Projects by engineering subject

The current catalog contains 49 project entries: 40 standalone build briefs, five stages of one continuing reading-list project, and four runnable AI references. Choose a brief after its chapter; its baseline and deeper follow-ups stay together. A build brief is an assignment, not a claim that its application is already implemented. The historical 45-brief count excludes the four later AI references.

Continue one system through five stages

Problem solving and AI-assisted engineering

• Specify tag behavior for independent implementers
• Evaluate generated rate-limiter code with a simple reference
• Build reusable AI instructions and evaluate transfer
• Collect evidence about changed behavior and affected callers
• Record and revisit uncertain engineering decisions
• Rewrite private commit history to explain a change
• Split a tagging feature into runnable changes
• Demonstrate CI enforcement in a disposable repository
• Automate deterministic review rules and retain human judgment
• Check the combined behavior of independently valid changes

Backend and APIs

• Trace requests through a bookmark API
• Enforce a single deadline across API dependencies
• Build an SSRF-resistant link preview fetcher
• Evolve tag responses without breaking old clients
• Move title lookup into restartable background jobs
• Build a link monitor with durable history and change alerts
• Build duplicate-safe form submission and CSV export

Databases and transactions

• Store receipts with reviewable extraction and corrections

Frontend and full-stack integration

• Build a shared reading-list UI with private reading state

Testing, debugging, and code review

• Measure whether existing checks detect real defects
• Protect API response types, units and compatibility
• Reproduce and remove order-dependent failures
• Measure API capacity with controlled arrival rates
• Design and run a synthetic reading-list journey

Security

• Prevent overlapping shifts and enforce manager access

CI/CD and progressive delivery

• Deploy from main and roll back compatible artifacts

Observability

• Diagnose a reading-list incident from existing telemetry

Reliability and incident response

• Define and calculate a user-facing save SLO
• Implement burn-rate alert and incident state rules
• Bound retries across browser, API and SDK layers
• Prioritize API work within a fixed capacity budget
• Recover a service trapped in expired work and retries
• Keep bookmark saves usable when title lookup fails
• Rehearse detection, rollback and service recovery

Data at scale

• Build ingestion with bounded backlog and explicit rejection

AI systems

The first four projects include runnable references in examples/ai-systems; the fifth is a separate build assignment. Local fixtures do not certify real-model quality or live AWS deployment.

• Build evidence-backed answers with permission rechecks
• Require exact human approval before agent actions
• Route extracted invoices through validation and review
• Track AI evaluation evidence and serving versions
• Decide whether an AI feature improves a reading list

Migrations and recovery

• Migrate tags across data, clients and workers

Technical decisions and engineering effectiveness

• Build a service template with overridable defaults
• Write a data-platform policy from concrete decisions
• Compare a small export script with a custom platform

Continuing reading-list project: five build stages

These are assignments that evolve one application, not five supplied completed applications. The bookmark editor is a bounded reference slice, not the whole P1–P5 implementation.