Replit expands Cloud Agents and introduces ViBench 2.0
Replit Cloud Agents are now available to Pro and Enterprise users, converting natural-language requests into tested pull requests that engineers approve before release. Replit also introduced ViBench 2.0 for evaluating long-horizon product building; the company says even its best models passed every check in only 49–65% of builds. Replit additionally says central remediation of recently disclosed kernel vulnerabilities reduced developers’ repair work.
The update expands agent-assisted coding with human approval while providing a benchmark that exposes the remaining difficulty of completing sustained software-building tasks.
This source is a digest of several news items (“This Week in Replit 1) The Self-Driving Company…”). Find this story in the original.
#Replit#Replit Cloud Agents#ViBench 2.0#GPT-6.1 Terra#Fireworks AI