Boris Cherny, who built Claude Code at Anthropic, recently shared advice he received from Anthropic co-founder Ben Mann:

Don’t build for the model of today, build for the model 6 months from now.

When Cherny first developed Claude Code, it wasn’t a great product. The models weren’t capable enough, and he only used the tool for about 10% of his own coding. But Mann pushed him to trust the scaling laws and design for where models would be, not where they were.

Six months later, with the release of Claude 3.5 Sonnet and Opus, the models caught up to the tool’s design. Cherny’s usage jumped to 80-90%.

This is a useful mental model for anyone building AI-powered tools: if you only design for today’s capabilities, your product may be obsolete by the time it ships.