We built Missions at Factory, and I wrote about the architecture that I led the design for to make multi-day autonomous coding reliable.
Agents are highly reactive to their context. Every design decision follows from keeping each agent's trajectory focused and directionally consistent.
Introducing model routing to Factory.
Factory Router picks the right model for every task, automatically.
Maintain frontier performance while cutting costs by 25%.
Today, we are excited to announce our $150M Series C led by Khosla Ventures with strong participation from Sequoia Capital, Blackstone, Insight Partners, Evantic Capital, Abstract Ventures, 20VC, NEA, and Mantis VC.
This puts our valuation at $1.5B and will accelerate our investment in research, product, and global go-to-market.
Long live developers.
They don't talk loud. They speak with the work.
Matan and 20 engineers at Factory just raised $150M at a $1.5B valuation, and the industry is still catching up to what they've shipped.
Long context compression: I pushed 40+ million tokens in a single session without starting a new chat. The agent held the thread the whole way. Every lab has since shipped some version of it. Factory did it first and barely mentioned it.
Missions: Droid can ship software across days. The codebase comes out cleaner than it went in (tests passing, edges handled, the kind of work a good engineer does at the end of a long week).
Agent Readiness: A framework for whether your codebase is even ready for agents. If you've wondered why AI coding tools feel uneven across projects, this is the answer. The environment around the agent is usually the bottleneck.
In three years, 20 people are outpacing AI agent innovations from frontier labs with millions in capital plus hundreds of employees.
If you're building with AI, spend a weekend with Droid. It cooooooks!
We built Missions at Factory, and I wrote about the architecture that I led the design for to make multi-day autonomous coding reliable.
Agents are highly reactive to their context. Every design decision follows from keeping each agent's trajectory focused and directionally consistent.
The post explains our design rationale, how missions actually run, and breaks down a real 16.5-hour mission: 185 agent runs, 778M tokens, 89% test coverage.
https://t.co/vCXs7B6MG1