"The company promised to move on from that mess, and on September 10 launched Agent 3, a coding helper that it says offers developers an easier way to build and test apps. Feedback on the new service, which Replit billed as "our most advanced and autonomous Agent yet" and "3x faster and 10x more cost-effective than Computer Use models", has been mixed, with the main complaint being that certain tasks take longer, and involve more checkpoints, so therefore cost surprisingly more."
"I think they're running a lot more under the hood with subagents which probably costs them more, but on older code where you're sort of 'editing as you go' and then having it reviewing old parts of your codebase (especially with very large files), it seems to charge a lot more than just asking for new app builds - it often calls many sub agents to review the code, plan the code, check for security, execute, then fix its issues and review thousands of lines -"
Replit launched Agent 3 on September 10 as a coding helper and billed it as an advanced, faster, and more cost-effective agent. Users reported mixed feedback, with many saying certain tasks take longer, involve more checkpoints, and unexpectedly cost more—particularly when editing pre-existing codebases. Several users described the agent invoking many subagents to review, plan, check security, execute, and fix code, creating frequent small charges that accumulate. Some users reported large weekly bills and a Reddit megathread collected multiple complaints about rapidly rising invoices and higher-than-expected costs for routine operations.
Read at Theregister
Unable to calculate read time
Collection
[
|
...
]