Vezryn Updates

Graph Community v0.1.1 released

Wed, 11 Nov 2026 00:00:00 GMT

Graph Community v0.1.1 is available on GitHub. Read the published release notes and download the release from the public repository.

In this release

Restore 22 engineering playbooks, 10 skills and six policies, with native task, model and reasoning guidance and editable library overrides.

Add readable future claims with public-safe worktree fields, truthful unknown attribution and private source-custody preservation.

Explain public issue reports and pull requests, maintainer-controlled readiness and contributor verification.

Release and documentation

These highlights reproduce the published release notes. Consult the release and repository documentation for requirements, verification details and current qualification limits.

Public sources

Introducing Graph Community v0.1.0

Wed, 11 Nov 2026 00:00:00 GMT

Graph Community is now open source under Apache-2.0: GitHub engineering coordination on your own machine, built around native Codex workers and recoverable execution.

From an issue to reviewable work

Vezryn has released Graph Community v0.1.0, the open-source edition of Graph, our autonomous engineering orchestration system.

Community watches explicitly enrolled GitHub repositories for actionable issues and launches a native Codex coordinator. The coordinator can claim work, delegate native workers in isolated worktrees, validate changes, and produce ordinary pull requests. Local ownership records and recovery safeguards help keep unfinished work visible across interruptions.

A workflow developers can inspect and adapt

We are making Community open source so developers can inspect how the coordination works, adapt its editable Markdown workflows, and experiment with engineering orchestration on their own machines.

The release includes source archives, a quickstart, platform documentation, a manifest, and checksums. Start with the accounts-free local checks, then try a bounded issue in a disposable private repository. Live delivery requires Python 3.11+, Git, authenticated GitHub CLI and native Codex access, trusted lifecycle hooks, and explicit repository enrollment.

Community is a headless, user-operated edition, released under the Apache License, Version 2.0. It does not include Vezryn’s complete private Graph system, company-specific playbooks, private metrics, a desktop interface, or model access. Friday and Wednesday remain in development.

Qualified on specific hosts, with documented limits

Linux x64 qualification includes native issue-to-worker-to-validated-PR delivery, lifecycle events, service controls, and recovery checks; Linux reboot recovery is not claimed. macOS arm64 qualification includes native delivery, stop, interruption, and a reboot/login trial; other architectures and client versions are not qualified.

A native Windows 11 x64 VM trial demonstrated issue-to-worker-to-PR delivery, lifecycle hooks, competing-watcher exclusion, and graceful drain/restart. Windows startup is manual; reboot recovery and Windows arm64 remain unqualified. WSL2 has no host qualification and is not native Windows evidence.

These are bounded host observations, not universal compatibility or support guarantees. The current workflow uses native Codex; other coding-agent harnesses are not qualified. Read the platform limits and verify the intended client and trusted hooks before unattended work.

Public sources

Vezryn Joins the NVIDIA Inception Program

Tue, 10 Nov 2026 00:00:00 GMT

Vezryn has been accepted into NVIDIA Inception, supporting our work on local-first AI software and engineering tools.

A new company milestone

Vezryn has been accepted into the NVIDIA Inception Program.

Resources for building with AI

NVIDIA Inception is a free program for eligible technology startups building with AI. It provides access to developer tools, technical training, and the NVIDIA ecosystem, along with preferred pricing on select hardware and software and offers subject to eligibility.

Aligned with our development direction

Vezryn is developing local-first AI software and engineering tools: Friday, a personal AI workspace; Wednesday, an engineering workspace; and Graph, an engineering coordinator. These products remain in development and are not generally available.

Access to technical resources and the NVIDIA ecosystem will help us evaluate AI capabilities, learn from the wider developer community, and make informed choices as we build. Our direction remains focused on useful software, local control, and the practical conditions in which people work.

Program membership does not imply NVIDIA investment in Vezryn or endorsement of any individual product.

Public sources

Start with the work. Build from there.

Mon, 09 Nov 2026 00:00:00 GMT

A practical approach to AI engineering: understand the intended outcome, shape the tool around the work, and learn from actual use.

Understand the intended outcome

Vezryn starts with the person, the problem, and the outcome that matters. Context and constraints shape what a useful tool needs to do.

Engineer for real conditions

Choosing technical capability is part of the work. Usability, reliability, cost, privacy, security, and adaptability also matter when that capability becomes working software.

Evaluate and refine

A working prototype gives an idea something concrete to be tested against. The approach described on our company site is to test assumptions in actual use, learn from those conditions, and refine the tool.

These are engineering principles, rather than a claim that a particular product is released or that every outcome has been independently verified.