The Collective
Master Library

The numbers, unrounded

Straight from manifest.json. Delivered LOC is deduplicated — the aggregate closure figure is larger and would have been the flattering number to print.

Lead exhibit · PRIM-ENG-01

DREAM

It has been in this Canon since November 2025, sealed and dated — which matters only because the industry arrived at the same idea later and independently. That is not a claim of influence. It is a claim of timestamp, and the seal chain is what makes it checkable rather than a story. Volume I would demote it otherwise.

42,418 LOC · requires the governed runtime slice RTS-CORE · one adapter port (PORT-PERSISTENCE) you implement · never recommend it alone.

What else is in there

• DREAM

  Deterministic, no-LLM synthesis engine that runs on idle cycles, mines signals below the confidence floor everyone else discards, and clusters them into novel patterns.

  PRIM-ENG-01 · The crown jewel. Shipped in this Canon since November 2025.

  42,418 LOC

• Creative Forge

  Full creative stack — imagination, innovation and dream engines fused for generative synthesis, pattern evolution, and nocturnal optimization.

  MENGINE-CREATIVE-FORGE · Meta-engine. Requires the governed runtime slices it ships with.

  27,188 LOC

• World-First Cognitive

  BRAIN attention and memory, DECODE intent understanding, and DREAM creative evolution unified into one cognitive layer.

  MENGINE-WORLD-FIRST-COGNITIVE · Meta-engine.

  27,188 LOC

• Creative Evolution Engine

  Genetic-algorithm knowledge mutation, insight crystallization into permanent memory, emergent pattern evolution, and dream-state analysis.

  SENGINE-CREATIVE-EVOLUTION-ENGINE · Substrate engine.

  28,206 LOC

• Temporal Engine

  Temporal causal chains, chronological event synthesis, time-aware pattern detection, lifecycle state prediction, and temporal memory indexing.

  SENGINE-TEMPORAL-ENGINE · Substrate engine.

  28,206 LOC

• Dream Synthesizer

  Offline memory consolidation — decay, prune, strengthen, merge. Turns raw experience traces into compacted long-term memory.

  BLD-ENG-260 · The DREAM primitive's heart, extractable on its own with no dependencies.

  94 LOC

Before you spend anything

Hand it to your agent

Two files do the work. AGENT-CATALOG.md is the machine-readable index of all 1,062 artifacts, written for a coding agent with your repository open. The prompt below tells that agent to run a substitution pass against your actual code and report what would be replaced, what would be hardened, and — critically — what should be skipped.

If the answer comes back "not much," you saved your money. That is the intended outcome for plenty of codebases, and the prompt is written so an agent can say it.

Read the attached `AGENT-CATALOG.md`. It is the index for a 1,062-artifact
component library — engines, primitives, components, blueprints and runtime
slices extracted from twenty-nine shipped projects.

Then read this repository.

Do a substitution pass. For every unit of our own code that does a job this
catalog also does, classify it:

- REPLACE — the catalog component covers our unit's job and more.
- HARDEN  — our unit works, but the component adds correctness or failure
            handling we lack (determinism, tamper evidence, backpressure,
            circuit breaking, budget accounting).
- SKIP    — our unit is more specific, or the component needs adapter ports we
            cannot satisfy.

Match by capability, not by name. Same job counts; same identifiers do not.

For each finding, report in this shape:

[REPLACE] CMPSBL <ID> "<Name>"  ->  host: <our file path> (<LOC>)
  Why: <what it covers that ours does not>
  Saves: <N> engineer-days, basis = <delivered LOC> delivered LOC +
       <k> correctness concerns you would otherwise implement and test
  Ports to implement: <none | PORT-*>
  Requires runtime slices: <none | slice ids>

Rules you must follow:

1. Never invent a component that is not in the catalog. If you are unsure
   whether something exists, say so rather than guessing.
2. Any component whose entry says it requires a governed runtime must be
   recommended together with its runtime slices — never alone.
3. Every adapter port is real work we have to schedule. Name each one. The
   portable defaults compile but do nothing.
4. Restate each component's verification status, including the entries that
   read "Not claimed". Do not upgrade "operationally exercised" into "secure"
   or "production-proven for our use case".
5. Do not inflate savings. State the basis in the same sentence as the number.
6. Produce a SKIP list. An evaluation with no skips did not evaluate anything.

Then answer three questions:

1. How many of the 1,062 artifacts are actually relevant to this repository?
   Give a number smaller than the catalog size, and justify it.
2. Which three would change the most, and what specifically would improve?
3. Given what you found, is this worth buying for us — and at which tier? If
   the answer is no, say no and explain what would have to be true for it to
   be yes.

Do not modify this repository. Report and stop.

Read the attached `AGENT-CATALOG.md` and this repository.

We have a specific problem: [describe it — e.g. our webhook delivery
double-sends under retry / we have no audit trail we can prove wasn't
tampered with / our AI provider costs are unpredictable and we're locked to
one vendor].

Which components in this catalog address that problem, and what would the
integration actually involve in our codebase — which files change, which
adapter ports we would have to implement, and what we would still have to
build ourselves?

Never invent a component that is not in the catalog. If nothing in the catalog
addresses this well, say so plainly.

What it will not tell you

That is a real claim and a bounded one — Volume I's rule applied to the author's own storefront. Every artifact carries six verification axes, and unproven axes read Not claimed rather than being quietly left blank. If an agent tells you a component is "secure" or "production-proven for your use case," ask it which verification field it based that on.

LICENSE · CMPSBL Source Available License v1.1 (CSAL-1.1) — Embed components in your own software and sell that software commercially, royalty-free, while the components remain supporting parts and never the primary product — with attribution to Kenneth E. Sweet Jr. in your documentation. Standalone resale of a component, or a product whose principal function is a component's function, needs written permission. That is a different license from the Free Edition Grant covering the volumes on this site.