“Current artifact” is a public owner-stated summary, not a maturity claim.
Supporting material may be public, non-public, or still awaiting a project-owned witness.
01 / 14
Writing · story systems
Whisper
What if a story could remember itself?
A writing studio where intent stays visible and the world behind a story can become
explorable—not another box that turns a prompt into prose.
Current artifactA reviewable intent → plan → draft → validation lifecycle.
Open the project
Vision
An agentic writer studio—where AI can carry out bounded, reviewable steps—with a knowledge graph: a map of characters, places, encounters, timelines, and connections that stays available while people write and read.
What exists now
A reviewable composition lifecycle that preserves audience, structure, voice, evidence, movement, review, and what the team learned.
Not yet
Not a finished writer studio or a live story knowledge graph. Broader transports and quality outcomes remain unproved.
Next experiment
Keep one bounded story map synchronized with one narrative excerpt as characters move, meet, and remember.
Possible collaborators
Writers, editors, narrative designers, readers, and knowledge-graph thinkers.
Talk about Whisper →
Possible contribution, not an open role. Confirm welcome and scope with a human owner.
02 / 14
Learning · explanation
User / Translate / Guide
What if a guide began with what you already know?
A candidate guidance pattern that translates an unfamiliar system through a reader’s
existing mental model—and then asks them to use it.
Current artifactA three-role guidance pattern and browser explainers.
Open the project
Vision
Explanations that locate a learner, translate through a known anchor, and test whether the new model can be retold and applied.
What exists now
A candidate User / Translate / Guide pattern plus concrete layered browser explanations.
Not yet
Preference is not mastery. Consent, updates, durable learner state, and universal-tutor claims remain unresolved.
Next experiment
Ask newcomers to retell and teach back one difficult concept without borrowing the original wording.
Possible collaborators
Educators, technical writers, workshop designers, and curious newcomers.
Talk about User / Translate / Guide →
Possible contribution, not an open role. Confirm welcome and scope with a human owner.
03 / 14
Music · creative tools
Agentic Music Studio
What if AI exposed musical decisions instead of skipping over them?
A studio concept where musicians edit shared musical intent, rules, structure, and
history instead of surrendering craft to one prompt.
Current artifactA closed research tower, interaction taxonomy, and claim ledger.
Open the project
Vision
A musician can shape an editable musical specification, hear a realization, revise one decision, and trace how the piece changed.
What exists now
Research on interaction processes, a bounded taxonomy, explicit claims, and a model of shared musical specifications.
Not yet
No working studio and no validated model of musical perception or delegated musical behavior.
Next experiment
Build one intent → editable specification → repeatable sound example and review it with musicians.
Possible collaborators
Musicians, live coders, audio-tool builders, interaction designers, and listening-study partners.
Talk about Agentic Music Studio →
Possible contribution, not an open role. Confirm welcome and scope with a human owner.
04 / 14
Games · agent evaluation
Poker Agent Arena
What makes an AI decision trustworthy enough to replay?
A constrained arena where agents compete under explicit rules and a human can inspect
legality, decisions, and replay before believing a ranking.
Current artifactA runnable local Kuhn-poker arena, bot contract, replay, and receipt.
Open the project
Vision
An arena for testing agent decisions where rules, seats, actions, randomness, and results stay reproducible and inspectable.
What exists now
A local Kuhn-poker command-line arena with constrained bots, legal decisions, seat rotation, replay, and a result receipt.
Not yet
Not Hold’em, hosting, real money, a final leaderboard, or a general proof of agent quality.
Next experiment
Add one bounded reveal or Leduc fixture without weakening legality, determinism, or replay.
Possible collaborators
Bot authors, game theorists, adversarial testers, tournament designers, and replay-tool builders.
Talk about Poker Agent Arena →
Possible contribution, not an open role. Confirm welcome and scope with a human owner.
05 / 14
Movement · game systems
Body War / TROQAR
Can a real run become a shared strategy game without betraying the runner?
Physical movement changes a common map—but only if safety, consent, fairness, and the
evidence behind a run stay visible.
Current artifactAn app-backed pilot and a fake-Strava end-to-end demonstration.
Open the project
Vision
A territorial strategy game shaped by real movement, shared rules, and evidence players can inspect and contest.
What exists now
An app-backed pilot and a fake-Strava end-to-end path that exercises a narrow activity flow.
Not yet
A narrow pilot, not a complete territorial game. Naming, safety, privacy, spoofing, fairness, and access remain open.
Next experiment
Validate one track-evidence and zone-update slice with real runners and explicit consent boundaries.
Possible collaborators
Runners, game designers, geospatial builders, privacy reviewers, and anti-spoofing thinkers.
Talk about Body War / TROQAR →
Possible contribution, not an open role. Confirm welcome, name, and scope with a human owner.
06 / 14
Design · reviewable iteration
UI Prototyping Studio
Can an interface change without losing why it changed?
Generate, compare, annotate, approve, apply, and trace interface variants without
letting visual iteration become an invisible automatic rewrite.
Current artifactA runnable L0 mutation loop, preview surface, and end-to-end evidence.
Open the project
Vision
A prototyping studio where alternatives, feedback, decisions, application, reversal, and receipts stay connected.
What exists now
A runnable minimum mutation loop, preview surface, explicit acceptance/revert path, and browser evidence.
Not yet
Later workflow layers and the canonical public/private root remain unresolved. The preview is not a complete product.
Next experiment
Carry one accessibility or layout change through generation, annotation, acceptance, application, evidence, and revert.
Possible collaborators
Designers, frontend builders, accessibility testers, reviewers, and UI-tool researchers.
Talk about UI Prototyping Studio →
Possible contribution, not an open role. Confirm welcome and canonical scope with a human owner.
07 / 14
Knowledge · disagreement
Tower System
How do we stop one agreeable AI voice from flattening real disagreement?
A learning and research system where sources, opposed stances, decisions, and ownership
remain inspectable instead of collapsing into one smooth answer.
Current artifactA product/research model and precedent map.
Open the project
Vision
Knowledge towers that retain provenance, tension distinct perspectives, and make the resulting decision path visible.
What exists now
A product and research model, sourced stance-pair work, and a precedent map for multi-perspective learning.
Not yet
The central persona → tension → grading assembly has no working example, and no live MOGT runs exist.
Next experiment
Hand-build one proof target with genuinely opposed sources, explicit ownership, and a reviewable decision.
Possible collaborators
Knowledge engineers, researchers, teachers, librarians, and argument designers.
Talk about Tower System →
Possible contribution, not an open role. Confirm welcome and scope with a human owner.
08 / 14
Developer tools · authorization
IDE Extension
What should happen before an AI-generated change becomes durable?
A reviewable workbench that keeps observation, proposal, approval, execution, and receipt
distinct when agent work approaches lasting state.
Current artifactA local workbench, schemas, kernel, L0 screen, and tests.
Open the project
Vision
A developer can inspect what an AI proposes, authorize a bounded action, and verify the resulting receipt without surrendering control.
What exists now
A local workbench, explicit schemas, a kernel boundary, an L0 screen, and focused tests.
Not yet
No autonomous worker, live provider, browser controller, or shipping IDE wrapper.
Next experiment
Run a bounded set of approve, reject, retry, and recovery trials against the local demonstration.
Possible collaborators
Developers, security reviewers, extension designers, protocol thinkers, and failure testers.
Talk about IDE Extension →
Possible contribution, not an open role. Confirm welcome and scope with a human owner.
09 / 14
Developer experience · routing
IDE Ergonomics
What if the useful fragment in a conversation could be routed when it appears?
Small, reversible ways to turn a selected phrase into the right next output or action
without turning the entire conversation into a workflow.
Current artifactA candidate product view, prototype, and fixture plan.
Open the project
Vision
Select the moment that matters, choose a bounded route, understand the consequence, and keep the rest of the conversation intact.
What exists now
A candidate product strategy, route-card interaction prototype, and fixture/evidence plan.
Not yet
No full plugin, execution layer, receiver integration, or tested productivity claim.
Next experiment
Observe real route choices and whether people understand the consequence without excessive choice burden.
Possible collaborators
IDE researchers, developers, interaction designers, technical writers, and working users.
Talk about IDE Ergonomics →
Possible contribution, not an open role. Confirm welcome and receiver scope with a human owner.
10 / 14
Explanation · visualization
X-Ray
What if a complex system could be read in layers?
A visual explanation surface for properties, components, flows, dependencies, and the
evidence boundary behind each claim.
Current artifactA public seed, schemas, validators, examples, and generated explainers.
Open the project
Vision
Let a reader move from the whole system to the layer they need without erasing sources, dependencies, uncertainty, or ownership.
What exists now
A public seed, input and lane schemas, validators, examples, and generated HTML explanations.
Not yet
Not promoted as a production visualization engine; reusable live-behavior and comprehension proof remain limited.
Next experiment
Build one source-marked layered explanation and ask cold readers to retell the system and its limits.
Possible collaborators
Information designers, technical educators, systems thinkers, accessibility reviewers, and cold readers.
Talk about X-Ray →
Possible contribution, not an open role. Confirm welcome and scope with a human owner.
11 / 14
Learning · practice by making
Arcanum Journey
What if documentation taught by making?
A guided path where each chapter produces, inspects, and tests one real output instead
of asking a newcomer to memorize a framework.
Current artifactAn L0 scaffold and the first two chapters.
Open the project
Vision
A learn-by-building journey through capabilities, composition, evidence, and reflection—one concrete output at a time.
What exists now
An L0 learning scaffold with chapters zero and one and a make-first instructional direction.
Not yet
Later chapters, full wiring, learner evidence, and promotion remain pending.
Next experiment
Have one novice complete and teach back one make-first chapter without facilitator repair.
Possible collaborators
Curriculum designers, newcomers, workshop facilitators, technical writers, and reviewers.
Talk about Arcanum Journey →
Possible contribution, not an open role. Confirm welcome and scope with a human owner.
12 / 14
Creative growth · evidence
AutoAds
Can ad generation learn without turning metrics into mythology?
A traceable creative loop where variants, measurements, grading, and owner-promoted
learning remain traceable.
Current artifactA validated seed and non-executed design.
Open the project
Vision
Generate creative variants, connect them to bounded evidence, grade with explicit criteria, and let an owner decide what becomes reusable learning.
What exists now
A validated project seed, product framing, and a non-executed design for the learning loop.
Not yet
No live ad network, spend, training system, market result, ROAS automation, or performance proof.
Next experiment
Grade a small synthetic or historical variant set without spending money or promoting a causal claim.
Possible collaborators
Marketers, copywriters, visual creatives, causal analysts, and experiment designers.
Talk about AutoAds →
Possible contribution, not an open role. Confirm welcome and scope with a human owner.
13 / 14
Quality · orchestration
Sonar Loop
What if quality work ended in a clear decision?
One auditable loop from selecting a quality gate to gathering evidence and deciding
continue, pause, halt, or complete.
Current artifactA docs-first TypeScript specification.
Open the project
Vision
Quality tools become one reviewable decision loop with evidence, ownership, explicit stop conditions, and a clear close.
What exists now
A docs-first TypeScript specification for the loop, boundaries, and decision vocabulary.
Not yet
Documentation does not prove implementation, integration, remediation behavior, or completion.
Next experiment
Manually run one thin repository slice through one reviewable evidence-to-decision pass.
Possible collaborators
Maintainers, quality engineers, repository owners, evidence reviewers, and tool integrators.
Talk about Sonar Loop →
Possible contribution, not an open role. Confirm welcome and scope with a human owner.
14 / 14
Metrics · decision trust
Ontologize
Can a business metric show how it came to be believed?
A practice for tracing metrics to events, boundaries, confounds, decisions, and owner
review instead of treating a dashboard number as self-explanatory truth.
Current artifactCandidate engagement and founding-pilot material.
Open the project
Vision
Business decisions can inspect where a metric came from, what could distort it, who owns its meaning, and what learning is justified.
What exists now
Candidate engagement material, a metric-trust framing, and a founding-pilot package.
Not yet
No broad buyer proof, committed brand, pricing, contracts, cross-vertical advantage, or autonomous metric truth.
Next experiment
Trace one disputed metric from source event through boundary and confound to a bounded human decision.
Possible collaborators
BI practitioners, data modelers, operators, domain owners, and skeptical decision makers.
Talk about Ontologize →
Possible contribution, not an open role. Confirm name, welcome, and scope with a human owner.