Shannon Labs

Open systems that expand who gets to create with intelligence.

Home of Codewhale, an open-source coding agent that works across models, tools, and platforms. Codewhale is Shannon Labs’ first product and full-time focus.

The work

Codewhale

Open source · Any model · Your machine

Give Codewhale a model and a task. It reads the repository, edits files, runs commands, checks the result, and stops when the job is done or it needs a decision from you.

The open-source runtime works in the terminal. The Codewhale app adds a browser and mobile surface for starting and supervising the same work, including sessions, approvals, connected computers, and projects.

License
MIT
Runtime
Rust
Models
Bring your own
Focus
Real workspaces

Codewhale · work session

A Codewhale browser session showing completed code changes and test results
The Codewhale app supervising a live workspace session.

A name I did not pick

Brother Whale

鲸鱼哥 Jīngyú gē

Codewhale began as an independent terminal project. Hunter opened a thread with “Hello, Whale Brothers,” and developers in China started calling him 鲸鱼哥—Brother Whale.

The project found an international open-source community before there was a company around it. Contributors became translators, testers, maintainers, advocates, and friends. Shannon Labs is building the next product layer without losing that origin.

Hunter Bown with members of the open-source community in Hangzhou
Hangzhou · May 2026 The early open-source community around the project that became Codewhale.

Soundings

How it got here

A sounding measures depth by making a noise and listening. The projects below form a practical lineage: each one supplied an idea or tool used by the next.

0 m

Current work

Codewhale

A coding-agent runtime for real repositories, with terminal, web, and API surfaces around the same work.

codewhale.net
400 m

Long context

Aleph

An MCP server that turns recursive-language-model analysis into tools an agent can call: search a large corpus, run code against it, and recursively query the material instead of forcing it into one prompt.

Source
600 m

Earlier harness

RLM Agents

An earlier agent harness for recursive context handling. Its useful parts led to Aleph’s MCP surface and informed Codewhale’s approach to long-running work.

Source
900 m

Information theory

Shannon Control Unit

Research on adaptive regularization and compression-based anomaly detection: using description length as a signal about what a system is learning or when data has changed.

Source

Why Shannon Labs

Murray Hill, 1950

My great-grandfather ran research at Bell Labs.

Shannon Labs is named for Claude Shannon. Hunter’s great-grandfather, Dr. Ralph Bown, was vice president of research at Bell Labs during the same era. The current principle is straightforward: build systems people can inspect, run, question, and improve.

01

Open systems

Useful work should be available to test and argue with.

02

Working artifacts

A runnable system says more than a claim about one.

Hunter Bown speaking at the Global Tech Leadership Conference
Hunter Bown Founder of Shannon Labs and full-time developer of Codewhale.

The founder

I taught band for seven years before I wrote any of this.

Hunter Bown’s path to software ran through music education and conducting, an MBA, and patent-law studies. AI-assisted development gave him a way to turn product ideas into working systems; Codewhale grew from an independent terminal agent into an international open-source project.

He now develops Codewhale full time through Shannon Labs, working in public with the contributors and users shaping the runtime and the app.

For investors

Codewhale is the first product.

Shannon Labs is building the open-source runtime and the product layer around it. If you invest in developer tools, AI infrastructure, or agentic work, contact Hunter directly.

hunter@shannonlabs.dev