The minimum viable unit of saleable software
Points and comments are a snapshot, not live.
LLMs lower software build costs but not to zero, preserving a zone of viability for paid products.
Brandur argues that while LLMs make building software cheaper, they don't make it free. He calculates that an engineer earning $200k/year costs ~$96/hour, so building an in-house Jira clone to save $400/month would only be cost-effective if maintenance took under 4 hours/month. Using a hypothetical 2-week build with 2 hours/month maintenance, he estimates 37 months to break even. He defines a "zone of viability" for software: complex enough that LLM rebuild is non-trivial, but priced reasonably enough not to incentivize rebuild. His side project River (Go/Postgres job queue) charges $125/month for up to 20 developers, offering advanced features he believes are non-trivial to replicate.
What commenters are saying
Several commenters underscore that the buy-vs-build decision involves more than raw cost: integration burden, ongoing maintenance, forced updates, supply chain risk, and procurement bureaucracy often dwarf the subscription price. One notes that individual developers often value tinkering over efficiency, citing xkcd. Another observes that enterprises frequently waste far more than the SaaS cost on internal approval processes, creating bizarre inefficiencies. A counterpoint argues that B2B customers are rational and eager to buy good products that free teams to focus on core work; the hardest challenge is marketing, not logic.