A Show HN project called Sunk Cost calculates how long it takes for a local large language model rig to pay for itself versus using API access. The tool estimates local inference speed as memory bandwidth divided by bytes read per token when no measured figure is available, and labels those estimates as such, while API speed only affects the time comparison. The project exposes its assumptions as user-editable inputs.
Where nothing has been measured, local speed is estimated as memory bandwidth ÷ bytes read per token × , and labelled as such. API speed only affects the time comparison.
I’m looking at a , the with of memory.
EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.