After leaving Stainless, I'm taking over full-time development of River, an open-source job queue for Go and Postgres that I hope can become a sustainable business. When I announced this move, several people questioned whether launching a software company makes sense in an age where LLMs can instantly generate alternatives to commercial products.

The math suggests most small SaaS tools aren't worth rebuilding. Take Jira at $400/month: even with LLM assistance, an engineer earning $200k annually would need to spend just 4 hours monthly on maintenance to break even—a completely unrealistic expectation. The refinement loops required for quality LLM-built software consume significant human time, making the buy-versus-build calculus favor purchasing for most tools.

However, there's a sweet spot for commercial software. River's Pro version charges $125/month for teams up to 20 developers, reserving advanced features like workflows and concurrent job limiting for paid customers. While LLMs could replicate these capabilities, the thoughtful API design and performance optimization we've invested creates enough novelty that rebuilding wouldn't be trivial.

The emerging principle is that software pricing determines its fate in the LLM era. Products priced within reason—where licensing costs stay below the cumulative expense of prompting and maintaining a custom build—remain commercially viable. River sits in this zone, but whether that's enough to sustain a business is something only the coming months will prove.