LIVE v1.0.0 acceptance testing → test.agenticmem.co | 📊 Reproducible results for peer review → testing methodology & outcomes
ai-memory · v1.0.0 · Engineering Methodology

Adversarially engineered.
Not vibe-coded.

ai-memory is the endpoint memory substrate for AI agent clusters, swarms, and hives. It is built to a single standard: that a Fortune 500 or a Federal, State, County, or Municipal government would stake mission-critical processes on it. This is how that standard is enforced — the AI-agent graph-engineering pipeline, the adversarial review, and the testing gauntlet that stand between a proposed change and the release branch.

Data integrity
Security
Performance
Reliability
Aggregate North Star · the entire epic
Endpoint memory for AI agent clusters, swarms & hives
One bar governs every gate, every issue, every test.
Adoption bar: a Fortune 500, Federal, State, County, or Municipal government must be willing to stake mission-critical processes on ai-memory v1.0.0 as a reliable endpoint memory substrate. Rock-solid on all four pillars — no exceptions, across every gate, issue, and test.

Data integrity

No silent loss, no partial or unsigned commit, no permanently-broken row.

Security

Operator permission rules never bypassed; no fail-open, no privilege or tenant leak.

Performance

Write-plane liveness under contention; bounded, non-DoS-able paths.

Reliability

Cross-backend parity, recoverability, no fleet kill-switch, no silent-wrong-result.

Binding GA mandate: v1.0.0 ships a 100%-certified AI-NHI enterprise-federation configuration — PostgreSQL + Apache AGE + pgvector, encrypted in transit, encrypted attestation, Batman mode. At mission-critical government/enterprise stakes, tolerance for any integrity, security, performance, or reliability defect is effectively zero — the default triage posture across the epic is FREEZE unless a defect is strictly fail-closed with the North Star fully intact.
§01
The development organism

AI-agent graph engineering

ai-memory is built by a structured organization of AI agents with an enforced separation of powers. The agent that writes code is never the agent that reviews it, and never the agent that merges it. No single agent can push an unreviewed change to the release branch.

Conductor · GatekeeperThe orchestrator writes no production code

A single conductor decomposes the epic, dispatches work, performs security & code review, and is the only actor permitted to merge. It holds the release protocol: signed merges, branch-protection lifts that are always re-armed, and the verify-at-tip discipline. Its job is judgment, not authorship.

Coders · Isolated worktreesTop-tier reasoning models, one problem each

Implementation is dispatched to strong coding agents, each in its own isolated git worktree so parallel work never collides. Difficult, security-critical, cross-cutting problems go to the top tier; mechanical work to a lighter tier. Every coder loads the Rust engineering standard and the code graph before writing a line.

Test swarms · Verification onlyCheap agents, massive breadth

A swarm of lightweight agents (GLM-5.3-Flash via OpenRouter) is a pure verification workload — never coding. Because they are inexpensive, they scale to dozens or hundreds of concurrent non-human identities (NHIs) to stress agent-to-agent flows, clusters, and hives against the real running system.

Separation of powersAuthor ≠ reviewer ≠ merger

This is the core anti-vibe-code control. A change is authored by one agent, adversarially reviewed by others, and admitted by a gatekeeper that re-runs the gates itself. Convenient self-approval — the failure mode of most AI-assisted codebases — is structurally impossible.

§02
Navigation as ground truth

CodeGraph

Every coding agent reasons over a real knowledge graph of the codebase before it touches a file — not a guess, not a grep, a graph.

CodeGraph is a queryable graph of the codebase's symbols, edges, and files. One call returns the verbatim, line-numbered source of the relevant symbols plus the call paths between them — including dynamic-dispatch hops that a text search can never follow — and a blast-radius summary of what a change touches.

It is pinned to the live release-development index, mechanically enforced, so an agent can never reason against a stale or wrong-branch view of the code.

Why it matters for the pillars

  • Reliability: a change's blast radius is known before it is written, not discovered in production.
  • Security: every caller of a sensitive funnel is enumerable, so an authz gate can be proven to cover all paths.
  • Integrity: cross-backend twins (SQLite / PostgreSQL) are located together, so parity gaps surface at design time.
§03
A written standard, not taste

The Rust 1.98 engineering standard

Every coding agent works to a prioritized, versioned rulebook — hundreds of rules with stable IDs, each cited by number in review. Correctness and safety are gated before anything else.

P0 — safety & correctnessGated before any other concern

  • Ownership & aliasing — single owner; many &T XOR one &mut T; never return a dangling reference.
  • Errors as values — recoverable failure returns Result; no unwrap in production; make illegal states unrepresentable and fail closed.
  • Concurrency — global lock order; publish with Release/Acquire; never hold a blocking guard across an .await.
  • Unsafe — every # Safety contract documented; no UB reachable from safe callers.
  • Numeric — no implicit overflow; TryFrom for narrowing; never float-compare for logic.

The disciplineRules are cited, not vibed

A reviewer writes "per ERRORS-09" or "flag OWNERSHIP-10." The standard also encodes the Edition-2024 migration hazards that silently change drop and lock behaviour — the traps that quietly corrupt a less rigorous codebase.

The result: the fail-closed, make-illegal-states-unrepresentable posture that the North Star demands is not a hope. It is a checklist an agent must satisfy and a reviewer verifies by rule number.

§04
Decisions are earned, not asserted

The 3×7 adversarial voting scheme

When a judgment call matters — does this defect block the release, or can it wait? — it is not made by one opinion. It is put to a panel of 21 independent jurors: 7 distinct adversarial lenses, 3 rounds each, every one judging on the merits under the North Star.

LENS 01

Data-integrity hawk

Hunts silent loss, partial or unsigned commits, permanently-broken rows.

LENS 02

Certified-federation guardian

Protects the certified pg + AGE + pgvector attestation chain.

LENS 03

Security / authz adversary

Fail-open gates, privilege bypass, "rules never bypassed by AI."

LENS 04

Cross-backend parity auditor

SQLite vs PostgreSQL divergence and silent-wrong-result.

LENS 05

Availability / DoS engineer

Write-plane liveness, fleet kill-switches, recoverability.

LENS 06

Scope disciplinarian

Argues ruthlessly for defer — resists scope creep, concedes only to the mandate.

LENS 07

Operator / upgrade-safety

Real blast radius on deployed NHIs, migration & fresh-install safety.

Worked example — a live 6-issue docket21 jurors · verdict + confidence + rationale · aggregated by a panel chair
#3191Coordination-plane atomicity — unsigned resolve, action TOCTOUFREEZE20–0
#2555schema_version unconstrained fleet kill-switchFREEZE19–1
#3111Capability token flips an explicit operator Deny → AllowFREEZE20–0
#3187Backend-confined metadata validation gapDEFER7–13
#3125Ungoverned-namespace posture — ship as opt-in knobDEFER0–20
#3305Fail-closed read-parity nit, North Star intactDEFER0–20

An issue's own author calling it "deferrable" does not settle it — the panel overruled self-assessments in both directions. A defect that lands on the certified-federation config freezes the release even when it looked minor; a defect that is strictly fail-closed with the North Star intact defers even when it was labeled a blocker.

§05
Cheap → exhaustive, nothing skipped

The testing gauntlet

A change earns the release branch by surviving a cascade that widens at every step — from a source-scanning invariant check, to real dual-OS certified database tiers, to a cloud swarm of hundreds of agents on the running system.

L1

Continuous integration

Self-hosted CI on every push: compile, clippy at pedantic denial, format, and per-module coverage floors that are never lowered — a floor is a ratchet, not a suggestion.

clippy -D pedanticcoverage floorsfmt
L2

Structural invariant suites

Source-scanning gates that fail the build on architectural drift: every write-path SQL is record-stop-gated or reviewed-allowlisted; the append-only spine is guarded; module-size and error-funnel ceilings are pinned; MCP parameter names and attestation strings are invariant.

record-stop gateappend-only spinefunnel ceilings
L3

Native dual-OS certified tier

The real enterprise data tier — PostgreSQL + Apache AGE + pgvector — exercised natively on both Linux and macOS hosts. Cross-backend parity between SQLite and PostgreSQL is asserted test-by-test, so the two backends can never silently diverge.

postgres + AGE + pgvectorLinux + macOSsqlite/pg parity
L4

AI-NHI acceptance matrix

Five deployment configurations — from a single SQLite node to encrypted certified federation — each run across five architecture tiers (single agent → many agents → multi-node swarm → data-center swarm → global hive), plus full agent-to-agent multi-NHI testing against the live daemon.

5 configs × T1–T5A2A multi-NHIat-rest machine-proof
L5

DigitalOcean final-final

The whole matrix re-run on cloud infrastructure, with a GLM-5.3-Flash agent swarm scaled to dozens–hundreds of concurrent NHIs to stress the data-center and global-hive tiers. Nothing ships until the entire cloud matrix is green.

cloud multi-archswarm at scalefinal gate
§06
Trust nothing; prove everything

Verify-at-tip & the signed chain

The last mile — admitting a verified change to the release branch — is itself a hardened protocol, because "the tests passed on the branch" is not the same as "the fix is present and correct at the tip."

The verify-at-tip ruleMerge ancestry is not trusted

A fix that "merged" once was later silently dropped in a rebase. So no issue closes on the claim that a PR merged. The gate is stricter: the fix must be present in the tip's code and demonstrated by a live regression test run against the merged tree — the full structural suite, the affected backend tests, and pedantic lint, re-run at the tip before the merge is admitted.

The signed spineEvery commit & record attested

Commits are cryptographically signed. Merges use a controlled admin-lift that is always re-armed and confirmed — branch protection can never be left open. At runtime, the product writes to an append-only, signed event chain; the certified tier machine-verifies its at-rest encryption and signs that proof into the chain. Attestation is proven, not asserted.

01

Issue

Triaged against the four pillars.

02

Graph + standard

CodeGraph + Rust rulebook loaded.

03

Implement

Isolated worktree, one problem.

04

Adversarial review

Independent lenses; vote if contested.

05

Verify-at-tip

Live regression at the merged tree.

06

Signed merge

Gatekeeper only; protection re-armed.

Most open-source AI code is written to look right. This is written to be proven right.

The difference is not effort — it is structure. Separation of powers so no agent grades its own work. A written standard so correctness is cited, not felt. An adversarial panel so hard calls are earned. A cascade of tests that ends on real certified databases and a cloud swarm of hundreds of agents. And a merge gate that trusts the tip's code, never the story of how it got there.

That is the machinery behind a memory substrate an enterprise or a government can stake a mission-critical process on.

§07
What the process buys you

The guarantees, tied to the pillars

Integrity, proven

Append-only signed chain; fail-closed writes; cross-backend parity asserted per test. No silent loss.

Security, gated

Operator Deny is terminal; authz funnels proven to cover all callers; pedantic lint denies the footguns.

Performance, bounded

Recursive and bulk paths are budgeted and non-DoS-able; write-plane liveness tested under contention.

Reliability, machine-checked

No fleet kill-switch, recoverable state, at-rest encryption machine-verified and attested on the certified tier.

This is the standard.

ai-memory v1.0.0 is not a demo and not a vibe. It is an endpoint memory substrate engineered — pillar by pillar, gate by gate, vote by vote — to be trusted with the work that cannot fail.

◆ Data integrity  ·  ● Security  ·  ▲ Performance  ·  ■ Reliability
ai-memory · v1.0.0 GA engineering methodology · AI-agent graph engineering · adversarial review · certified enterprise federation
ai-memory™ · endpoint memory for AI agents · Apache 2.0 · github.com/alphaonedev/ai-memory-mcp An open-source project of AlphaOne LLC · © 2026