Estimate details

Both figures refer to the same estimated 1.22 million written tokens, not equal-quality software or completed work.

Human: 220 working days × 8 paid hours = 1,760 paid hours. We assume one active typing hour per day at 60 words per minute: 220 typing hours, with other time available for thinking, review, meetings and breaks. These are illustrative assumptions, not measured productivity.

AI: about 68 minutes to generate that volume at a continuous 300 tokens per second. This uses an advertised maximum, not a measured sustained API rate, and excludes request latency.

Costs: human pay, benefits and employer taxes total about $213,000. The AI estimate is about $440 in short-context API charges, assuming one input token per output token ($60 input + $300 output per million). Tools, additional reasoning, retries, infrastructure and supervision are excluded.

Code token counts use an estimated tokenizer mapping.