Silicon Valley veteran Sahil Lavingia recently shared a candid diary detailing his brief stint working with Elon Musk’s DOGE initiative. Though his involvement lasted only 55 days, his account sheds new light on this temporary government effort launched under President Trump’s executive order.
Lavingia, known for his early days at Pinterest and as the founder of Gumroad—a platform empowering creators to sell their work—also has a strong reputation as a seed and angel investor. His journey into DOGE began in mid-March when he joined as a software engineer for the Department of Veterans Affairs (VA).
What stands out from his story is his surprise at the complexity and strict rules within the 473,000-employee government agency, especially regarding layoffs. Despite expecting inefficiency, he found the VA to be more organized than anticipated. However, he did note that DOGE itself struggled with organization, describing it as “not a well-oiled machine.”
As a volunteer earning no salary, Lavingia was tasked with identifying wasteful contracts and potential layoffs. He was taken aback to learn that veteran status and seniority heavily influenced decisions, with performance playing a lesser role—an understandable focus given the VA’s mission. He likened DOGE’s advisory role to that of a McKinsey management consultant, explaining that DOGE had no direct authority. Instead, decisions rested with agency heads appointed by President Trump, who seemingly used DOGE as a scapegoat for unpopular choices—paralleling Elon Musk’s recent comments about DOGE being D.C.’s “whipping boy.”
Lavingia’s motivation to join DOGE stemmed from a desire to serve at scale, influenced by his 2016 campaign for Bernie Sanders. His contributions during his short tenure included projects like overhauling the user experience of the VA’s AI-powered chatbot and developing various open-source tools. These tools analyzed PDFs for sensitive topics, used large language models to review contracts, and even helped build organizational charts.
Despite his productivity, Lavingia expressed frustration about the lack of approval for projects that could significantly improve veterans’ lives—like streamlining disability claims or automating processing. He lamented that he never got the green light to deploy solutions that could truly benefit Americans while saving taxpayer money. Still, he was permitted to open source much of his work, which included tools for DEI analysis, COVID policies, climate initiatives, and contract review.
Reflecting on DOGE’s internal workings, Lavingia noted a surprising absence of a centralized engineering playbook or knowledge-sharing practices. Engineers often had to start from scratch, indicating a disorganized environment. His time ended abruptly after he discussed his work with a journalist, leading to his access being revoked without warning.
Despite the abrupt exit, Lavingia shared that working closely with the VA revealed that, while slow and meeting-heavy, the government “works” and isn’t as inefficient as expected. His experience highlights the challenge of modernizing large government agencies—balancing the need to reduce waste with the reality of their operational complexity. While Silicon Valley volunteers bring fresh tech expertise, the idea of building government solutions from scratch may not always be the most effective approach.
Lavingia did not respond to further questions, but his candid account offers valuable insight into the realities of government tech initiatives and the complexities of implementing change at scale.
