Replit Agent 3 launch triggers $1,000/week bills - users report 5-20x monthly cost spikes
Replit launched Agent 3 on September 10, 2025. Within a week, users reported single-week bills hitting $1,000 against previous monthly bills of $180-200. The Register and InfoWorld both covered the backlash.
$1,000+ / week reported for same workload
Launching a more sub-agent-heavy Agent on top of opaque effort billing produced order-of-magnitude bill increases that users could not predict; multiple outlets documented the pattern.
Vendor marketed Agent 3 as more cost-effective while user-observed bills increased 5-20x; the gap between claim and reality is the structural harm.
Even granting more capability per task, the absence of pre-submission cost preview combined with multi-subagent fan-out is a textbook surprise-billing pattern.
Replit did not respond inside the 30-day window. This receipt now reads as uncontested in the public record.
Agent 3 ran 'a lot more under the hood with subagents' for reviewing, planning, security checks, and fixes - each consuming an effort-metered checkpoint under the June 2025 billing model.
Agent 3 ran 'a lot more under the hood with subagents' for reviewing, planning, security checks, and fixes - each consuming an effort-metered checkpoint under the June 2025 billing model. One Register-quoted user reported: 'In the last week alone it charged me $1K since the new agent dropped whereas before it was never more than $180-200 a month for the same effort.' Another user said they 'blew through $70 in one night' at launch. Replit's own marketing called Agent 3 'our most advanced and autonomous Agent yet' and '3x faster and 10x more cost-effective than Computer Use models,' a claim users widely disputed. This event is the direct cause of the November 2025 forced billing migration already catalogued.
Open-source Claude Code / Cursor replacement that lives in the terminal. Pick when you want zero vendor lock-in and any LLM backend.
Permissive (MIT) open-weight model targeting ChatGPT/Claude-class chat and agentic coding without API lock-in - the strongest open option for a self-hosted stack.
ChatGPT's Code Interpreter, but local and against any model.
Near-frontier open-weight model to back a self-hosted ChatGPT/Claude replacement or an agentic coding stack.