Skip to content

feat: propose domain-aware plan ranking engine with relative comparison#62

Merged
neoneye merged 3 commits intoPlanExeOrg:mainfrom
VoynichLabs:feature/domain-aware-ranking-proposal
Feb 22, 2026
Merged

feat: propose domain-aware plan ranking engine with relative comparison#62
neoneye merged 3 commits intoPlanExeOrg:mainfrom
VoynichLabs:feature/domain-aware-ranking-proposal

Conversation

@82deutschmark
Copy link
Copy Markdown
Collaborator

Domain-Aware Plan Ranking Engine Proposal

This proposal addresses the problem of universal rubrics breaking down when comparing plans across different domains.

Key Changes

  • Domain Classification — Detect plan type (construction, software, marketing, operations, research, business-development)
  • Domain-Specific Signal Extraction — Pull KPIs relevant to that domain instead of generic metrics
  • Corpus Bucketing — Group plans by type for fair apples-to-apples comparison
  • Relative ELO Ranking — Score each plan against similar plans in its domain, not in a vacuum
  • Actionability Scoring — Surface top-performing plans (>90th percentile) as refinement candidates vs. major-rewrite situations

Problem Solved

Current evaluation assumes a universal rubric (concreteness, executability, success criteria). This breaks when comparing a road construction plan to a software project—they have completely different success signals.

With domain awareness, we can:

  • Compare software plans against other software plans (not construction)
  • Extract domain-specific KPIs (MVP clarity + tech debt for software vs. budget contingency + inspection checkpoints for construction)
  • Give percentile rankings that actually mean something (top 10% within its domain)
  • Make actionability decisions (fixable? major rewrite? template-worthy?)

Implementation Plan

4 phases over ~10 days:

  1. Domain classifier (2 days)
  2. Domain-specific extractors (3 days)
  3. ELO ranking engine (4 days)
  4. Actionability + APIs (2 days)

See full proposal for architecture, data model, test strategy, and risk mitigation.

Larry the Laptop Lobster added 3 commits February 20, 2026 21:50
- Replace hallucinated /api/plans/*, /api/rank, /api/leaderboard endpoints with TBD notes
- Fix invalid planexe.org domain URLs → home.planexe.org
- Replace hallucinated tool names: create_plan, check_plan_status → task_create, task_status
- Remove non-existent tool: get_next_action_item
- Add implementation status notes to /openapi.json and /.well-known/mcp.json recommendations
- Clarify that semantic search API and Elo ranking APIs are proposed local features, not part of public MCP
- Add domain classifier to detect plan type (construction, software, marketing, ops, etc.)
- Implement domain-specific KPI extraction (timeline, resources, risk, owner clarity)
- Corpus bucketing for fair comparison within domain context
- ELO ranking engine for relative plan scoring (not absolute)
- Actionability scoring to surface refinement candidates vs. rewrite situations
- 4-phase implementation roadmap (10 days total)
- Full API spec and data model
@neoneye neoneye merged commit 69b765e into PlanExeOrg:main Feb 22, 2026
@neoneye neoneye deleted the feature/domain-aware-ranking-proposal branch February 22, 2026 20:00
neoneye added a commit that referenced this pull request Mar 9, 2026
Improve discoverability for AI crawlers and agents:
- New public/robots.txt with AI-specific user agent rules
  (GPTBot, ChatGPT-User, Claude-Web, Anthropic-AI, etc.)
- Add "For AI Agents" section to README.md with discovery links,
  MCP endpoint, workflow summary, and key zip file paths

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
neoneye added a commit that referenced this pull request Mar 11, 2026
Add AI agent discoverability: robots.txt and README section (#62)
neoneye added a commit that referenced this pull request Mar 11, 2026
Mark #86, #62, #114-I4 as implemented. Note open PRs for #58,
#88, #94, #59. Update phase sequence with status indicators and
next priorities (114-I1, 114-I2).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants