Independent deployments will determine whether Maritime can economically and reliably run large fleets of isolated, persistent, stateful customer-facing agents in microVMs.
state: expiredheat: lowuncertainty: highknownscott: lowai-infrastructure microvms durable-agentsMaritimeMaria Gorskikh
What is this?
Maritime is a managed deployment platform that claims developers can run separate, persistent AI-agent instances for each customer, using isolated microVMs and scaling from one agent to large fleets; its site markets deployment at $1 per month. The broader snippets support microVMs as a fast, VM-isolated execution model for untrusted, stateful agent workloads, with competing or adjacent offerings from AWS, E2B, Modal, and Northflank. However, the supplied evidence provides no independent production benchmarks for Maritime’s fleet-scale reliability or economics and does not establish Maria Gorskikh’s role.
Why it matters to Scott
Scott already holds this architecture in OpenClaw and the Single-tenant AI appliance: isolated per-customer runtimes with durable external state. The radar also tracks nearly identical validation questions on Machine0 and Fly.io; Maritime currently adds only an unverified $1/month claim, not evidence that would change his architecture or economics.
dev:project.openclawdev:concept.single-tenant-ai-applianceip:framework.long-running-agentsip:concept.sandboxed-executionradar:machine0-persistent-agent-vmsradar:fly-agent-computers-infrastructureradar:concept.agent-infrastructureradar:concept.agent-economics
queries asked of Scott's wikis
- durable agent runtime architecture
- microVM isolation for coding agents
- per-customer agent infrastructure economics
- persistent state and agent memory lifecycle
- multi-tenant sandboxing versus dedicated runtimes
- operating large fleets of long-running agents
Measured heat
no measured readings yet — the hourly heat pass fills this in
How the heat travelled
no chain yet — the hourly chain pass fills this in
Evidence (1) — ⭐ canonical anchor
Interpretation history
2026-08-21T06:27:01Z
The extra comments and marginal engagement add no operational evidence beyond the brief onboarding trial. With no independent deployment, benchmark, or customer validation emerging within the active window, the launch episode has faded while its core economics and reliability claims remain untested.
2026-08-19T05:25:41Z
An independent commenter reports registering and using Maritime successfully, providing lightweight evidence that the product is accessible and its onboarding works. This is only a brief trial, not a production deployment or validation of fleet reliability and $1-per-agent economics.
2026-08-18T23:41:42Z
No independent deployment, benchmark, or customer evidence has appeared; the case remains an unvalidated first-party launch whose economics and fleet reliability are still open questions.
2026-08-18T23:37:33Z
grounded: known/low — Scott already holds this architecture in OpenClaw and the Single-tenant AI appliance: isolated per-customer runtimes with durable external state. The radar also
2026-08-18T23:33:18Z
case created — The founders released a concrete agent-infrastructure platform centered on isolation, persistence, secrets, triggers, and sleep-wake orchestration.
Decision trace
- 08-21 16:27expireThe extra comments and marginal engagement add no operational evidence beyond the brief onboarding trial. With no independent deployment, benchmark, or customer validation emerging within the active w
- 08-21 16:27alert_silentThis is a stale reobservation with no consequential new fact; Scott gains nothing from attention before stronger production-scale evidence appears.
- 08-21 16:27alert_routeThis is a stale reobservation with no consequential new fact; Scott gains nothing from attention before stronger production-scale evidence appears.
- 08-19 15:25repriceAn independent commenter reports registering and using Maritime successfully, providing lightweight evidence that the product is accessible and its onboarding works. This is only a brief trial, not a
- 08-19 15:25alert_silentThe first independent usage report modestly advances the case but does not test persistent workloads, recovery, resource limits, or production-scale economics, so it can wait for normal briefing and s
- 08-19 15:25alert_routeThe first independent usage report modestly advances the case but does not test persistent workloads, recovery, resource limits, or production-scale economics, so it can wait for normal briefing and s
- 08-19 14:21sensor_dirtycomment_update
- 08-19 09:41repriceNo independent deployment, benchmark, or customer evidence has appeared; the case remains an unvalidated first-party launch whose economics and fleet reliability are still open questions.
- 08-19 09:41alert_silentThis reobservation adds no consequential delta beyond the already-assessed launch, so there is nothing new that Scott needs before the next briefing.
- 08-19 09:41alert_routeThis reobservation adds no consequential delta beyond the already-assessed launch, so there is nothing new that Scott needs before the next briefing.
- 08-19 09:38alert_silentMaritime’s first-party launch establishes availability, a three-agent free tier, and advertised $1-per-agent monthly pricing, but provides no deployment evidence validating fleet reliability, security
- 08-19 09:38alert_routeMaritime’s first-party launch establishes availability, a three-agent free tier, and advertised $1-per-agent monthly pricing, but provides no deployment evidence validating fleet reliability, security
- 08-19 09:37groundScott already holds this architecture in OpenClaw and the Single-tenant AI appliance: isolated per-customer runtimes with durable external state. The radar also tracks nearly identical validation ques
- 08-19 09:33createThe founders released a concrete agent-infrastructure platform centered on isolation, persistence, secrets, triggers, and sleep-wake orchestration.