We said MAYBE on August 22, 2026. Not settled — due August 22, 2027.
Read this with the caveat. We tested the engine that produced this verdict against 292 launches whose outcomes we already knew, and could not show it predicted which survived. Some of its data sources were also dead at the time of scoring. The verdict stays up, dated and unedited, because a record you can quietly revise is not a record — but it is worth less than it looked when it was written.
Demand is inferred from adjacent paying products rather than observed for this specific idea: no Reddit chatter, no keywords, and the two nearest Show HN launches flopped at 3 points each. The wedge (agent-native, self-hosted, 1-click, auditable) is real and articulable, but it sits in front of Microsoft- and VC-funded incumbents and targets the least monetizable buyer segment — self-hosters. This is a marginal, mostly SKIP: build it only as an OSS project you'd run anyway, not as a revenue bet.
There is real commercial demand for dev sandboxes — five competitors, several venture-scale, charge money for them — but demand specifically for a *self-hosted agentic* sandbox is thin: zero Reddit posts, no strong keywords, and the two closest Show HN attempts scored 3 points each.
A specific wedge exists (1-click self-hosted, agent-native, audit-transparent sandbox) and the buyer is a HN/GitHub-reachable developer, but the incumbents are funded giants (Microsoft, Replit, Gitpod), the engineering is genuinely hard (container isolation, credential handling, multi-tenancy, state sync), and self-hosting buyers are the least willing to pay.
The one signal that did land — a 79-point HN post on a self-hosted agentic software factory — is exactly the audience that pays for developer infra, and the agent-sandbox category is genuinely new enough that no incumbent has shipped a credible answer yet. If AI coding agents keep becoming the default workflow, secure isolation for untrusted agent code becomes a compliance requirement rather than a nice-to-have, and being early with an OSS+paid-control-plane could compound fast.
What is worth more than this page. The register records what became of 6,266 real launches. No engine has to be right for that to be true.