The useful maintained tool is the exhaustive time–cost comparison. The two old source pages now have one Library home, with the article immediately below the tool. Original source hashes and all retained/corrected features are recorded in README.md.

Concrete findings addressed

  1. The original budget input had no effect on its optimiser. All knee, value and manual selections now respect the applied budget; unaffordable frontier points remain visible for comparison but disabled.
  2. The original payment expression produces zero for non-negative option costs and its guarantee claims lacked a specified procurement model. The executable payment output is removed; the original incentive-design ambition, externality idea and pilot survive as an explicitly unfinished proposal with primary theory background.
  3. Uncapped Cartesian enumeration could freeze the page. Bounds reject excessive cases before application, and accepted runs yield/cancel in chunks. No incomplete run is advertised as exhaustive.
  4. Original import replacement occurred before full validity was established. Input validation is atomic; async sequence guards protect both reading and enumeration. Imports cannot forge calculated schedule, selection semantics or welfare results.
  5. Float cost comparisons and duplicate rounded points could misrepresent decimal limits. Integer millionths determine schedules, exact budget feasibility and frontier comparisons; integer products determine value choice.
  6. Unsafe dynamic text/CSV paths and incomplete exports were replaced with DOM text nodes, quoted/formula-neutralised CSV and a reproducible result envelope containing source menus, applied settings and derived schedule/frontier context.
  7. The explanation now distinguishes a declared baseline, a deterministic idealised schedule, a geometric knee, a stated economic objective and a future mechanism. Resource/risk/authorisation limitations are attached to the concepts they qualify.

Suggested browser journeys

  • Open default example: declared baseline 25, 648 portfolios; compare net-value choice and knee. Set budget 0 and rerun: selected cost is 0; above-budget frontier choices are disabled.
  • Enter invalid activity JSON or an edge cycle and Apply: prior completed result remains. Apply a valid tiny network: old comparison clears until rerun.
  • Choose a frontier point with Enter or Space: choice becomes manual and keyboard focus remains on the corresponding control after rerender.
  • Export input/result JSON and both CSVs; reimport the result: outputs recalculate, with context intact. A download-request status means dispatch, not proof of filesystem save.
  • Open #explainer, #pilot and #procurement-proposal; the last opens its disclosure. Check phone width and internal table/chart scrolling.

Suggested catalogue copy

  • Title: Project Time Exchange
  • Subject: Delivery dynamics
  • Description: Compare bundles of task acceleration options against whole-project finish time and a budget.
  • Try: Set the A–G example’s budget to zero, then compare the knee and highest net-value choices with a larger budget.
  • Limit: Deterministic precedence-only CPM; declared costs and value, bounded exhaustive menus, no shared resources or supplier-payment mechanism.

Thirteen maintained model/UI tests passed on 2026-10-03. Independent generated-case review passed 80 cases. No browser interaction, commit, push, deployment or old-source edits were performed by this agent.