Elements of Scalable Web Development

Part III

Compare — Evaluate workflows and scaling responses#

Implementations A and B expose what changed across two single-agent development systems, but they do not yet show what orchestration adds or costs. Compare produces implementation C through a declared agent fleet, applies common scaling pressures to all three applications, and turns their evidence into a justified target architecture and development workflow.

III.1 From orchestration to an evidence-based decision#

The three chapters form one progression:

  1. When do agent fleets improve web development? grows one agent into a bounded fleet, makes authority and integration explicit, and treats failure correction as part of implementation C.
  2. How should web applications respond to scaling pressure? identifies approaching limits, applies formal models, and compares possible responses without treating a technique catalog as a design.
  3. What do three agent-assisted implementations teach us about web scalability? connects the application, development, and agentic loops and turns the three evidence packages into a final decision.

Complete Compare with implementation C and its fleet trace, a three-way application and workflow comparison, and a target design whose claims distinguish observation, inference, and proposal.

III.2 Implementation C — Work Activity Tracker#

The Compare project specification defines the third fresh implementation and its bounded fleet workflow. The common behavior, permitted acceptance material, pressures, and reporting categories remain stable while delegation, orchestration, authority allocation, integration, and fleet controls vary deliberately. Use B's small model as the reference configuration; an approved fleet may use small models, large models, or a mixture with declared role or task assignments. When model assignment also changes, the full B–C builds compare development systems rather than isolate orchestration. The supplied matched-trace activity holds the small model fixed; students may optionally run their own matched pair or vary one model assignment with the task graph fixed.

Analyze the actual fleet run through a compact redacted execution record and debugging findings about attempts, failures, handoffs, validation, correction and human intervention. A curated log excerpt or table is sufficient. As an optional extension, transform and load the trace into the Work Activity Tracker, then use its views to investigate coordination and correction.

III.3 Compare applications and development systems#

Explain what each setup made easier or riskier, which application architecture best fits the selected pressure, and what the team can sustain.

Record architectural, configuration, and practice differences as alternative explanations for the result. The final decision should name the conditions under which you would choose the application and development workflow.