The Minimum Viable Unit of Saleable Software
5 hours ago
- #Buy vs Build
- #LLM Software Economics
- #Sustainable Tech Business
- The author discusses transitioning from Stainless to focus on their side project, River, aiming to build a small sustainable business.
- Addresses concerns about AI displacing software by exploring the 'buy vs. build' dilemma in the age of LLMs, using anecdotes like replacing Jira with a custom LLM-built tracker.
- Highlights that LLMs reduce software development costs but not to zero, with ongoing maintenance and human oversight remaining significant expenses.
- Presents calculations showing that for lower-cost software like Jira ($400/month), building an internal solution with LLMs may not be cost-effective due to engineer hours and break-even timelines.
- Contrasts with higher-priced software like Salesforce ($500/seat/month), where building with LLMs might be more justifiable due to higher licensing costs.
- Introduces the 'zone of viability' concept: software priced reasonably and with sufficient novelty to discourage rebuilds by LLMs remains viable for purchase over building.
- Posits a 'minimum viable unit of saleable software' threshold below which building is more cost-effective than buying.
- Applies these principles to River, an open-source job queue project, arguing its Pro version's advanced features and pricing model ($125/month for teams) keep it in the viable zone as a plausible business despite AI advancements.