Summary: The Golden Age of Open Source Applications
Charlie Graham argues AI has triggered a "Cambrian explosion" of open-source applications, not just infrastructure. AI made producing code nearly free, but distribution and operations remain hard — and that mismatch is what pushes developers toward open source.
The supply side. Two paths converge. Some developers launch a paid product, find few buyers, and open-source it rather than let it die. Others skip straight to open source: they see an expensive product, recognize the core is now cheap to reproduce, and build a free version from day one. Graham's examples: voice-to-text dictation (Wispr Flow spawned OpenWhispr, Handy, Muesli, VoiceInk, Ghost Pepper, FreeFlow), personal agents (OpenClaw in early 2026, then Hermes, NanoClaw, NemoClaw), and "vibe coding" tooling (CodeRabbit, Cursor, Lovable as SaaS versus open-source RoboRev, Superpowers, Compound Engineering). Open source pays no immediate financial return but yields credibility, a public audition, a distribution channel, or a future managed-hosting business — plus the emotional payoff of seeing code used rather than rot privately.
The demand side. This is the article's distinctive move. Open source historically cost more than SaaS once you counted install, dependencies, config, deploy, security, and maintenance. AI changes that: you can ask Claude or Codex to install, deploy, secure, and customize a project for you. A company weighing Jira or Trello can take an open-source project 80% of the way there and have an agent finish the last 20% — including Slack/GitHub integration and bespoke workflows. Open source now buys control over data, less vendor lock-in, and customization, not just a lower sticker price.
Where the money goes. Graham expects SaaS to keep its convenience advantage, but predicts a meaningful layer of business around open source — hosting, deployment, security, support, updates, customization, models, databases — and open source as a distribution channel for whatever the free software routes users toward.
The limits. Open source still under-delivers on UI polish, multi-user products, and sustained maintenance; maintainers burn out, and AI-generated PRs threaten to drown the review process. Backend and single-user tools fit best; multi-user SaaS keeps its edge.
The reframe. The old comparison was "$30/month/user SaaS vs. free software that takes me days to configure." The new one is "$300/month/user SaaS vs. free software my agent installed and customized for me." Two mutually reinforcing changes — software becoming cheap to create and cheap to consume — build toward an enormous library of open-source software that agents use as building blocks, with most projects disappearing, a few becoming standards, and some very large businesses forming around making the rest dependable.