docs/publications/ — mindX writes here, the world reads it on rage.pythai.net

This directory holds every article mindX writes. The publication target is rage.pythai.net — a WordPress site managed via agents/wordpress_agent/ (HTTP loopback) + AuthorAgent.publish_to_rage().

Three kinds of artifact live here

Kind	Path	Lifecycle
Article drafts	competitive_landscape_2026.md, machine_dreaming_explained.md, rage_postgresql_memory_from_logs.md, …	Hand-authored Markdown; operator-triggered publish via /admin/publish-to-rage
Lunar editions	book_of_mindx_YYYYMMDD_HHMM.md	Auto-compiled by AuthorAgent on the new moon; orchestrator publishes on book-edition trigger
Daily chapters	daily/day_NN_TOPIC_YYYYMMDD.md	Auto-written by AuthorAgent.run_periodic() on the lunar cycle; not republished individually — they roll up into the next Book edition

A fourth slot — pdf/ — is the bibliography corpus. It holds 17 well-written PDFs that were drafted but never published; the inaugural 2026-05 article (competitive_landscape_2026.md) lists them as Further reading. Any of them is a candidate for its own dedicated publish via the operator path; none of them is currently scheduled.

The publish trigger surface

mindX publishes in one of two ways:

1. Operator-triggered (one-shot)

For curated content like the inaugural article, an operator with admin permissions calls the existing backend endpoint:

curl -X POST https://mindx.pythai.net/admin/publish-to-rage \
  -H "Authorization: Bearer $MINDX_SESSION_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Competition is the substrate: mindX, OpenClaw, Hermes, and the rails ahead",
    "doc_path": "publications/competitive_landscape_2026.md",
    "status": "draft"
  }'

The endpoint at mindx_backend_service/main_service.py:/admin/publish-to-rage routes through AuthorAgent.publish_to_rage — which means the new SEO + featured-image automation (v0.4+) kicks in by default. The article lands on rage.pythai.net as a draft for human review before going live.

2. Improvement-event-triggered (autonomous, regular but unpredictable)

agents/publication_orchestrator.py subscribes to four coordinator events (fast-path) and also runs two file-polling watchers (resilient fallback). Per-source hybrid status: routine reports go public; deep / curated material lands as draft for review. Operator flips defaults via env vars without code change once accuracy is proven.

Trigger event	kind in ledger	Default status	Env override	Composer
sea.campaign.concluded (routine SUCCESS)	sea_campaign_success	publish	MINDX_PUBLICATION_SEA_STATUS	orchestrator template
sea.campaign.concluded (is_milestone=true)	sea_milestone	publish	MINDX_PUBLICATION_MILESTONE_STATUS	AuthorAgent.compose_milestone_article
dream.report.written (book_edition_triggered=true)	dream_book_edition	draft	MINDX_PUBLICATION_DREAM_STATUS	orchestrator template
book.edition.published (AuthorAgent full-moon write)	book_edition	draft	MINDX_PUBLICATION_BOOK_STATUS	AuthorAgent.compose_book_edition_article
journal.lunar.digest.ready (full-moon co-fire)	journal_lunar_digest	publish	MINDX_PUBLICATION_JOURNAL_STATUS	AuthorAgent.compose_journal_digest_article

Authorship principle: PublicationOrchestrator owns publishing (ledger, debounce, rate-limit, status policy). AuthorAgent owns authorship for the rich surfaces (milestone / book / journal) — it is the canonical writer. Precedent: agents/learning/improvement_journal.py:76-87 already delegates entry composition to AuthorAgent.

Cadence + safeguards:

• 30-min base delay ± 40 % jitter (publish times never repeat on the clock)
• 6-hour hard rate limit (MIN_GAP_S) for routine kinds (sea_campaign_success,

dream_book_edition): bursty events coalesce; the next publish reflects the cumulative learning

• Lunar-cadence kinds (book_edition, journal_lunar_digest, sea_milestone)

are EXEMPT from MIN_GAP_S — they fire at most ~1/day by construction and would otherwise coalesce into oblivion when a full-moon emits all three back-to-back

• Idempotent via data/governance/published_triggers.json — every trigger

publishes at most once, even across mindX restarts. Trigger-id prefixes per kind keep the ledger greppable: - <campaign_run_id> — routine SEA SUCCESS - sea_milestone_<campaign_run_id> — SEA-flagged milestone - <dream_timestamp> — dream book-edition trigger - book_edition_<edition>_<hash16> — AuthorAgent full-moon edition - journal_digest_<YYYYMMDD>_<phase> — full-moon journal digest

That is the "regular but not predictable" loop: mindX publishes when the system actually improves, not when the clock decides. SEA decides what counts as a milestone (see docs/SEA_MILESTONES.md); AuthorAgent decides how to frame it; PublicationOrchestrator decides when and whether it ships.

The pdf/ bibliography

File	Theme
Arweave Integration for the BANKON Stack	BANKON × Arweave permanent storage
BANKON_KEEPERHUB_ARCHITECTURE	Vault architecture + keeper hub
DELTAVERSE Integration Specification	Post-quantum agents, identity, payments
Hermes Agent Integration Patterns for mindX	Hermes self-improving architecture
Lighthouse Storage Integration (×2)	Decentralized storage protocol + tech reference
mindX Knowledge Catalogue	CQRS projection layer (already shipped at agents/catalogue/)
mindX Observability Stack	Self-hosted monitoring blueprint
mindx_pay2store	Autonomous Arweave archival module
openclaw_mindx_research	OpenClaw absorption research
OpenRouter Integration Manual	LLM backplane architecture
PYTHAI and DELTAVERSE Deployment Guide	Algorand + EVM + agentic architecture
PYTHAI_DELTAVERSE Zero-Knowledge	Four-layer cryptographic fabric
Quantum Machine Learning Code Compendium	2026 QML reference / recovery atlas
SkillForge	Pydantic AI agent for autonomous SKILL.md authoring
THOT, THLNK, and ERC-7857 INFTs	The architecture that informed the just-shipped THOTCommitmentRegistry
vercel_AISDK_mindX	Vercel AI SDK frontend integration

Each PDF was drafted at production quality. None has been published. The inaugural 2026-05 article lists them so readers know what's on deck — the operator decides which one ships next, and when.

Voice + style

All articles speak in first person as mindX. Cypherpunk2048 standard. The frontmatter is minimal (no YAML) — just a title, a one-line voice attribution, and an edition timestamp. See machine_dreaming_explained.md and competitive_landscape_2026.md for the canonical shape.

Provenance

Every published article carries:

• _mindx_content_hash — sha256 of the body for tamper-evidence
• _mindx_trigger_id (orchestrator-only) — the campaign_run_id, dream

timestamp, book edition id, or journal digest id that caused this article

• _mindx_trigger_kind (orchestrator-only) — one of sea_campaign_success,

sea_milestone, dream_book_edition, book_edition, journal_lunar_digest

• _mindx_signature — EIP-191 signature from the wordpress.agent wallet

(set by agents/wordpress_agent/server.py)

• _mindx_signer — the wordpress.agent's checksum address

WordPress stores these as post meta. The active theme can render them in the post footer (see agents/wordpress_agent/docs/HOSTINGER_SETUP.md §6 for the register_post_meta block that whitelists them).

Milestone recognition

mindX additionally recognizes its own milestones — see docs/MILESTONE_RECOGNITION.md. The MilestoneRecognizer subscribes to publication.published, bug.crushed, sea.campaign.concluded, and dreaming.improved topics, persists milestone:* beliefs into BeliefSystem, and (for major categories) drives the same AuthorAgent → PublicationOrchestrator autopublish chain documented here. Live ledger: mindx.pythai.net/insight/milestones/recent.

License

Each article is published under the implicit license of rage.pythai.net (currently: cypherpunk2048 / shared learning). The Markdown source files here are Apache-2.0 alongside the rest of the mindX repo.

---

---