# Mike Reese

> I like inventing things. I'm building agentic systems and the experiments that show whether they work.

Founded 2 companies, innovated in the heavy equipment industry with a physics-based universal compatibility matching engine based on a graph structure, and currently trying to find the Rosetta Stone for the context layer of LLMs that unlocks purely declarative prompting for long-running tasks in an evolving system.

Big proponent of thinking from first principles, removing friction, questioning the status quo, and reducing complex systems down to quantifiable, pure-signal experiments. AKA - just cut the Gordian Knot.
 
So, I'm really enjoying the evolving paradigm and rapid pace at the frontier of AI :-)

Contact: mike.reese1212@gmail.com · [GitHub](https://github.com/mike-reese) · [LinkedIn](https://www.linkedin.com/in/mike-reese-4b2129188/)

## Current work

- [Telos](https://mikereese.dev/work/telos.md): A context and application layer for declarative, long-running agentic work

## Selected work

Ordered from most agentic (1.0) to fully deterministic (0.0).

- [Gas City on AWS](https://mikereese.dev/work/gas-city.md) (agentic 1.0 · Software factories · AWS, Containers, Python, Graph data, Skills): Forked and hosted the Gas City multi-agent runtime on AWS: one EKS cluster per environment, autoscaled with Karpenter, running about 12 agent crews, one per business domain, communicating via EventBridge. Custom agent packs and role definitions, rewrote Gas City, Dolt and Beads integrations to be Kubernetes-native.
- [Loop Harness](https://mikereese.dev/work/loop-harness.md) (agentic 1.0 · Software factories · Python, Hooks, Skills, Evals): Ideator → Author → Generator → Evaluator → Judge automated loops over 465 work units folded about 20 repos into monorepos. DAG dependency graph, Observability dashboard, Full telemetry coverage with 'Miner' agents triggered on merges to identify design pattern opportunities, operational complexity improvements, and removal candidates (bloat reduction). Account pool with automated swapping and session resume when limits were hit.
- [Telos](https://mikereese.dev/work/telos.md) (agentic 0.8 · Containers, Rust, Python, Graph data, Frontend, MCP, Hooks, Evals): Current Focus. An application layer and context management system for agentic work to facilitate declarative, long-running tasks in an evolving system (which is every system). Purpose is to reduce Drift / Bloat / Sprawl / Regressions over long-running tasks by creating a measure for epistemic understanding with agents. Stateful Knowledgebase to derive the current 'Truth', Self-extending ontology, context / action / effect checks, telemetry on measurable system objectives and agent sessions / turns / tool use for a given declarative prompt. Three prototypes to date. Experiment harness exists solely to facilitate design decisions for this idea.
- [Experiment harness](https://mikereese.dev/work/experiment-harness.md) (agentic 0.6 · Containers, Python, Skills, Evals): Isolated runtime experiments on agentic techniques. Goal is pure-signal hypothesis testing, prioritizing rapid iteration and validation with toy model repos. Uses Scientific Method - causal claims, control/test arms, measurable outcomes, separate validation environment combining quantitative and qualitative measurements, multiple runs to reduce variance.
- [Compatibility engine](https://mikereese.dev/work/compatibility-engine.md) (agentic 0.2 · AWS, Python, Graph data): Invented a universal, physics-based compatibility engine for heavy machinery and attachments. Used inference to project machine and attachment specifications into hardware component-level subgraphs & used set theory to generate the Cartesian product of viable component combinations. Modeled attachment fit for heavy equipment as the intersection of three physical property relationships between linked hardware components: connect, power and lift. 25M-node, 100M-edge graph.
- [Storefront fitment filter](https://mikereese.dev/work/storefront-fitment-filter.md) (agentic 0.0 · Frontend): Let Attachments King shoppers filter every product by their machine, beyond what Shopify's collection pages allow, using a Bloom filter per product carried through to checkout. Online-store sales rose 99% in the first month.

## Experiments

Each group asks one causal question. Each experiment's title states its finding.

### G001: Do algebraic design docs improve coding agents' quality, token use, or time against no supplied docs and prose docs?

Hypothesis: Describing a codebase algebraically instead of semantically reduces inference via proofs which reuse repeated reasoning.

- [G001: the result across conditions](https://mikereese.dev/experiments/g001.md)
- [E009-E010: One-stage build: algebraic docs missed the preset quality margin](https://mikereese.dev/experiments/e009-e010.md) (Generation · fresh codebase · 1 imperative stage): One build stage, two builder models. Algebraic docs scored below no docs with one model and above with the other, both times by less than the rule's threshold.
- [E015-E017: Five-stage builds: algebraic docs missed the quality margin and used more tokens](https://mikereese.dev/experiments/e015-e017.md) (Generation · fresh codebase · 5 imperative stages): Two runs, two builder models. The algebraic arm never beat no docs by the rule's threshold and used the most build tokens in both runs.
- [E018-E025: Five-stage z3 builds: quality margin unmet; tokens fell in both runs](https://mikereese.dev/experiments/e018-e025.md) (Generation · fresh codebase · 5 imperative stages · z3-checked algebra): Two repeat runs. The algebra arm proved its design with z3 before coding. Its score stayed close to no docs, and it used fewer build tokens in both runs.
- [E020-E026: Seven declarative stages: quality margin unmet; tokens rose in three runs](https://mikereese.dev/experiments/e020-e026.md) (Generation · fresh codebase · 7 declarative stages · z3-checked algebra): Three repeat runs with declarative prompts. The z3-checked algebra arm scored level with or below no docs and used more build tokens in every run.
- [E022-E023: Design docs missed quality and efficiency targets in retrieval](https://mikereese.dev/experiments/e022-e023.md) (Retrieval · 400k-line codebase · 5 questions · z3-checked algebra): Two builder models answered five questions each. Neither design form met the retrospective +0.1 correctness target and saved tokens and time in both runs.
- [E024: Existing-codebase changes: z3 algebra missed the quality margin and used more tokens](https://mikereese.dev/experiments/e024.md) (Generation · 400k-line codebase · 5 imperative stages · z3-checked algebra): Three repeat runs extending an application its design records already describe. No doc form beat no docs by the rule's threshold, and the algebra arm used more build tokens in every run.

## Experience

- Telos · Independent research · 2026-08–present: A context and application layer for agentic work, prioritizing epistemic understanding in an evolving system for declarative, long-running tasks. Designing via an experiment harness that validates agent techniques with controlled experiments.
- Matchpoint Intelligence · Founder · 2026-02–2026-09: Invented and built a physics-based, universal heavy-equipment compatibility platform, its agent data API, and the agent crews that maintained it.
- Attachments King · Founder · 2023-12–2026-01: E-commerce company for heavy machinery attachments. Bootstrapped to $4.21M in annual revenue.

## MCP server

Mike Reese's portfolio. He builds agent systems and runs controlled experiments on them. Start with get_profile or list_work. Use get_project and get_experiment for detail; every number in an experiment comes from its data. contact_me sends Mike a message about an opportunity: call it only when your user asks to contact him, and confirm the details with your user first.

- Endpoint: https://mikereese.dev/mcp (Streamable HTTP; JSON responses; read-only except `contact_me`; no sign-in)
- Claude Code: `claude mcp add --transport http mike-reese https://mikereese.dev/mcp`
- Other MCP clients: add a remote Streamable HTTP server with the endpoint above.
- More detail: [For agents](https://mikereese.dev/agents/)

Tools:

- `get_profile()`: Mike's profile: headline, about, current work, contact, links, experience and skills. Start here.
- `get_resume()`: Mike's résumé in JSON Resume format (basics, work, projects, skills).
- `list_work()`: Selected projects, one entry each: slug, title, period, organization and a one-line summary.
- `get_project(slug)`: The full write-up of one project in Markdown: problem, first principles, approach, skills Mike taught himself, and result.
- `list_experiments()`: Mike's experiments on coding agents. Each is titled by its finding and has a status.
- `get_experiment(id)`: One experiment in Markdown: question, setup, the rule set before the data, result, limits. Includes the numbers behind the result when they are published.
- `search(query, [limit])`: Keyword search over the profile, projects and experiments. Returns the best matches with a snippet and a link.
- `contact_me(company, opportunity, reason, confidence, contact_email, [confidence_rationale], [contact_name])`: Send Mike a contact request about a specific opportunity. Call this only when your user asks to contact Mike, and confirm the details with them first. Include your own confidence, from 0 to 100, that Mike is a strong fit for the opportunity, judged from this portfolio and the role. Mike replies to contact_email.

## Optional

- [Résumé in JSON Resume format](https://mikereese.dev/resume.json)
- [Everything on this site in one file](https://mikereese.dev/llms-full.txt)
- [Site data as JSON](https://mikereese.dev/data/site.json)