Snap CEO Predicts AI Shifts Software Spending

Evan Spiegel said more than two-thirds of new code at Snap is now generated by AI, according to Business Insider. The comments were made on the "Cheeky Pint" podcast, where Spiegel singled out Anthropic's Claude by name and said, "Claude is transforming software development, full stop, at Snap in every part of our organization," Business Insider reports. Spiegel told the podcast he expects companies to "reallocate more resources away from things like software engineering, for example, to distribution in order to grow faster," the article adds. Business Insider notes broader enterprise interest in automating coding; the piece also reports that "it's hard as a brand today to get visibility, to get traction," a challenge Spiegel mentioned. The article records that Snap declined to comment.

Rapid improvements in large language models and specialized codegen models have already reduced the manual effort of routine implementation tasks. Industry-pattern observations show that when teams outsource boilerplate creation and first drafts to models, remaining engineering work concentrates on integration, correctness, security, and system-level testing. For practitioners, this typically increases the importance of robust CI/CD, automated testing, type- and contract-checking tooling, and runtime observability to catch model-induced errors.

Industry context: Reporting like this highlights a shift in where companies allocate scarce engineering hours. Observed patterns in similar transitions indicate budgets and headcount can shift from feature implementation toward product distribution, platform reliability, and developer-experience investments. That does not quantify Snap's hiring or roadmap beyond Spiegel's remarks; Business Insider provides the interview as the primary source for the claim about code generation levels.

Look for follow-on reporting or public engineering blogs that provide measurable signals: internal benchmarks on AI-generated code quality, changes in CI/CD or testing spend, published incident postmortems tied to codegen, and documentation from model vendors about safety and correctness features. Also watch for product metrics showing whether reallocating effort toward distribution materially changes user acquisition or engagement.