# perardua.dev — Jeremy McEntire

> Engineer, researcher, and author. More than twenty years building and leading engineering teams, from Intel to startups to independent research. 5 US patent filings, granted and pending. 28 active research papers across AI and organizational theory. 5 books.

## About

Jeremy McEntire is a structural thinker who applies systems reasoning to AI and organizational theory. His research examines exploitable structure in AI training trajectories, agent governance architectures, and the information-theoretic foundations of organizational behavior.

Career highlights span Mashery and Intel (through the acquisition, to SWE VII), Twilio (built the Edge Services organization), Appreciate (Chief Product & Technology Officer), and independent research. He is a co-inventor on three granted US patents (Twilio) covering edge communication locations and API request throttling, with two pending provisional filings on the Stigmergic Coordination Protocol and decentralized multi-agent coordination.

Primary contact: jmc@perardua.dev
Website: https://perardua.dev
GitHub: https://github.com/jmcentire
LinkedIn: https://www.linkedin.com/in/jandrewmcentire/
Publisher: https://cageandmirror.com

## Research — Structural Compression Theory (SCT)

A unified information-theoretic account of how structure compresses — in neural networks, in organizations, and in epistemic systems. Published as a monograph on SSRN and developed across 28 active papers in four domains.

### Activation Geometry (13 papers)

Exploitable structure in the forward pass of transformer models. Consolidated papers cover training trajectory structure (regime detection, cross-seed convergence, speculative weight prediction), constellation-indexed model composition, domain-invariant structural signatures, topological detection of model distillation, RLHF social cognition damage, the geometry of failure (terminal measurement, concentration barriers), causal basis discovery, the inter-instance compression barrier, the dissociation of geometry and function, context management (chain degradation saturates logarithmically; meta-commentary priming is harmful), universal entanglement, the Entanglement Theorem, and entanglement under fine-tuning.

### Bridge Theory (1 consolidated paper)

The Cage and the Mirror: consolidated bridge theory combining the Generative Lossy Channel, The Cage, Dysmemic Pressure, and The Mirror, integrating key results on stigmergic measurement, strategic rate-distortion-perception, and structural immunity.

### Organizational Theory (5 papers)

Substrate-independent dysfunction in multi-agent AI. Stigmergic coordination as ambient structure discovery. Emergence as a programming paradigm for constraint-shaped agency. Contraction dynamics and Gaussian fixed-point convergence. The strategic rate-distortion-perception tradeoff.

### Applied and Market Analysis (9 papers)

Forced ranking, structural immunity, inversion of doubt, topology of influence, drone-swarm coordination, entropy-bounded time-series decomposition, welfare-optimal matching, conjunctive hiring and the dilution of exceptional-skill selection ("No Concerns"), and a replication and sensitivity analysis of "The AI Layoff Trap" ("Three Headlines from One Equation").

## Books (5)

- The Cage and the Mirror (Cage & Mirror Publishing, 2026)
- The Key and the Current (Cage & Mirror Publishing, 2026) — coming soon
- Introduction to Applied Synthesis (Cage & Mirror Publishing, 2026)
- Beyond Code (Packt Publishing, 2026)
- Privacy: The Architecture of Forgetting (Cage & Mirror Publishing, 2026)

## Essays (15)

The Drunkard's Walk, The Tragedy of Process, Gödel's Children, The Fiduciary Trap, The Legibility Trap, The Scale Trap, The Context Trap, The Cage, The Mirror, The Cage Yields, The Mirror Breaks, The Geometry of Talent, The Last Website, One Algorithm, Many Skies, Why We Can't Have Nice Things.

## Articles (21)

- No Concerns (LinkedIn, 2026-10-04): https://www.linkedin.com/pulse/concerns-jeremy-mcentire-kqhic
- Lead by Need (LinkedIn, 2026-09-27): https://www.linkedin.com/pulse/lead-need-jeremy-mcentire-2zrxc
- A Bright Idea (and where to put it) (LinkedIn, 2026-09-27): https://www.linkedin.com/pulse/bright-idea-where-put-jeremy-mcentire-t4ugc
- It's in the Pudding (LinkedIn, 2026-09-13): https://www.linkedin.com/pulse/its-pudding-jeremy-mcentire-fjkhc
- Discovering Excellence (LinkedIn, 2026-09-12): https://www.linkedin.com/pulse/discovering-excellence-jeremy-mcentire-zvdqc
- Bring Your Own Bot: Support (LinkedIn, 2026-09-06): https://www.linkedin.com/pulse/bring-your-own-bot-support-jeremy-mcentire-dsmhc
- Be Careful What You Wish For (LinkedIn, 2026-09-06): https://www.linkedin.com/pulse/careful-what-you-wish-jeremy-mcentire-9rfdc
- The Factory runs on BITs (Medium, 2026-09-04): https://medium.com/@j.andrew.mcentire/the-factory-runs-on-bits-6998978bbf35
- Regulate the Harm, not the Tool (LinkedIn, 2026-07-21): https://www.linkedin.com/pulse/regulate-harm-tool-jeremy-mcentire-lf4xc
- K3 is Amazing (If Software Engineering Is Code Generation) (LinkedIn, 2026-07-20): https://www.linkedin.com/pulse/k3-amazing-software-engineering-code-generation-jeremy-mcentire-nnu1c
- Contracts Before Code (LinkedIn, 2026-06-12): https://www.linkedin.com/pulse/contracts-before-code-jeremy-mcentire-dtzxc
- The Delegation Problem (Medium, 2026-06-11): https://medium.com/@j.andrew.mcentire/the-delegation-problem-9e4205dbaace
- Why We Can't Have Nice Things (Medium, 2026-05-20): https://medium.com/@j.andrew.mcentire/why-we-cant-have-nice-things-9667c3c36a70
- One Algorithm, Many Skies: The Application Space of Decentralized Swarm Coordination (Medium, 2026-05-01): https://medium.com/@j.andrew.mcentire/one-algorithm-many-skies-the-application-space-of-decentralized-swarm-coordination-757eb4978892
- Operators, Users, and Little Bobby Tables (Medium, 2026-04-27): https://medium.com/@j.andrew.mcentire/operators-users-and-little-bobby-tables-444a61152519
- Why the Wage Stagnation Chart Is Right and Wrong at the Same Time (Medium, 2026-04-26): https://medium.com/@j.andrew.mcentire/why-the-wage-stagnation-chart-is-right-and-wrong-at-the-same-time-0cf5b100a815
- The Tool That Changed (Medium, 2026-04-26): https://medium.com/@j.andrew.mcentire/the-tool-that-changed-22c9efd863aa
- Communal Wealth Theory (Medium, 2026-04-23): https://medium.com/@j.andrew.mcentire/communal-wealth-theory-d9173e6e7eb2
- I Have a Hunch (Medium, 2026-04-08): https://medium.com/@j.andrew.mcentire/i-have-a-hunch-49e6a503cf51
- The Geometry of Talent (Medium, 2026-04-01): https://medium.com/@j.andrew.mcentire/the-geometry-of-talent-e52a37479923
- Your Agentic AI Is Recreating the Meetings It Was Supposed to Replace (Solutions Review, 2026-03-02): https://solutionsreview.com/your-agentic-ai-is-recreating-the-meetings-it-was-supposed-to-replace/

## Open Source (30 projects)

### Reference Implementation
- Exemplar — governed code review service and reference implementation for the entire governance stack.

### AI & Agents
- Signet — personal sovereign agent stack (~62K lines Rust). https://signet.tools
- Signet Eval — deterministic policy enforcement for AI agent tool calls.
- Agent-Safe — SPL capability tokens; six SDKs; ~2us eval.
- Pact — contract-first engineering; plan-only by default, with explicit implementation handoff and MCP tools.
- Apprentice — adaptive model distillation; routes traffic between frontier API and local model.
- Constrain — interviews engineers and produces induced-understanding prompts and structured constraint sets.
- Transmogrifier — register-aware prompt translation.
- Webprobe — generic site state-graph auditor.
- Advocate — six-persona adversarial review engine.
- Herald — webhook relay and message queue for AI agents and local services.
- Simulacrum — cognitive simulacrum with two-phase classifier dispatch between recall and adversarial-dialog branches.

### Stack Governance
- Baton — cloud-agnostic circuit orchestration.
- Arbiter — access auditing, consistency analysis, blast-radius classification, and trust enforcement.
- Sentinel — production attribution and contract tightening.
- Ledger — schema registry and data obligation manager.
- Cartographer — stack adoption and compatibility scanner.
- Stack Smoke — cross-stack smoke test harness.
- Aegis — resource-budget primitive.
- Covenant — contract validation runtime.
- Scram — emergency termination service.
- Vigil — sub-threshold anomaly detection.
- Witness — human-in-the-loop coordination primitive.

### Org & Management
- Chronicler — event collection and story assembly.
- Stigmergy — organizational awareness via ART-based signal processing.
- Ascend — AI-powered engineering management CLI.
- Kindex — persistent knowledge graph for AI workflows. https://kindex.tools

### Infrastructure & Privacy
- Tessera — self-validating executable document format.
- HermesP2P — decentralized ephemeral P2P messaging. https://hp2p.net
- BlindDB — client-side encryption architecture demo.

Current project homes:
- Exemplar: https://github.com/wandercom/exemplar (maintained by Wander)
- Signet: https://github.com/jmcentire/signet
- Signet-Eval: https://github.com/wandercom/signet-eval (maintained by Wander)
- Agent-Safe: https://github.com/jmcentire/agent-safe
- Pact: https://github.com/wandercom/pact (maintained by Wander)
- Apprentice: https://github.com/wandercom/apprentice (maintained by Wander)
- Constrain: https://github.com/wandercom/constrain (maintained by Wander)
- Transmogrifier: https://github.com/wandercom/transmogrifier (maintained by Wander)
- Webprobe: https://github.com/wandercom/webprobe (maintained by Wander)
- Advocate: https://github.com/wandercom/advocate (maintained by Wander)
- Herald: https://github.com/jmcentire/herald
- Simulacrum: https://simulacrum.tools
- Baton: https://github.com/wandercom/baton (maintained by Wander)
- Arbiter: https://github.com/wandercom/arbiter (maintained by Wander)
- Sentinel: https://github.com/wandercom/sentinel (maintained by Wander)
- Ledger: https://github.com/wandercom/ledger (maintained by Wander)
- Cartographer: https://github.com/wandercom/cartographer (maintained by Wander)
- Stack Smoke: https://github.com/wandercom/stack-smoke (maintained by Wander)
- Aegis: https://github.com/wandercom/aegis (maintained by Wander)
- Covenant: https://github.com/wandercom/covenant (maintained by Wander)
- Scram: https://github.com/wandercom/scram (maintained by Wander)
- Vigil: https://github.com/wandercom/vigil (maintained by Wander)
- Witness: https://github.com/wandercom/witness (maintained by Wander)
- Chronicler: https://github.com/wandercom/chronicler (maintained by Wander)
- Stigmergy: https://github.com/jmcentire/stigmergy
- Ascend: https://github.com/jmcentire/Ascend
- Kindex: https://github.com/wandercom/kindex (maintained by Wander)
- Tessera: https://github.com/wandercom/tessera (maintained by Wander)
- HermesP2P: https://github.com/jmcentire/HermesP2P
- BlindDB: https://github.com/jmcentire/BlindDB

## Patents (5)

- US 12,041,145 — Edge Communication Locations (Twilio, 2024)
- US 11,720,396 — API Request Throttling (Twilio, 2023)
- US 11,601,519 — Edge Communication Locations (Twilio, 2023)
- US Provisional 63/981,369 — Stigmergic Coordination Protocol (2026, pending)
- US Provisional 64/054,727 — Decentralized Multi-Agent Coordination (2026, pending)

## Claude Code Slash Commands

Jeremy authors reusable slash commands for Claude Code. These live in project repos and are listed in the agent-skills index at https://perardua.dev/.well-known/agent-skills/index.json.

- /project:craft-task (Pact) — interactive interview that generates task.md, sops.md, and pact.yaml for Pact's contract-first pipeline. Research-backed: verb-led specs, explicit constraints, boundary naming, failure modes.

## Machine-readable endpoints

- /robots.txt — bot policy + content signals
- /sitemap-index.xml — full sitemap
- /llms.txt — short LLM summary
- /llms-full.txt — this file
- /articles.json — external articles as JSON
- /.well-known/capabilities.yaml — YAML capabilities manifest
- /.well-known/agent-skills/index.json — Claude Code slash commands
- /specs/capabilities — capabilities specification page