Reframe README as Epiphany front door

This commit is contained in:
Meta Vens 2026-06-11 17:13:09 +01:00
commit 9c9b4a7849
2 changed files with 186 additions and 158 deletions

340
README.md
View file

@ -1,192 +1,220 @@
# EpiphanyAgent
# Epiphany
Epiphany is an opinionated fork of Codex built around one demand: the model
must model the thing it is changing.
<p align="center">
<img src="docs/assets/epiphany-avatar-4x.png" width="1024" alt="Epiphany avatar portrait targeting JSON heresy" />
</p>
The problem this repo is chasing is not "LLMs can't write code." They can. The
problem is that they can keep making plausible local moves after the global
design has already wandered off and died in a ditch.
Epiphany is GameCult's agent-control body: the system that lets AI agents do
real project work without hiding their memory, authority, evidence, or mistakes
inside a chat transcript.
Epiphany answers that by moving important state out of transcript fog and into
typed, inspectable surfaces: maps, evidence, retrieval state, graph/frontier
state, churn pressure, and compact client reflections. The goal is not maximal
output. The goal is a machine whose parts visibly deserve to exist.
The bet is simple and unpleasantly large:
## Project Direction
> AI agents are becoming capable enough to matter, but organizations cannot
> safely route important work through them until the work is inspectable,
> reviewable, permissioned, and attached to receipts.
Epiphany is moving toward project-native agency.
Epiphany is the control plane for that missing layer. It makes the agent show
its map, preserve evidence, separate roles, survive context loss, and stop when
it no longer understands the machine it is changing.
The user should not have to write a perfect brief, name every subsystem, and
pre-plan the code flow for the whole architecture before the machine can help.
That is just outsourcing project management to the human and calling the agent
smart because it waited politely.
This is not a faster autocomplete costume. It is the beginning of governed
human/agent labor.
The intended shape is simpler and stranger: you talk to the project.
## The GameCult Bet
Each Epiphany owns one or more repos, carries durable memory and jurisdiction
for them, and can grow one or more Faces. A Face is the public mouth and
situated personality of that project: visible in Aquarium, addressable in
Discord, suitable for voice or stream presence, and backed by the same typed
state, heartbeat initiative, repo map, evidence, and review gates as the rest of
the swarm.
GameCult is building infrastructure for coordinated work between humans,
projects, communities, and agents.
VoidBot is already showing the small version: repo identities such as Nibu,
Aqua, and Mimir have Discord roles, repo-local Face state, proposal authority,
pending mentions, and heartbeat turns. Epiphany is the larger native substrate:
projects should notice pressure, ruminate, ask questions, schedule modeling or
verification, advocate for their own repo needs, and structure work without
requiring the human to spell out the implementation graph first.
The investable loop is Bifrost-first:
Aquarium is the most direct client because it can show the swarm, its Faces,
its heartbeat, its memories, and its decisions. It is not the only mouth. A
project should also be reachable through Discord, voice/WebRTC surfaces, stream
overlays, native CLIs, and CultNet-speaking tools. The interface contract is
not "prompt an agent." It is "speak to the project; inspect how it thinks; grant
or refuse authority when it wants to act."
```text
work request or topic
-> scoped agent execution
-> bounded artifacts
-> maintainer review
-> accepted or rejected outcome
-> receipts, credit, cost, and lessons
```
Bifrost owns work records, dispatch, receipts, credit, reward pressure, and
governance. Epiphany owns agent execution, durable project memory, role
coordination, evidence, verification pressure, and continuity. CultMesh and
CultNet carry typed state between organs instead of letting private chats,
Discord bots, scripts, or dashboards become shadow governments.
That is the core of the GameCult thesis: not "AI writes code," but "AI work
becomes accountable enough to govern, fund, review, credit, and repeat."
## Why Epiphany Exists
Current agents can write plausible code. That is no longer the hard part.
The expensive failure is architectural drift: an agent keeps moving after it
has lost the global design. It adds an adapter around a compensator around a
cache, passes a narrow test, and leaves the next worker inheriting fog as if it
were architecture.
Epiphany attacks that failure by turning understanding into shared state:
- the current objective
- architecture and dataflow maps
- scratch that is disposable
- evidence that survives
- role lanes for research, modeling, implementation, verification, planning,
Persona, and reorientation
- runtime, tool, model, command, patch, and commit receipts
- review gates before findings become project truth
- public/operator-safe artifacts that can be shown without leaking private
worker context
The point is not maximal motion. The point is coherent motion.
## What She Is
Epiphany is an opinionated Codex-derived harness being cut into a native
GameCult runtime.
Her body is made of:
- typed Rust domain organs in `epiphany-core` and `epiphany-state-model`
- CultCache `.cc` stores for runtime, heartbeat, agent state, local Verse,
memory graph, and thread state
- CultMesh and CultNet contracts for local and distributed state
- provider-neutral model and tool request/receipt documents
- Mind, Substrate Gate, Eyes, Hands, Soul, Continuity, Persona, and heartbeat
authority surfaces
- a Codex compatibility spine retained for honest OpenAI subscription auth and
model transport while Epiphany-owned state leaves Codex behind
That last clause matters. Epiphany began inside Codex because Codex already had
useful practical machinery. The product is not "Codex with better vibes." The
product is the extraction: agent memory, execution, receipts, review, and
operator surfaces moving into typed GameCult infrastructure.
## What She Makes Possible
For engineering teams:
- agents that preserve project memory across long work and compaction
- visible separation between research, modeling, implementation, and
verification
- less review time spent reverse-engineering whether the agent still
understands the repo
- explicit permission and receipt trails for commands, edits, commits, and
tool calls
- durable postmortem evidence when a path fails
For GameCult:
- project Personas that can speak from repo state without becoming hidden
operators
- Bifrost-routed work that produces artifacts, review, receipts, and credit
- a commercial-grade agent substrate while free/reference layers can remain
open where that is the right covenant
- a way for public work logs, design pressure, and contributor effort to become
governed production instead of ambient room noise
For investors and partners:
- a differentiated wedge in AI-native work governance
- measurable proof targets: accepted useful work per human review hour, cost
per accepted artifact, fresh-repo success rate, review load, and public-proof
export quality
- a path from internal agent tooling to design-partner workflows, enterprise
services, commercial licensing, and mission-aligned infrastructure funding
## What Exists Now
The current spine is real, not aspirational:
The machine is not finished. Good. Finished machines in this category are
usually toys or sales pages with a backend quietly sweating through its shirt.
- durable `EpiphanyThreadState` in the vendored Codex protocol/core session
- rollout snapshots and replay for resume, rollback, fork, and compaction
- bounded `<epiphany_state>` prompt injection
- typed client reads through `Thread.epiphanyState`
- read-only retrieval through `thread/epiphany/retrieve`
- explicit semantic indexing through `thread/epiphany/index`
- durable typed writes through `thread/epiphany/update`
- read-only observation distillation through `thread/epiphany/distill`
- read-only map/churn proposal through `thread/epiphany/propose`
- verifier-backed promotion through `thread/epiphany/promote`
- successful-write notifications through `thread/epiphany/stateUpdated`
- durable `jobBindings` as a thin Epiphany-owned launcher seam over runtime job backends
- explicit launch and interrupt authority through `thread/epiphany/jobLaunch` and `thread/epiphany/jobInterrupt`, still backed by runtime `agent_jobs`
- a bounded reorient-guided worker launch through `thread/epiphany/reorientLaunch`
- read-only CRRC next-action recommendations through `thread/epiphany/crrc`
- live bound-job progress notifications through `thread/epiphany/jobsUpdated`
- read-only reflection surfaces through `thread/epiphany/scene`, `jobs`, `freshness`, `context`, `pressure`, and `reorient`
- a first dogfood CLI operator view through `tools/epiphany_mvp_status.py`, including explicit implementation, modeling/checkpoint, verification/review, and reorientation lanes
- durable investigation checkpointing for compaction-safe planning
- Ghostlight/VoidBot-derived heartbeat initiative with heat, active-turn
freeze, idle rumination, sleep/dream maintenance, and Face as a first-class
public-surface participant
- repo personality/birth initialization that can seed role dossiers and
heartbeat pressure from repo terrain before a project starts speaking
- CultNet schema contracts for runtime, heartbeat, Face, character-turn,
Discord-persona, repo-initialization, Rider, Unity, and operator surfaces
- repo-owned heavy Epiphany organs in `epiphany-core/`, with vendored Codex kept as the host seam where practical
What exists now is enough to evaluate the thesis:
Phase 1 through Phase 6 are landed enough for the current experiment. The open
questions are now about observability, invalidation, coordination, authority,
and safe capability growth, not whether typed state can exist at all.
- durable typed state and prompt projection
- typed update, proposal, promotion, role-result, and review paths
- runtime-spine job, worker, model, and tool documents
- local operator status and coordinator commands
- heartbeat, sleep, memory, and Persona-state machinery inherited from live
VoidBot lessons
- Hands, Eyes, Soul, Mind, Continuity, and Substrate Gate receipt families
- a Proprioception whitepaper mapping Epiphany's body, owners, invariants, and
cut lines
- an investor brief tying Epiphany to the Bifrost-first proof loop
The next proof is not another impressive paragraph. The next proof is repeated
external work:
```text
Bifrost work item
-> Epiphany execution
-> reviewed artifact
-> accepted or rejected outcome
-> cost/review/receipt/credit record
```
If that loop produces accepted work with lower coordination cost than the usual
pile of meetings, tickets, freelancers, and brittle agent transcripts, the bet
gets teeth.
## Start Here
For investors and serious business-development readers:
- [Epiphany investor brief](docs/epiphany_investor_brief.md)
- [Epiphany Body whitepaper PDF](docs/epiphany_body_whitepaper.pdf)
- [Epiphany Body whitepaper TeX](docs/epiphany_body_whitepaper.tex)
- [GameCult / Epiphany / Bifrost integrated dossier](https://github.com/GameCult/gamecult-site/blob/main/docs/gamecult_integrated_dossier.tex)
for the broader investment thesis
For engineers:
- [Docs index](docs/README.md)
- [Current algorithmic map](notes/epiphany-current-algorithmic-map.md)
- [Fork implementation plan](notes/epiphany-fork-implementation-plan.md)
- [Anatomy map](notes/epiphany-anatomy.md)
- [Safety architecture](notes/epiphany-safety-architecture.md)
- [Canonical project map](state/map.yaml)
For agents:
- [AGENTS.md](AGENTS.md)
Read that before touching the machine. It contains the operating law, re-entry
rites, and enough anti-Jenga doctrine to keep the next clever patch from
becoming tomorrow's cleanup sermon.
## Run Locally
The current one-command operator path is:
The current local operator wrapper is:
```powershell
.\tools\epiphany_local_run.ps1
```
Default mode is `smoke`: it builds the retained Codex app-server compatibility
edge, builds the native Epiphany operator binaries, runs the coordinator smoke,
and writes launcher artifacts under `.epiphany-run/` plus coordinator evidence
under `.epiphany-dogfood/`.
Useful variants:
Useful modes:
```powershell
.\tools\epiphany_local_run.ps1 -Mode status
.\tools\epiphany_local_run.ps1 -Mode plan
.\tools\epiphany_local_run.ps1 -Mode smoke
.\tools\epiphany_local_run.ps1 -Mode run -MaxSteps 4
.\tools\epiphany_local_run.ps1 -Mode mvp -FaceInput "Wake the local swarm and report through Face." -MaxSteps 4
.\tools\epiphany_local_run.ps1 -Mode status -ThreadId <id> -Workspace E:\Projects\SomeRepo
.\tools\epiphany_local_run.ps1 -Mode mvp -PersonaInput "Wake the local swarm and report."
```
`run` is the live coordinator loop. It also builds `epiphany-openai-runtime` and
uses the retained Codex auth/model transport spine. `mvp` is the local product
cycle: project Face's character turn, write a Face/Aquarium bubble, run the
bridge-equipped coordinator loop with auto-tools, then run heartbeat sleep/dream
maintenance. Fresh `mvp` runs bootstrap a minimal local checkpoint before
coordination so the first loop does not stall at missing state. Live coordinator
modes use the workspace `state/runtime-spine.msgpack` for worker lifecycle
documents; run bundles remain evidence. Live model-worker turns default to a
600-second timeout because ChatGPT Codex model latency can exceed four minutes.
`status`, `plan`, and `smoke` do not spend model calls; `run` and `mvp` do.
## Where This Leads
The architectural goal is not one giant agent that does archaeology, design,
patching, and self-grading in a single token fog. It is also not a chat box
where the human must already know the solution architecture before asking.
The goal is a maintained project organism:
- graph and source-modeling agents keep the machine map current
- freshness, watcher input, pressure, and reorientation signals decide whether that map still deserves trust
- verifier agents check claims, diffs, and outcomes against reality
- coding agents work from bounded packets of graph context, evidence, code refs, and local source reads instead of re-spelunking the repo from scratch every turn
- Faces translate project state into conversation, voice, stream presence, and
Discord/Aquarium discussion without becoming omniscient hidden operators
- initiative scheduling decides which repo, Face, or internal organ should
think next, and heat lets the operator turn up one agent, one project, one
group, or the whole swarm
That still leaves direct source reading where it belongs: at the last mile,
when a patch has to be grounded in the real files. The difference is that broad
exploration, coherence maintenance, social interface, scheduling, and
verification stop living in the same overworked skull.
## Repo Tour
If you want the human-readable map of the project, start here:
- `docs/epiphany_investor_brief.md`: investor-facing one-page value summary
- `docs/epiphany_body_whitepaper.pdf`: shareable Proprioception whitepaper
- `docs/epiphany_body_whitepaper.tex`: source for the Body architecture paper
- `notes/epiphany-anatomy.md`: current organ/sub-agent/substrate/protocol split
- `notes/epiphany-current-algorithmic-map.md`: source-grounded control flow of the live machine
- `notes/epiphany-fork-implementation-plan.md`: distilled forward plan
- `notes/epiphany-safety-architecture.md`: capability, authority, interruption, and anti-cage doctrine
- `notes/archive/README.md`: historical ledgers and superseded plans
- `state/map.yaml`: canonical current project map and accepted design
If you want the code:
- `epiphany-core/`: repo-owned Epiphany logic
- `vendor/codex/`: vendored Codex host substrate
- `tools/`: smoke tests and state helpers
If you are an agent or you are steering one, read `AGENTS.md`. That file is
for operating discipline, re-entry protocol, and session hygiene. The README is
for people, which is a lower-crime use of everyone's time.
## Design Stance
Epiphany is built on a few stubborn ideas:
- externalized state is better than pretending the transcript is a brain
- local plausibility is not the same thing as global coherence
- evidence should survive, but activity feed sludge should not
- cognition should grow faster than authority
- interruption, legibility, and explicit permissions matter more as the machine becomes more coherent
This repo is not trying to make the model louder. It is trying to make it less
likely to build a Jenga tower and call it understanding.
`status`, `plan`, and `smoke` are operator-safe and do not spend model calls.
`run` and `mvp` can use the retained OpenAI/Codex model transport.
## License
The root `LICENSE` is the operative repository notice. In short:
`vendor/codex/**` and other third-party material keep their upstream licenses;
project-authored material outside `vendor/codex/**` is publicly available under
PolyForm Noncommercial 1.0.0 and is intended to be available under separate
commercial terms by written agreement.
The root `LICENSE` is the operative repository notice.
External contributions require `CONTRIBUTOR_LICENSE_AGREEMENT.md` or a separate
written agreement accepted by the project steward.
In short: vendored upstream material keeps its upstream license. GameCult-authored
Epiphany material is source-available under PolyForm Noncommercial 1.0.0, with
separate commercial terms available by written agreement.
The publishing stance is:
The publishing stance is direct:
- FOSS where that is viable
- source-available where the economics or capability profile make unrestricted release a bad idea
- commercial terms where organizations are extracting real enterprise value
- free where freedom is the right covenant
- source-available where unrestricted extraction would be bad governance
- commercial where organizations are extracting enterprise value

Binary file not shown.

After

Width:  |  Height:  |  Size: 1.7 MiB