Sahil Lavingia has published a diary recounting his time as a member of Elon Musk’s DOGE workforce. It’s a short read — Lavingia’s DOGE stint lasted just 55 days — but it is does provide new details on the temporary government organization formed by President Trump’s executive order.
Lavingia is a well-known name in Silicon Valley from his days as an early employee of Pinterest, to his current gig as founder of Gumroad, a platform where creators can sell their goods. He’s also a well-known seed and angel investor.
He joined DOGE as a software engineer for the Department of Veterans Affairs (VA) in mid-March, he wrote. What stands out from his account is his surprise that the 473,000-employee government agency had strict rules on who could be targeted in a layoff; and he quickly learned that all things at the VA were not as inefficient as he imagined. He also lamented that DOGE itself isn’t a well-oiled machine.
As a volunteer who had a salary of $0, he was immediately tasked with identifying “wasteful” contracts and the people the VA should layoff, he wrote. But he was surprised to discover facets like seniority and the person’s veteran status (this was the VA, after all) determined who could be targeted. Performance could be factored in lower on the list, in Lavingia view.
He also described DOGE’s advisory role as like a McKinsey management consultant, and said DOGE is not responsible for the actions taken by the orgs. “DOGE had no direct authority. The real decisions came from the agency heads appointed by President Trump, who were wise to let DOGE act as the ‘fall guy’ for unpopular decisions,” he says.
This is similar to what Musk was decrying this week to the Washington Post. Musk described DOGE as D.C.’s “whipping boy” blamed for every unpopular decision.
Lavingia said he joined DOGE after campaigning for Bernie Sanders in 2016 because he dreamed of writing code for the government that helped people at scale. Because his DOGE missives didn’t take much time, he said he worked on projects that interested him, including overhauling the UX of the VA’s already-in-use LLM-based chatbot.
He said he built a fairly long list of stuff in his less-than-two month stint, but didn’t get a chance to do enormous projects, like “improving the UX of veterans’ filing disability claims or automating/speeding up claims processing.”
And, he wrote, “I was never able to get approval to ship anything to production that would actually improve American lives — while also saving money for the American taxpayer.”
He was, however, given permission to open source much of his work. His work included a tool that scanned internal PDFs for terms “related to DEI, gender identity, COVID policies, climate initiatives, WHO partnerships,” he described on the tool’s page, as well as tools that used LLMs to analyze contracts, and a tool for building org charts.
He also made observations about the lack of organization in DOGE itself. “I wondered why there wasn’t a centralized DOGE software engineering playbook with all of our learnings; overall, I was surprised by the lack of knowledge-sharing within DOGE. It seemed like every engineer started from scratch.”
He was unceremoniously axed from DOGE on Day 55 after he discussed his work there with a reporter from Fast Company. “I got the boot from DOGE,” he wrote. “Soon after publication, my access was revoked without warning.”
In that FC interview, however, he also said working up close with the VA taught him that, while it was slow like a giant enterprise, it still “works.”
“I would say the culture shock is mostly a lot of meetings, not a lot of decisions,” he says. “But honestly, it’s kind of fine—because the government works. It’s not as inefficient as I was expecting, to be honest. I was hoping for more easy wins.”
His experience captures perfectly the dilemma of keeping enormous government agencies modern as they remain functional. While all taxpayers would like less waste, and the government can surely benefit from more programmers immersed in the latest tech, perhaps Silicon Valley volunteers swooping in like they are building a startup from scratch isn’t the answer.
Lavingia did not immediately respond to our request for additional comment.
