Ethen topic
An AI coding agent is an AI agent that works inside a software repository: it reads the code, plans a change, edits files, and checks the result, usually by running tests.
The practical questions are where the agent runs, what it can see, and how its changes are checked. Conversational help, cloud workspaces, and local repository setups trade convenience against access to the real repository and its tests.
Ethen Code is Ethen's coding product. The articles below describe where it runs and how it keeps one user's history separate from another's. Ethen Research Lab has not yet published coding-agent-specific research. The publications listed here define methods that apply to coding agents among other kinds of agent.
Ethen Code is designed as one system with three surfaces: light conversational coding in Chat, a full cloud engineering workspace in Platform, and a local coding-agent environment in Desktop. The article separates what is registered in code today from the intended product direction.
Match the environment to the work. If the work fits in one message and leaves no repository trace, conversational help is enough. If it spans files, branches, builds, or team review, use a workspace with repository scope. If it must stay on your machine or use a local runtime, use a local environment with explicit local controls.
Ethen Code scopes local run history to the signed-in account with keyed storage, sweeps it on sign-out, and broadcasts across tabs so one user's sessions stay isolated from another's.
Not specifically. VerifiedWork is a benchmark design that includes executable software environments. The failure-genome paper draws on published coding-agent failure studies. The context-compaction proposal and the VerifiedWork Context design address what long-running agents must remember. None of these reports Ethen coding-agent measurements.
Ethen Code is designed as one system with three doors: quick help in Chat, a full workspace in Platform, and a local environment in Desktop.
A durable checklist for matching the task to conversational help, a cloud workspace, or local repository work.
Ethen Code keeps local run history behind a per-account key, discards the previous scope on identity change, and tells other tabs to do the same.
None of these publications reports measured coding-agent results. They are benchmark designs, a research paper that synthesizes published work, and a proposal.
A benchmark design for AI agents that take action: executable environments, state-based verification, cost reporting, permissions, recovery and reproducibility.
A research paper proposing a multi-axis AI agent failure taxonomy, the failure genome, built on recent work on failure modes, attribution and critical steps.
A research proposal for context compaction that keeps obligations, permissions, deadlines and evidence out of lossy summaries while cutting an agent's token cost.
A benchmark design measuring what AI agents must remember: obligations, constraints, permission changes, superseded facts and revoked data, across compaction.
Ethen Code is moving from assistive coding — explaining code, writing functions, fixing snippets — toward software tasks that end with evidence a person can check. The direction has seven parts: reproduce a problem before changing anything; plan a minimal, reviewable change before executing it; work in isolated environments with scoped permissions; treat tests, builds and review as evidence rather than as a finish line; put approvals in front of merges, deployments and other consequential steps; recover from failures in long-running work without repeating effects; and keep the same task model across Chat, the cloud workspace and Desktop. This article describes that direction. It is not a release schedule, and it makes no availability or date claims.
Using AI agents in software development has changed how Ethen is built, but not in the way people usually expect. The biggest change is not raw speed. It is where human attention goes. Agents now do much of the mechanical work — exploring unfamiliar code, implementing scoped changes, writing and running tests, and drafting reports of what they did. People spend more of their time defining tasks precisely, reviewing the evidence an agent produces, and deciding what is claimed, merged and deployed. Every agent-made change has to carry an evidence report saying what ran, what passed and what did not run, using the same status words as our release certificates. And no agent merges or deploys to production without explicit human authorization. This article describes that practice, what the research says about AI-assisted development, and what we have learned.
Ethen is being built across three kinds of surface because no single one is best at everything. The web is where Ethen is quickest to reach: conversation in Ethen Chat and cloud workspaces you can open from any browser. The Ethen desktop app is where Ethen can work with your local files, your own toolchain and local coding. Local AI models, reached through the desktop app, let some work run entirely on your own machine — useful for privacy, offline work and experimentation. What ties them together is not that every surface does everything, but one account, shared project state where it makes sense, and the same rules about memory and approvals everywhere. Each surface and device keeps only the access it needs. This article explains that design. It describes product direction and published engineering; it does not announce downloads, release dates or capabilities beyond those already described in Ethen's engineering posts.