The Year of Autonomous Engineering
Long Live Human Best Practices
“I… I think the robots are writing better, more reliable code than our dev team”
— my CTO,
in my nightmaresin December
The last 3 months were the most exciting and terrifying acceleration of AI coding capabilities Kyle and I have seen since we started Hominy.
Exciting, because it’s now faster and better (not to mention 100-1000x cheaper) for us to produce fully-tested, QA-verified, web-scale software managing agentic coding tools instead of humans.
Terrifying, because as operators and technical leaders we’re now outnumbered 20-to-1 by profoundly capable robots working 24/7 on our most critical systems and business processes.
Think of it like L3 autonomous driving; we can now safely take our eyes off the road most of the time, but god-forbid we fall asleep at the wheel.
What unlocks autonomous engineering
We’ve learned a tremendous amount about steerability and reliability of agentic AI systems building our owner- and tenant-protecting replacement for rental property management (and helping several very large enterprises deploy agentic coding tools themselves).
Beyond navigating the rapidly-evolving ecosystem of models and tools, the most impactful investments you can make to unlock the power of your robots are the same you should have been making all along to unlock your engineers:
- Thoroughly-documented code
- Comprehensive automated tests
- Well-supported modern frameworks
- Rigorously-researched technical specifications
- Production-equivalent development environments
- Data- and expert feedback-driven quality systems
- Resilient infrastructure as code and CI/CD
The good news is that agentic coding tools, paired with well-trained architects, can bootstrap those best practices in weeks, not years.
The bad news is that inexperienced operators can use those same tools just as easily to produce what looks like best practices, and instead send your human devs and robots spinning indefinitely.
2026: The year it becomes real
This is the year hands-off, eyes-off autonomous engineering becomes possible for mature organizations… at least the ones that learn how to hold agentic coding tools and their operators accountable to high-quality outcomes.
Kyle and I have spent the last few years figuring out how to do just that; manage dozens of agentic coding agents working simultaneously and reliably hold them to producing software that exceeds the expectations of product & engineering requirements.
Our battle-hardened in-house tools and practices are finally in a place we can share them, and we’re eager to help others responsibly deploy autonomous engineering. We may even build a product or two to help along the way.
Let’s build together
If your organization would benefit from building software much better and faster (and cheaper), or you want to help us build, use, and teach these tools, we’d love to partner with you.