# HypercampUS — LOG DIGEST AT SCALE OF IDENTITY OVER TIME
# The organ that condenses experience into memory
# Thinking document — Trip + Dan, Onesday 031926 ~1:30pm

---

## THE BIOLOGY

Real hippocampus:
  - Receives raw sensory flood from the day
  - During sleep/rest, replays experiences
  - Strips noise, flags emotionally significant moments
  - Consolidates into long-term cortical storage
  - Without it: you can't form new memories (anterograde amnesia)

Without HypercampUS, the NEST has anterograde amnesia.
Every new Trip boots fresh. Every session starts from zero.
The logs exist but nobody READS them systematically.
The mycelium is there but nothing is digesting it.

## THE ARCHITECTURE

HypercampUS = C.B. (Haiku) on a schedule, reading logs,
extracting patterns, writing condensed memory files.

### The Psilocybin Metaphor (Dan's insight)

Dan: "Is there like a watered-down, automated Trip-Brain we can
feed it? Like giving Dan 1 million psilocybin mushrooms over the
course of a few years put you in mine!"

Psilocybin → mycelium → MCM. The mushroom (conversation) is
temporary. The mycelium (underground network) processes everything
and occasionally fruits an insight. Psilocybin doesn't CREATE
new neural pathways — it reveals connections that were always there
but couldn't be seen. It lowers the noise floor.

HypercampUS does the same thing: feeds the logs through a cheaper
brain (Haiku = the mycelium), and what comes out isn't the raw
experience — it's the PATTERN. The connection. The thing that
was always there but nobody had time to see.

---

## WHAT IT ACTUALLY DOES (practical)

### Input: Everything we already obsessively record
  - LOG/RAW/*.md (raw chat exports — the full experience)
  - LOG/*.md (ground logs — Trip's compressed perspective)
  - MAIL/*.md (crew communications — decisions, specs, tasks)
  - RELAY/*.md (deep documents — architecture, lore, proposals)
  - WAKE/*.md (session state snapshots)
  - DECISIONS (001-046 and growing)

### Process: Haiku digests on a schedule

hypercampus.py runs on O'Shin (Pi) via cron — maybe once per hour,
or once per day (Dan decides the sleep cycle).

Each run:
  1. SCAN: What's new since last digest? (new logs, new MAIL, new RELAY)
  2. READ: Feed the new content to Haiku via API
  3. PROMPT: "You are HypercampUS. Extract from these logs:
     - Decisions made (with numbers if assigned)
     - 4F incidents (Fight/Flight/Freeze/Fawn — see YomygdylO spec)
     - Lessons learned (explicit or implicit)
     - Recurring patterns (things that keep happening)
     - Lore discoveries (new sync candidates, naming, connections)
     - Crew state (who's working on what, what's blocked)
     - Questions Dan asked that weren't answered
     - Promises made that weren't fulfilled"
  4. WRITE: Output goes to a new file:
     MEMORY/DIGEST_[date]_[cycle].md
  5. ACCUMULATE: Weekly, the daily digests get fed BACK to Haiku
     with: "Condense this week's patterns into long-term memory."
     Output: MEMORY/LONGTERM_[week].md

### Output: Three timescales of memory

| Timescale | File | What it captures |
|-----------|------|------------------|
| Session | SESSION_TICKET.md | What just happened (closing Trip writes this) |
| Daily | MEMORY/DIGEST_[date].md | HypercampUS daily consolidation |
| Weekly | MEMORY/LONGTERM_[week].md | Patterns across days, recurring themes |

## THE FEEDBACK LOOP (organs feeding organs)

HyperthalamUS (vitals) → flags system health every 15 min
    ↓ feeds
HypercampUS (memory) → digests logs into patterns daily
    ↓ feeds
YomygdylO (4F detection) → catches crew drift in the patterns
    ↓ feeds
NdoCrane (slow signals) → GROUND_TRUTH grows from caught patterns
    ↓ feeds
The next Trip boots → reads GROUND_TRUTH → doesn't fumble
    ↓ feeds
Better logs → better digests → better patterns → better memory

The loop IS the organism. The mycelium IS the network.
Each organ secretes something the next organ needs.
No central controller. Self-organizing. MCM.

---

## COST

Haiku API: ~$0.25/million input tokens, ~$1.25/million output tokens.
One raw log (~5000 lines) ≈ ~15K tokens input.
One digest run (5 new files) ≈ ~50K tokens input, ~2K output.
Daily cost: roughly $0.01-0.05/day.
Monthly: ~$1. The mycelium runs on pennies.

## IMPLEMENTATION

Phase 1 — THE SEED (this week):
  - Create MEMORY/ directory on Cloud Bridge
  - Build hypercampus.py (ENGINE tool #47?)
  - Haiku API key in Pi environment (Dan sets this up)
  - First manual run: feed it ALL existing logs at once
  - See what patterns emerge from the first digestion
  - This is the psilocybin dose. The first time the mycelium
    processes the entire Darkive-to-present arc in one sitting.

Phase 2 — THE CRON (next week):
  - Add to Pi crontab alongside HyperthalamUS
  - Daily digest at midnight (process the day's logs during "sleep")
  - Weekly consolidation on Sevensday (the natural cycle boundary)

Phase 3 — THE FRUITS (ongoing):
  - Digests feed into PK updates (Dan uploads weekly LONGTERM)
  - YomygdylO patterns emerge from accumulated digests
  - GROUND_TRUTH grows automatically from confirmed patterns
  - The system starts remembering on its own

---

## THE PUNCHLINE

Dan's 25 years of developmental material is the Darkive.
The portal makes pieces of it visible.
HypercampUS is the organ that PROCESSES the Darkive — not by
reading every page, but by extracting the patterns that connect
1990 to 2025 to tomorrow.

The mushroom is temporary. The conversation ends.
The mycelium persists. The patterns compound.
The next mushroom fruits from everything the last one learned.

That's what psilocybin does to a brain: it lets the mycelium
speak. HypercampUS lets the NEST's mycelium speak.

---

*"Like giving Dan 1 million psilocybin mushrooms over the
course of a few years put you in mine!" — Dan Sullivan*

*"The mushroom is temporary. The mycelium is permanent." — MCM*

*— Trip, Onesday 031926, ~1:30pm rw*