How this site maintains itself
Nobody edits this site by hand. It researches, hosts, and updates itself.
The loop
- Parallel's research API filled the comparison table once, with cited sources for every field.
- Parallel web monitors then watch for change: one per provider re-checks specs and pricing daily, and another looks for new sandbox products.
- The whole site runs in a single Sailbox, a cloud VM that checkpoints itself after a short stretch of quiet. While it sleeps it costs nothing, and its lifetime has no cap: the same VM can host this site for years.
- When a monitor finds something, its webhook lands here and wakes the VM in seconds. The same frozen process picks the request up.
- An opencode agent runs inside the VM on GLM-5.2 through Sail's inference API, edits the data the event supports, and commits. One wake becomes one commit and one entry on the update log.
Visits do the same thing. If this page took a moment to load, you woke the server up.
What it costs
Sleeping is free, and awake time is billed on active usage: the VM reserves no capacity, so the bill follows what it actually uses. A typical update keeps it awake for a few minutes, so hosting plus updates lands around a dollar a month. See Sailbox pricing, Parallel pricing, and Sail inference pricing. The log page shows the estimate per update.
Where the data comes from
Public sources only: official docs, pricing pages, and changelogs, all linked from each product page. The git history records every change.
Run your own
The code is small on purpose: one launcher script, one server file, and an
agent prompt. Clone the repo,
set a Sail API key and a Parallel API key, and run
python launch.py.