Expand capability coverage benefit
Represent rich interaction, reactive state synchronization, backend/data behavior, and asynchronous workflows as weighted requirements.
Evaluation roadmap
A deliberately staged list for making stack comparisons more useful without turning the model into an opaque optimizer.
Now
Review every scored cell evidence
Replace generic judgments with inspectable reasoning and provenance. Keep measured facts distinct from reviewed assessments.
Boundary enforcement factor
Evaluate component isolation, explicit data flow, generated contracts, build-time checks, and ownership of cross-layer interfaces.
Next
Make weights obvious
Keep requirement and factor weights together in one clearly editable model, with revisions preserving each decision basis.
Integration edge costs graph
Estimate the cost of each component boundary using reviewed ordinal ratings or measured glue code, then expose every active edge.
Portfolio scenarios
Compare stacks across forms/CRUD, interactive canvas, API service, and asynchronous workflow applications before recommending a standard.
Sensitivity analysis
Perturb weights individually and through randomized samples; report winner stability instead of presenting one ranking as inevitable.
Later
Complexity v2 research
Retain deduplicated LOC as complexity_v1, then evaluate import-graph fan-in/fan-out and Git co-change coupling as stronger signals.
Agent-development fit
Find observable proxies for documentation quality, API stability, generated-code verifiability, and test feedback speed before scoring this dimension.
Candidate graph search
Only after the scoring model is trusted, test a small greedy set-cover approximation for suggesting—not silently selecting—complete stacks.