Roadmap

What we're building, in what order, and what gates each phase. No hard dates — phases advance when conditions are met, not on a calendar.

How we build

Every feature passes the three-question test before it goes live:

1. Is it transparent? Does it explain, or obscure to create intrigue?

2. Does it point through? Does it lead toward evidence, or keep you on-site?

3. Is it invitational? Can you read it and close the tab with no friction?

Every financial mechanism passes four additional checks:

4. Opacity: Functional abstraction only — no engineered complexity.

5. Response: Signals the problem, not the participant.

6. Attention: Points at the work, not at the money.

7. Termination: Has a designed end — no infinite engagement loops.

Money is a void. The framework says so. So the project scores its own financial design against the same diagnostic it uses on everything else. Full disclosure →

Active

Alpha — Ship the diagnostic

Get the core tools and research live. Static site, no accounts, no tracking.

Score tool — evaluate any system for manipulation risk
Three research papers (v12.2, v4.1, v6.2)
Bounty board — 22 falsification conditions, counter-examples pay 2x
Learn pages — architecture, evidence, safe design
Articles — framework applied to real-world systems (3 published)
Vocabulary scorer — check language for drift patterns
Methodology — codebook, rubric, and protocols downloadable
Deploy site
Publish LessWrong article
Submit Paper 1 to arXiv
Set bounty amounts ($500 USD per test, paid in $MORR at claim-day rate)
Write 2-3 more articles

Gate to Phase 2: Site live. At least one paper or article published externally. Score tool used by people who didn't build it.

Phase 2 — Validation and traction

Other people use the framework. We measure whether it works when the author isn't running it.

Independent blind application — someone else applies the diagnostic to a new domain
Inter-rater reliability at scale — multiple scorers, same systems, measure agreement
Crowd-sourced scoring — structured template for community domain analysis
Paper 2 and Paper 3 external submission
Backend deployment — data collection for submitted scores
Evidence browser — interactive 90-domain analysis explorer
Community governance — $MORR token voting on void classifications (productive vs destructive), scored against the same four financial design checks

Gate to Phase 3: Independent replication of at least one core result. Inter-rater kappa > 0.7 with non-author scorers. Peer review submitted.

Governance Triggers

When governance changes happen

The project starts as a solo founder operation. That's an honest 6/12 on the self-score. These are the specific, measurable conditions that trigger governance transitions. No condition, no change. No calendar — thresholds.

100 unique scorers (distinct IPs/wallets submitting void index assessments) → Community advisory input on classification disputes (productive vs destructive). Founder retains final call.
1 independent replication of any core experiment by a non-author researcher → External reviewer appointed with public kill-condition authority. Changes governance geometry from single-point to two-point.
Inter-rater kappa > 0.7 with 5+ non-author scorers → Community scoring accepted into the public database without founder review.
3 independent replications by separate teams → Framework stewardship committee (3 members: founder + 2 elected). Committee governs methodology changes. Founder retains veto on kill condition adjudication only.
$50K annual service revenue → Public treasury dashboard with quarterly operational reporting.
Peer review acceptance at a recognized venue → Full governance proposal published for community input.

The principle: Governance decentralizes when evidence justifies it, not when popularity demands it. Each threshold is measurable, published before the condition is met, and visible in the git history if changed.

Future

Phase 3 — Scale

If the framework survives independent validation, build infrastructure for widespread use.

Framework assistant — constrained AI chatbot with full void disclosure
Automated scoring API — programmatic access to the diagnostic
Continuous monitoring — track systems over time
3D risk map — interactive visualization of scored systems
Preservation layer — papers on IPFS + Arweave, verifiable hashes
Structured engagement study — opt-in measurement with human participants

Gate: Framework accepted for peer review at a top AI safety venue. Independent teams using the diagnostic without author involvement.

Ongoing

Research — always running

The framework has 22 falsification conditions. Any one of them could kill it. We're looking.

Kill condition monitoring — tracking all 22 conditions against new evidence
Domain expansion — testing universality beyond 90 domains
Productive void mechanics — when the same architecture produces discovery instead of harm
Cross-domain thermodynamic measurements
Epistemological implications — what the framework says about how we know things

What we won't build

We're keeping the site simple and transparent by design. No engagement tricks, no retention mechanics, no financial mechanisms that fail the four checks.

User accounts or profiles — the tool works without them
Algorithmic feeds or personalization — everyone sees the same content
Engagement optimization — no dark patterns, no retention hooks
Analytics beyond basic traffic counts — we don't optimize for your attention
Yield or staking mechanics — compounding rewards have no designed end (fails termination check)
Leaderboards or earning streaks — gamified contribution is observer-targeting (fails response check)
Token marketing from treasury — steepens the attention gradient toward the money (fails attention check)
Complex DeFi integrations — layered financial products add extractive opacity (fails opacity check)

Want to break the framework? Bounties pay 2x for counter-examples.
Track progress: github.com/AnthonE/morr-papers