Longterm Wiki

Project Roadmap

This file tracks future infrastructure improvements and project priorities. For content-specific tasks, see the Enhancement Queue.

Current State (January 2026)

The project has mature infrastructure:

SystemStatusNotes
Validation Suite✅ Complete14 validators covering style, links, MDX, Mermaid, staleness, redundancy
Quality Grading✅ CompleteClaude API-powered grading (0-100 quality/importance)
Dashboard✅ Complete/dashboard/ with quality distribution, /dashboard/graph/ for relationships
Knowledge Base✅ CompleteSQLite cache with article/source extraction and AI summaries
Resource Linking✅ Complete<R> component with bidirectional tracking, metadata extraction
Page Type System✅ Completeoverview/content/stub classification
Freshness Tracking✅ CompletereviewBy, contentDependencies, staleness validator

Future Work

Content Quality

  • Batch content improvement - Systematically upgrade high-importance, low-quality pages

    • Use node scripts/page-improver.mjs --list to identify priorities
    • Target: All pages with importance > 60 and quality < 40

  • Citation coverage - Many pages lack external citations

    • Resource manager can convert links: npm run resources process [page] --apply
    • Goal: Every substantive claim has a source

Tooling (Low Priority)

  • Vale integration - Prose linting for terminology consistency

    • Would catch mechanical style issues (hedging, passive voice)
    • Deferred: Current validators cover structural issues well

  • VS Code snippets - Scaffolding templates for new pages

    • Low effort (~30 min), zero installation
    • Templates for: risk, response, model, debate pages

Style Guide

  • Common writing principles - Cross-cutting style guide for epistemic honesty, language neutrality, analytical tone

    • Created at /internal/common-writing-principles/
    • Added 7th rating dimension: objectivity (0-10)
    • All domain style guides reference it

  • Re-grade pages with objectivity dimension - Existing pages don't have objectivity scores

    • Run grade-content.mjs with --apply to add objectivity ratings
    • Prioritize analysis/model pages where epistemic honesty matters most

  • Audit existing pages for insider language - Many pages likely use "EA" as adjective

    • Could build a crux validator to flag "EA money", "non-EA", "EA causes" etc.
    • Apply going forward; retroactive cleanup is lower priority

  • Model page guidance - Models style guide complete at /internal/models-style-guide/

  • Component usage patterns - Document when to use:

    • EstimateBox vs plain tables
    • Mermaid diagrams vs prose explanations
    • KeyQuestions component

Completed

December 2025

Infrastructure:

  • Quality dashboard at /dashboard/
  • Freshness tracking (reviewBy, contentDependencies, staleness validator)
  • Page type system (overview/content/stub)
  • Redundancy detection validator
  • Wiki statistics on About page
  • Mermaid diagram validator
  • Comparison operator validator
  • Dollar sign escaping validator

Content:

  • Entity naming refactor for consistency
  • Parameter pages restructured
  • Concepts directory page (LessWrong-style layout)

Earlier

  • kb-2.0 style guide
  • Full validation suite (style, links, MDX, data integrity, etc.)
  • Knowledge base system (scan + summarize)
  • Resource manager CLI
  • Quality grading system

Notes

  • Don't over-engineer - Add tooling only when manual process becomes painful
  • Style guide is guidelines, not law - Adapt to content, don't force templates
  • Run npm run validate before committing to catch issues early