The 'Selfware' Panic Is Missing the Point

Software stocks dropped 15% this week on “selfware” fears. The theory: if Claude Code lets anyone spin up custom tools on demand, why would they pay for pre-built SaaS?

I’ve been building with Claude Code since launch. The panic is both overblown and underblown—just in opposite directions from what the market thinks.

Overblown: Most business software isn’t valuable because it’s hard to code. It’s valuable because of integrations, support, compliance, and the institutional knowledge baked into workflows. A developer can absolutely prompt-engineer a task tracker in an afternoon. They cannot prompt-engineer Salesforce’s ecosystem.

Underblown: The type of software that’s vulnerable isn’t the enterprise stack. It’s the long tail of $10-50/month tools that do one thing adequately. The screenshot annotation app. The simple scheduling tool. The niche analytics dashboard. These are exactly the things that take 4 hours to build with an AI coding assistant.

The actual disruption is the collapse of the market for adequate-but-not-great tools. If you can spin up “good enough” for free, you’ll only pay for “genuinely excellent.”

SaaS companies should be asking: “Is our product good enough that someone wouldn’t just build a replacement in a weekend?” For a lot of them, the honest answer is uncomfortable.