From Cloud Everything to Local-First: How Heroku Co-Founder Adam Wiggins Rethinks Software Architecture

Adam Wiggins, who co-founded Heroku and authored the influential 12-Factor App manifesto, now leads a different mission: proving that the pendulum swung too far toward cloud-only computing. In a recent podcast conversation, he outlined a vision for local-first architecture—not as a rejection of the cloud, but as a necessary rebalancing act that preserves user agency, offline capability, and low latency while retaining the collaborative benefits of cloud-connected systems.

The Heroku Lesson: Too Much Complexity for Simple Tasks#

Wiggins's shift in thinking emerged from years running Heroku's infrastructure. He witnessed the pain of maintaining mission-critical systems where even seconds of downtime caused real financial harm to customers. The frustration crystallized around a simple question: Why does checking a checkbox on a to-do list require routing through layers of complex, geographically distant cloud infrastructure?

That operational experience, combined with observing what users lost when moving from local files to cloud-only tools, planted the seeds for local-first thinking. Yes, cloud deployment solved real problems—it made getting software to users faster and easier. Yes, real-time collaboration through tools like Google Docs was a genuine leap forward. But something valuable disappeared in that shift: the sense of control, privacy, and agency that came with owning files on your own machine.

Technologies Ready for Prime Time#

The local-first vision isn't theoretical. Conflict-free Replicated Data Types (CRDTs) have matured from academic curiosities into production-grade infrastructure. Tools like Linear demonstrate that CRDTs can power robust collaborative features without relying entirely on a central server, a crucial capability that was unavailable during Heroku's heyday.

Wiggins sees this as opening the door to the next frontier: bringing version control primitives—branching, merging, diffing—beyond software engineering. Imagine a spreadsheet or document where non-technical users could make changes on a branch, diff them visually, and merge them back with the same power that Git gives developers. This would democratize practices currently exclusive to engineers and unlock new possibilities for creative and scientific work.

AI and the Next Productivity Inflection Point#

The integration of artificial intelligence presents another reason to revisit centralized architecture. Rather than routing every task to large, expensive cloud-based language models, Wiggins believes the productivity gains will come from small, high-performance local models handling 80% of routine tasks—with cloud-based LLMs reserved for truly compute-intensive work. This approach would reduce latency, lower costs, and preserve data privacy.