On June 23, 2026, Justin Poehnelt published a thread on X that reached 1.1 million views. It is one of the more precisely ironic stories to come out of the tech industry this year.
He built a Google Workspace CLI. It went viral. It hit #1 on Hacker News. It gained thousands of GitHub stars and many thousands of users in days. Then Google fired him.
Two days before they fired him, Google Cloud Next announced that an official Workspace CLI was coming.
"I think the cause was that Workspace and certain leaders (and projects) were afraid of being disrupted. But the fear wasn't specific to my CLI, it was a broader fear in what agents meant for Workspace."
Poehnelt framed his post as personal — he wanted the story out in his own words, as part of his healing after nearly seven years at a company. But the story it tells is bigger than one engineer's termination. It is about what happens when a large company's internal immune system fires at one of its own builders at exactly the moment the company needs to move faster.
What He Built
The tool is the Google Workspace CLI — an open-source command-line interface designed for both humans and AI agents.
It covers the full Google Workspace surface: Drive, Gmail, Calendar, and every Workspace API. It ships with 40+ agent skills built in — the kind of structured, tool-callable interfaces that let AI agents perform Workspace actions as part of automated workflows.
Addy Osmani — himself a prominent Googler and engineering lead, now also ex-Google — announced the tool on March 5, 2026:
"Introducing the Google Workspace CLI: built for humans and agents. Google Drive, Gmail, Calendar, and every Workspace API. 40+ agent skills included."
The post spread immediately. Within days, the CLI was trending on Hacker News — hitting the top spot — and the GitHub repository was accumulating thousands of stars from developers who had been waiting for exactly this kind of agent-native Workspace integration.
Where It Came From
One of the first questions people asked on X after Poehnelt's post: was this a 20% project? Something he built on his own time without authorization?
No.
Poehnelt was on the Workspace DevRel (Developer Relations) team. In his own words from the thread replies:
"I was on the Workspace DevRel, we regularly create open source layers and abstractions over APIs."
Building open-source tooling on top of Google APIs is explicitly part of what Workspace DevRel does. The CLI was built in that context — not as a rogue side project, not as a personal experiment, but as the kind of developer-facing abstraction that DevRel teams exist to produce.
The controversy, according to Poehnelt, centered on a narrower issue: the Google logo and brand colors appearing on the GitHub repository. Legal questioned him about why an unofficial tool was visually presenting itself with Google's brand identity on Google's own GitHub organization.
This is a genuinely complicated area. A tool under the googleworkspace GitHub organization using Google's visual identity reads as an official Google product to most developers — which creates liability questions around what the tool promises and who is responsible if it breaks. That is a real concern. It is also, notably, the kind of concern that has a resolution short of termination.
The Irony: Two Days Before He Was Fired
Poehnelt does not specify the exact sequence of events in detail, but the central fact is stark: Google Cloud Next announced an official Workspace CLI two days before he was fired.
This is not a situation where Google saw his tool, decided to build their own, and then let him go as the project moved in-house. The official announcement and his termination were effectively simultaneous — close enough in time that they occupied the same news cycle.
The read Poehnelt offers for what actually drove the decision:
"I think the cause was that Workspace and certain leaders (and projects) were afraid of being disrupted. But the fear wasn't specific to my CLI, it was a broader fear in what agents meant for Workspace."
This is the more interesting layer. The Workspace CLI did not just hit #1 on Hacker News — it surfaced a question that product leaders inside Google were apparently uncomfortable having surfaced publicly: what does an agent-native future look like for Workspace?
When developers can interact with Drive, Gmail, and Calendar through a CLI — or, more significantly, when AI agents can do so through 40+ structured skills — the nature of Workspace as a product starts to shift. The value proposition of the applications changes. The workflows change. The integration patterns change. The fact that an internal DevRel engineer built a proof of concept that developers immediately understood and adopted at scale made that future concrete and visible in a way that internal planning documents do not.
The hypothesis: some leaders within Workspace found that concrete visibility uncomfortable, not because the CLI itself was a threat, but because the agent-native direction it implied was a threat to existing projects and roadmaps.
What the Reaction Reveals
The X thread generated extensive debate beyond Poehnelt's own account. A few threads worth noting:
On brand usage: Several commenters argued that using Google's logo and name on an unofficial tool is a clear violation regardless of context — "No matter how big or small, using a brand's name and logo in a project without their consent is infringement of many things, if the product is not an official product of that brand."
This is technically accurate. But it collides with how developer relations at large tech companies actually operates. DevRel teams routinely produce samples, libraries, and tools that carry their employer's branding, hosted in their employer's GitHub organizations, because that is how developers discover and trust them. The line between "unofficial OSS from DevRel" and "official Google product" has always been blurry — deliberately so, because blurriness increases adoption.
On organizational dynamics: "There has to be way more to the story. Any director+ would immediately have picked up the hype and based on my experience with legal, their escalation would not have resulted in a firing, just in a PR cleanup campaign."
Poehnelt's own account acknowledges this ambiguity — he describes directors and leaders asking what they could learn from the tool alongside legal questioning him about the brand. The fact that both reactions occurred simultaneously — senior leaders wanting to learn, legal wanting to know who authorized the Google logo — is consistent with a large organization where different functions respond to the same event in contradictory ways.
On the outcome: Poehnelt is explicit that he had wonderful teammates and a manager who supported him through the final months. The decision, wherever it came from, was not the universal view inside Google. It was specific to whoever made the call.
The Timing with Addy Osmani
Addy Osmani's departure from Google — he was a Senior Staff Engineer and widely recognized for his work on Chrome DevTools and web performance — followed in the same period. Osmani was the one who publicly announced the Workspace CLI in March 2026 before it went viral.
The proximity of these departures, both connected to the Workspace CLI story, has not been officially explained. Whether they are causally connected, coincidentally proximate, or part of a broader pattern is not public.
What is public: the tweet Osmani posted announcing the CLI — which Poehnelt refers to as "the tweet that got me fired" — is still live, still circulating, and still directing developers to the googleworkspace/cli repository.
The Agent-Native Question at the Center
Strip away the employment story and there is a product question underneath this worth naming directly: what does an agent-native interface to Workspace actually look like, and why does it make some product leaders nervous?
The Google Workspace CLI answers part of that question concretely. You can interact with Drive, Gmail, and Calendar from a terminal. An AI agent can call those same operations as tools, using the 40+ skills the CLI ships with, without needing a GUI or a browser session.
This is the same category of shift happening across every major productivity suite. Microsoft 365 has been building agent-native interfaces through Copilot and the Microsoft Graph API. Notion, Slack, and Linear all have structured tool interfaces for AI agents. The question is not whether Workspace will have one — Google itself announced it will — but how it gets there and what organizational dynamics shape the transition.
The Poehnelt story suggests that at least some internal Workspace stakeholders viewed the agent-native future as a threat rather than an opportunity, and acted accordingly when someone on their own team made that future visible at 1.1 million impressions.
What Happens to the CLI
The repository at github.com/googleworkspace/cli remains live. It has not been taken down. The code Poehnelt and his collaborators wrote is still there for developers to use.
Google's official Workspace CLI — the one announced at Cloud Next 2026 — is presumably in development. The relationship between the two (whether the official version incorporates any of the open-source work, whether the open-source version continues to receive updates, whether they diverge into separate tools) has not been addressed publicly.
Poehnelt is now building at helensfoundry.com. He published a post on his personal blog titled "Rewrite Your CLI for AI Agents" — part of the thinking that informed the tool's architecture — which remains worth reading for anyone building agent-native developer tooling.
Claude for Work
Use Claude as a thought partner for writing, research & decisions — no coding required. 2 live sessions with Yash Thakker.
Claude for Work is a 2-day live workshop on using Claude to supercharge your daily work — writing, research, analysis, and decision-making — without any coding required. Learn how to set up Claude Projects with custom instructions, run deep-research sprints, co-write documents that sound like you, and build repeatable prompt systems for your team. August 1–2, 2026. Hosted by Yash Thakker, founder of AISOLO Technologies, instructor to 350,000+ students.
Includes 1-year access to all session recordings, a personal prompt library, Discord community access, and a certificate of completion. No coding or technical background required. Designed for managers, marketers, founders, and writers.
The Broader Pattern
The Poehnelt story is not isolated. Large technology companies regularly fire or push out internal developers who build things that embarrass or threaten existing product lines — even when those developers are building on sanctioned time, in sanctioned repositories, in ways consistent with their job description.
The mechanism is usually not a straightforward "you built something we didn't want." It is more typically a combination of factors: a brand concern that gives legal a handle, a product concern that gives leadership a motivation, and an employment relationship that gives the organization a lever.
What makes the Poehnelt case specific to 2026 is the agent framing. The fear he describes is not fear of a CLI. It is "a broader fear in what agents meant for Workspace." That fear is going to surface repeatedly in the next several years as AI agent interfaces challenge the application-layer assumptions that enterprise software is built on.
Companies that resolve that fear by removing the internal people who make it visible will be slower to adapt than companies that resolve it by hiring and promoting them.
The two-day gap between Poehnelt's firing and the official Workspace CLI announcement is a precise illustration of which approach Google chose — and then immediately recognized was wrong.
Justin Poehnelt's thread is at x.com/JPoehnelt. The Google Workspace CLI is at github.com/googleworkspace/cli. His post on CLI design for AI agents is at justin.poehnelt.com.