About
Archit Dwivedi
Software that thinks. AI agents that do real work.

I build software that thinks. Specifically, AI agents — the kind that reach across APIs, tools, and your inbox to do real work, not just answer questions.
By day, I work with teams at Gale Partners on shipping AI into production — the awkward part where research-grade models meet real users, real budgets, and real reliability constraints. By night and on weekends, this blog is where I write about what I’m learning along the way.
What I write about
Building agents
Step-by-step walkthroughs of agents I’ve actually shipped — trip planners, sports bots, project managers, news digesters. The plumbing, not the demo.
LLMs in plain English
Mental models for people who want to stop nodding along when someone says “context window.” Diagrams over jargon.
The tooling
Claude Code, OpenCode, Codex CLI, Cursor, the new MCP world. What’s worth using, what isn’t, and why.
Where to find me
If you’re working on something interesting in agents, evals, or developer tooling — I’d genuinely like to hear about it.