5-Step Framework for Scaling Up Your Coding Agents — Kiet Ho, Superset
Learn how to scale to 100 coding agents in parallel by treating software development as a manufacturing process.
Kiet Ho from Superset presents a framework for running coding agents at scale, borrowed from lean manufacturing and Toyota's production system. He maps software development to a five-step process — trigger, plan, code, review, merge — and argues that throughput is bounded by your slowest step, which, for most teams, is review.
Kiet shares the patterns his customers use to add capacity at each step: Sentry-triggered cron jobs that auto-create PRs, well-documented codebases that give agents better plans, worktrees and database branching for parallel coding, agent-driven review with tools like CodeRabbit, and canary branches that let you merge recklessly without breaking production. The big idea, drawn from Eliyahu Goldratt's book The Goal: define your unit of throughput, find the bottleneck, and lift capacity in every node so the system scales together.
Superset is a desktop IDE designed for running dozens of AI coding agents in parallel. Each agent runs in its own isolated Git worktree, eliminating merge conflicts and allowing developers to review and merge their work on their own schedule. The platform is agent-agnostic — it works with Claude Code, OpenCode, Codex, Gemini, Cursor Agent, and any CLI-based coding tool — and is currently free to use. Superset is part of YC's X26 batch and is used by developers at Google, Amazon, Cloudflare, and Vercel. Kiet Ho is co-founder of Superset, where he focuses on building infrastructure for parallel agent coding. He previously co-founded Onlook in the YC W25 batch and was an engineer at Amazon and ServiceNow.
Recorded at the TypeScript AI Demo Day in San Francisco, April 2026, hosted by Mastra.
KIET HO
/ kiet-ho
https://superset.sh/team/kiet
SUPERSET
https://www.superset.sh
📚 MASTRA RESOURCES
https://mastra.ai
https://x.com/mastra_ai
https://mastra.ai/community/discord
https://github.com/mastra-ai
https://mastra.ai/course
https://mastra.ai/books/principles-of...
https://mastra.ai/books/patterns-of-b...
WHAT IS MASTRA?
Mastra is an open-source TypeScript framework designed for building and shipping AI-powered applications and agents with minimal friction. It supports the full lifecycle of agent development — from prototype to production.
0:00 Introduction
0:34 More tokens, but where to spend them?
1:13 Applying lean manufacturing to software development
3:34 Where your throughput actually breaks down
4:13 Step 1: Triggers
4:40 Step 2: Planning
5:14 Step 3: Coding
5:45 Step 4: Review
6:32 Step 5: Merge
7:14 Wrap-up
7:46 Q&A: agent reviews vs human reviews