Replit CEO Amjad Masad said on stage at TechCrunch Disrupt that the AI coding platform now uses OpenAI models as the default for Replit Agent. He cited cost, results on internal and public benchmarks, and the partnership with OpenAI. He later clarified that customers can still choose Claude models. Replit had until recently been one of Anthropic's top five customers, while people familiar with the relationship insist the partnership is still in place.

What it means: Changing a default looks like a technical detail, but in practice it decides where most traffic and spending go. Most users never touch their settings, so Replit's decision could noticeably cut Anthropic's revenue from one of its key customers. The episode also shows how fluid the model market is: platforms that relied on one provider yesterday replace it today, using cost and performance criteria that shift every few months. For Greek software companies building products on large language models, the message is to design for switching. An abstraction layer that lets you change models without rewriting code, your own test suites and per-task cost monitoring all reduce dependence on a single vendor. Masad's remark that enterprise contracts now make up a larger share of revenue matters too. Businesses buying tools like these should ask which model runs behind the service, where their data is processed and what changes when the vendor switches its default.

Read the original on TechCrunch

Related service: Custom software

Related stories