justinwp

The launch was canceled after the fact.

The "G" logo is a GitHub organization setting that applies to all repos in GitHub.com/googleworkspace and has for years.

> they were about to announce in 2 weeks

Market validation can change roadmaps. As I stated on Twitter/X, CLIs for agents was not a very interesting thing outside a small group earlier in 2026.

Wasn't ignored or skipped.

Yes there was a launch with eng bit flipped by manager.

There was a Ariane/Launch with bits flipped including the eng bit from my manager.

There is a two "calendar" launch process for OSS at Google, one "calendar" is org specific, the other OSS. I followed the process and each of the bits were flipped in the same way I had always done it.

Yes

You are assuming that it was "personally" releasing something and that the process wasn't followed.

To clarify, I was on the Google Workspace Developer Relations team, the majority of my work was that exact OSS release process. It is not clearly documented and always changing. You can read some of it here, https://opensource.google/documentation/reference/releasing/..., but like I said it is always changing. Relevant: https://www.theregister.com/software/2023/01/27/what-is-goog...

I am not going to share much more than what I already have, but I think this speaks to the experience of working in big tech and the disruption caused by AI both at the level of teams/roadmaps/incentives and changing user behavior.

Location: Durango, Colorado

Remote: Yes

Willing to relocate: Yes

Technologies: Rust, TypeScript, Python, Go, SQL, React, SvelteKit, GCP, AWS, Kubernetes, Docker, Applied ML, MCP

Résumé/CV: https://justin.poehnelt.com | https://github.com/jpoehnelt

Email: justin.poehnelt@gmail.com

GitHub: https://github.com/jpoehnelt

LinkedIn: https://www.linkedin.com/in/justin-poehnelt/

I am a Senior Software Engineer with a decade of experience bridging deep machine learning foundations with modern agent infrastructure. Previously a Senior Developer Relations Engineer at Google and a Founding Engineer at Descartes Labs.

Recently, I architected the open-source https://github.com/googleworkspace/cli in Rust—built entirely via AI agents which hit 20k+ stars and #1 on HN. My background spans building petabyte-scale Kubernetes architectures, engineering visual search engines over billions of satellite images, and developing full-stack embedded Rust IoT firmware with custom LoRa mesh networking for endurance races.

I excel at taking systems from 0 to 1 and driving adoption developer experiences. I'm looking for Senior/Staff or Founding Engineer roles where I can tackle complex distributed systems or build secure, agent-first infrastructure.

It's complicated :)

Wrote a blog post on my experience and reasoning for a MCP server to drive my gpui app.

It's deliberate, because you generally do not want calories from fat or protein during a marathon or other running race.

Try building a CLI with a complex JSON as flags approach. :)

You don't need to install all of them.

You don't need to use gcloud if you already have:

1. A GCP project (needed for OAuth) 2. Enabled APIs in said project

You can skip that setup if you already have the OAuth credentials.

There are many gotchas in this process and unfortunately there is no easy way to deal with the OAuth setup.

Yeah, still no way around this unfortunately.

Title is a little clickbait, but not really! :)

---

Human DX optimizes for discoverability and forgiveness. Agent DX optimizes for predictability and defense-in-depth. These are different enough that retrofitting a human-first CLI for agents is a losing bet.

Drive, Gmail, Calendar, and every Workspace API. Structured JSON output. 40+ agent skills included.

I recently formalized a `.agent/workflows/knowledge-extraction.md` to extract high-signal engineering patterns from Antigravity’s persistence layer to make them explicit and portable.

Use a rules file to change the Git identity for agents in your repository.

I had a teammate asking me about calling MCP servers from Apps Script, so I wrote this up!

An example of using AI for my blog, https://github.com/jpoehnelt/blog/pull/493.

I was using Jules for AI assistance with my blog (tagging, components, linking, etc), but git LFS stopped working and I decided to reassess its usage in today's dev environment.

there are multiple cycles from highs to lows and back and then typically a larger dominant split similar what was discussed here for the marathon but scaled to the distance.