About Mitchell Hashimoto
Mitchell Hashimoto is the co-founder of HashiCorp, the company behind some of the most widely-used infrastructure tools in software: Terraform, Vagrant, Vault, Consul, Nomad, and Packer. After stepping back from day-to-day operations at HashiCorp, he is currently building Ghostty, a modern GPU-accelerated terminal emulator written in Zig, and advising stealth startups in the developer tools space.
Hashimoto has become one of the most thoughtful voices on AI-assisted development, bringing 20 years of infrastructure engineering experience to the question of how developers should actually work with AI agents. His approach — disciplined, pragmatic, and deeply considered — contrasts with both AI hype and AI skepticism.
Career Highlights
- HashiCorp (2012-2023): Co-founder. Built Terraform (infrastructure as code), Vagrant, Vault, Consul, Nomad, Packer. IPO in 2021, acquired by IBM in 2024.
- Ghostty (2023-present): Building a GPU-accelerated terminal emulator in Zig
- Advisor: Working with stealth startups in developer infrastructure
- Open source: One of the most prolific open source maintainers in infrastructure tooling
Notable Positions
On AI Coding Methodology
Hashimoto developed a disciplined workflow centered on always having an agent doing useful work:
“I endeavor to always have an agent doing something at all times. If I’m coding, I want an agent planning. If they’re coding, I want to be reviewing.”
He coined “harness engineering” — building test infrastructure and validation tooling that agents can invoke to self-correct, creating compounding quality improvements.
On AI and Open Source
Managing Ghostty, he faces 2-3 low-quality AI-generated PRs daily and advocates for a shift from default-trust to default-deny in open source:
“AI makes it trivial to create plausible looking but incorrect and low-quality contributions. Before we had default trust. Now it’s default deny.”
On Git’s Future
“This is the first time in 12-15 years that anyone is asking whether Git will be around — without laughing.”
Key Quotes
- “It’s really allowed me to choose what I want to actually think about.” (on AI’s core value)
- “I hesitate to say more productive. There’s an expectation they could do more.” (on AI productivity claims)
- “This is the first time in my 20-year career that so much is on the table for change at one time.” (on the scope of disruption)
Related Reading
- Agentic Coding - The methodology he practices and evolves
- AI Agents - The systems behind his workflow
Video Mentions
New way of writing code with AI agents
Detailed walkthrough of his AI coding workflow: always have an agent running, harness engineering to prevent repeat mistakes, competitive agent runs, and why Git may not survive the agentic era.