Every new technology spends some of a team’s capacity to cope with surprises

Dan McKinley imagines companies having a small budget of innovation tokens. The number is a metaphor, but the constraint is real: teams can absorb only so many unfamiliar failure modes while also building their actual product.

Mature technology is useful partly because its limitations are understood. Adding a new database or language brings monitoring, debugging, deployment and training obligations, even when it is locally well suited to one feature. The relevant optimization spans the whole organization.

His Etsy example is an activity feed built using an existing shared platform. It required compromises initially, but later accommodated twentyfold growth without dedicated changes to that feature. Shared operational investment benefited it automatically.

McKinley does not ban new tools. He proposes explaining why the existing stack makes a problem prohibitively difficult, considering the organization-wide cost, and planning migration where the new tool replaces an old one. The 2015 product examples have aged; the accounting for attention and maintenance remains the central argument.