in full; nothing from the unread parts is stated
Authors and date of the CNCF platforms white paperthe page shows neither
How the 3,246 WatchAlgo problem statements were generatedno record was found in any repository or history
Token use, cost and duration of WatchAlgo sessionsno telemetry exists; the author's estimates are labelled as estimates
Who reviewed WatchAlgo contentthe review files found are model-written reviews; no human review record was found