What we’re building next.

Eliminate remaining edge cases in the cross-platform hook layer: PowerShell path quoting, signal handling on Windows, and CI matrix coverage for all six hooks.

Port the session-start greeting, onboarding wizard, and genome scribe loop to the Cursor and Goose adapters in /ports so every supported host gets the same first-run experience.

The session-start notifier already ships — it compares the installed version against the latest GitHub release and prints a one-line notice at most once per day per version. Q2 work focuses on richer presentation: changelog excerpt in the notice, opt-in one-command apply, and a quieter status-line indicator for users who'd rather not see stderr output.

A metered plan for large engineering teams that prefer per-token billing over per-seat. Requires a usage-reporting endpoint in the plugin and a Stripe metered subscription integration on the backend.

This page, plus a bespoke MDX blog for technical deep-dives, release notes, and engineering transparency posts.

Upgrade the team genome encryption model from a single shared AES-256-GCM key to per-member X25519 key exchange with encrypted key envelopes, so a member rotation does not require a full re-encrypt of the genome store.

A one-command deploy path for the Hono/Bun backend onto Fly.io, with a generated fly.toml and environment variable checklist. Teams that want full data sovereignty can run their own instance.

A JSON dashboard definition that works against the existing Postgres schema — per-team token savings, genome hit rate over time, LLM summarizer latency histogram. Import it in three clicks.

Allow a parent genome to pull sections from child repo genomes at build time, so a monorepo or multi-repo org can keep per-package genomes without duplicating shared architecture documentation.

When a CRDT merge produces a conflict in the genome store (two writers edit the same section within the same clock tick), surface a side-by-side diff in the terminal and prompt for a resolution strategy.

Users who opt in can publish their anonymized savings stats to a public leaderboard page, enabling community benchmarking across codebases and team sizes.

A GitHub Actions step that posts a summary of genome changes as a PR comment: which sections were added, modified, or removed, and what token-savings impact the change has on the benchmark.

A JetBrains plugin that exposes the same MCP tools to AI assistants inside IntelliJ IDEA, PyCharm, and Rider.

A JupyterLab extension that integrates the genome retrieval index with notebook context, so AI tools inside Jupyter see a compressed view of the project rather than full file contents.

A fully air-gapped deployment option: local inference, local genome store, SAML/OIDC SSO, and no outbound network calls except to the customer's own identity provider.

A browser extension that instruments ChatGPT, Claude.ai, and Gemini web interfaces and reports token usage to a unified dashboard alongside the plugin's native Claude Code stats.

Expose the genome retrieval index and compressed read/grep tools to GitHub Copilot via VS Code's MCP extension host, extending the savings to the Copilot workflow.

A feedback mechanism where the agent rates genome sections by usefulness after each session, and the genome scribe loop uses those ratings to weight future section selection.

Integration with Zed's AI assistant panel via the MCP protocol, giving Zed users access to compressed reads and genome retrieval.

TypeScript hooks, Windows PowerShell shell selection, multi-OS CI matrix, VS Code extension packaging.

Per-section random nonces, keys at ~/.ashlr/team-keys, server stores ciphertext only. Admin dashboard and public status page.

Auto-firing first-run wizard, transactional HTML email templates, session greeting.

Shared genome store with CRDT merge, conflict resolution, per-org retrieval index.