Tyler Martin

Founder · Product Engineer · Researcher · Inventor

Tyler Martin

Pro-human. Pro-AI. Anti-extraction.

Madrid / San Francisco / San Juan

I build AI-native products and infrastructure that compound human creativity and judgment: agents, content protection, secure document review, and developer tools with accountable human-agent collaboration built in. Founded Lobster Labs, an agent-native software studio. Inventor with twelve USPTO filings.

My thesis: AI should compound human capital, not extract from it. That means provenance, permission, royalty rails, and reviewable agent systems — products that move value back toward creators, professionals, and independent builders.

The strategy is to connect dots across domains, then ship tools that fix broken incentives and open new markets.

Approach Human experience and agency as drivers. AI as leverage.
Reading Craig Alanson’s Expeditionary Force, The Expanse series, Suleyman and Mostaque. I mine science fiction and AI operators for high-fit, high-probability, high-commercial-value invention context.
Availability Building with good people

Shipped

Redline Live

Share confidential documents without losing the keys.

AI-native document review via time-limited access tokens. Share confidential docs with AI reviewers without permanent exposure.

diskspace v0.9.0

Find the dead weight in your cargo hold.

Personalized, reversible disk-cleanup CLI for macOS. 85+ declarative YAML rules. Drivable by humans and agents. Signed, notarized universal binary.

Provisionally Live

Non-confidential fit briefs for protected inventions.

Structured disclosure format that surfaces the market relevance of a patent-pending invention without revealing confidential details — designed for early conversations with partners, customers, and investors.

FounderMode Studio Live

Tools for high-agency founders.

Teleprompter purpose-built for 60-second pitch videos. Practice, refine, and ship the pitch that opens doors — no camera crew required.

Research

The CMI Trap

Compound Statutory Liability for Inference-Time AI Retrieval

Tyler Martin, Nicholas Vincent

Reviewed by Prof. Xiyin Tang (Stanford Law)

Inference-time piracy is not training. Different legal exposure, different liability architecture. DMCA §1201/§1202 compound liability for AI systems that strip or ignore content management information at retrieval time.

SSRNStanford LawUC Davis

Also Directed-exhaust p-11B fusion propulsion — a phase-space argument on collimating nonthermal fusion products

Problem Domains Under Filing

10 Applications Filed
2 At Nonprovisional
Public Problem Statements

Each filing answers one of these problems. The public version names the problem space and accountable outcome, not the protected mechanism.

#1 Provenance Creators need a way to prove downstream AI use without exposing the whole workflow. Filed
#2 Metering Publishers need automated web use to stay countable before it disappears into invisible context. Filed
#3 Attribution Rights holders need value attribution when AI systems transform work through compute. Filed
#4 Rights Signals Web content needs enforceable protection when bots learn to optimize around visible rights signals. Filed
#5 Agent Authority Autonomous agents need bounded domains of authority so action stays inside delegated intent. Filed
#6 Accountability Gates High-risk AI actions need accountable routing before private decisions become public harm. Filed
#7 Adaptive Defense Layered web defenses need to adapt as extraction pressure escalates across surfaces. Filed
#8 Deferral Logic Agents need to know when to act independently and when to defer to a person or policy. Filed
#9 Portable Capability Agent capabilities need to stay portable, sealed, and licensable across runtimes. Filed
#10 Auditability Generated documents and figures need deterministic, auditable production paths. Filed

From Assistant Workflow to Agent Runtime

Built on existing model interfaces and third-party tools, this thread shows the operating layer I designed around them: workflow routing, review boundaries, memory, and human approval.

Dr. Forth

2024 – 2025

A Claude-based personal operating workflow for inbox triage, scheduling, and daily planning. The experiment was not "a better chatbot"; it was learning which parts of a founder's day could be drafted, queued, checked, and handed back for decision.

Claude Workflow Ops Prototype Review Loop
workflow → runtime

Skippy

2026 – Present

The current private agent layer I use across active projects. It connects model calls to real tools—GitHub, Railway, email, document review, Telegram—with explicit routes for drafting, approval, and audit. The work is the integration layer: permissions, handoffs, memory, and operating discipline.

Private Runtime Human-in-the-loop Tool Access

The throughline is accountability. Every step moved more work into AI-assisted systems only where the action could stay bounded, inspectable, and reversible.

Ecosystems

Aposema

7 repos

AI content protection and accountability infrastructure for publishers and agentic systems.

Envelope

2 repos

Agent accountability infrastructure — governance through auditability, boundaries, and reviewable records.

Redline

1 repo

AI-native document review — share confidential docs with AI reviewers via time-limited access tokens.

MCP Licensed Servers

5 repos

AI tool wrappers that natively respect creator rights

Expatriator

4 repos

Immigration assistance platform — Telegram bot, API, appointment automation

What I build with

TypeScript
Rust
JavaScript
HTML
CSS
Python
TOML
Shell

TypeScript, Rust, and Python lead — across 50+ public and private repositories, from CLIs and agent infrastructure to web apps.

Repos curated

diskspace

Personalized, reversible disk-cleanup CLI for macOS. 85+ declarative YAML rules. Drivable by humans and agents. Signed, notarized universal binary.

Rust 1mo ago
View →
governor

Trust router for autonomous agents — hardens tool use against sloppy, over-eager, manipulative, or malevolent AI activity. Open source.

Rust 2mo ago
View →
envelope

Agent accountability infrastructure built on email primitives — governance through auditability, boundaries, and reviewable records.

TypeScript 2mo ago
View →
redline

AI-native document review via time-limited access tokens. Share confidential docs with AI reviewers without permanent exposure.

TypeScript aposema 3mo ago
View →
provisionally

Non-confidential fit briefs for protected inventions. Structured disclosure format that surfaces market relevance without revealing confidential IP.

2mo ago
View →
foundermode-studio

Tools for high-agency founders. Teleprompter purpose-built for 60-second pitch videos — practice, refine, ship.

TypeScript 3mo ago
View →

Milestone timeline

2024 – Present

Lobster Labs

Founder / Builder / Inventor

Agent-native software studio building AI-native products and infrastructure: agent governance, accountable tool-use systems, email-as-infrastructure, secure review workflows, and developer tooling. Public IP posture: twelve USPTO filings across protected problem domains; filed applications, not granted patents.

2021 – Present

Expatriator.com

Founder

Relocation and immigration product work rooted in the long-running SpainExpat audience: matching expats with vetted lawyers, service providers, and practical admin paths.

2015 – 2023

Actigram Labs / MCS

Product

Digital health and behavior-change products, including MCS work that continued through 2023 around clinical-trial and patient-support programs.

2014 – 2016

Turf.ly

Founder

Location-based social fitness game that turned walking into a team sport.

2012 – 2014

Curelator

Head of Product

Personalized medicine startup; led product for migraine tracking and trigger-analysis workflows.

2004 – Present

SpainExpat.com

Founder & Editor

Independent resource for expats in Spain and the original long-running audience behind later relocation/admin products.