Shadok AI claims its open-source scheduler makes unattended recurring Claude Code workflows practical while incurring no usage cost on days when no jobs run.
state: expiredheat: lowuncertainty: mediumknownscott: lowcoding-agents agent-harnesses inference-economicsShadok AI
What is this?
Shadok AI claims to have released an open-source scheduler for running recurring Claude Code workflows unattended in the cloud, including code reviews and security scans. Its distinguishing claim is that a deterministic guard command runs before invoking the model, so quiet days incur no model-usage cost. The supplied web snippets confirm that scheduled Claude Code tasks and GitHub Actions-based unattended execution are established patterns, but they do not independently document Shadok AI’s implementation or verify its zero-cost-on-idle-days claim.
Why it matters to Scott
Known via `Linux router with autonomous AI operations`, where Scott already runs scheduled Claude Code agents behind deterministic health checks, and the Heartbeat Supervisory Program already specifies scheduler-awakened agents with deterministic gates. Shadok AI is another implementation of that pattern; without independent verification or a demonstrated architectural extension, its idle-cost claim does not yet change what Scott would build or argue, while the radar already tracks adjacent scheduled-Claude systems in Cairn Wake and Jaipilot.
dev:project.routerip:framework.heartbeat-supervisory-programradar:cairnwake-persistent-business-agentradar:jaipilot-hosted-claude-maintenance
queries asked of Scott's wikis
- deterministic guards before agent invocation
- event-driven coding agents versus polling
- zero-cost idle agent infrastructure
- unattended Claude Code harnesses
- scheduled agents safety and permission boundaries
- recurring code maintenance automation
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 (2) — ⭐ canonical anchor
Interpretation history
2026-08-30T09:27:13Z
The case has produced no independent validation, adoption, or operational results within its horizon. It remains a minor implementation of a scheduled-agent pattern Scott already uses, with no reason to expect near-term movement.
2026-08-28T08:38:21Z
No new evidence or engagement independently validates adoption, operational reliability, or the claimed economics. The artifact remains a small implementation of a scheduled-agent pattern Scott already uses, so its meaning has not advanced beyond the initial seed.
2026-08-28T08:34:29Z
grounded: known/low — Known via `Linux router with autonomous AI operations`, where Scott already runs scheduled Claude Code agents behind deterministic health checks, and the Heartb
2026-08-28T08:32:48Z
origin walked (codex/luna, conf 0.97): anchor hn.story.49475656 -> echo.github.94ed90cfb3 by shadok-ai-dev
2026-08-28T08:29:39Z
case created — The repository is a concrete reusable artifact for scheduled coding-agent operation rather than a report about one autonomous-agent experiment.
Decision trace
- 08-30 19:27expireThe case has produced no independent validation, adoption, or operational results within its horizon. It remains a minor implementation of a scheduled-agent pattern Scott already uses, with no reason
- 08-30 19:27alert_silentThe staleness trigger adds no consequential evidence, and neither topic-level heat nor the existing artifact changes the case’s meaning for Scott.
- 08-30 19:27alert_routeThe staleness trigger adds no consequential evidence, and neither topic-level heat nor the existing artifact changes the case’s meaning for Scott.
- 08-28 18:38repriceNo new evidence or engagement independently validates adoption, operational reliability, or the claimed economics. The artifact remains a small implementation of a scheduled-agent pattern Scott alread
- 08-28 18:38alert_silentThis reobservation adds no consequential delta; the implementation and idle-token mechanism were already assessed and can wait for evidence of adoption, architectural novelty, or measured operation.
- 08-28 18:38alert_routeThis reobservation adds no consequential delta; the implementation and idle-token mechanism were already assessed and can wait for evidence of adoption, architectural novelty, or measured operation.
- 08-28 18:34alert_silentThe repository establishes an open-source implementation of scheduled Claude Code jobs with a deterministic pre-model guard, but this repeats an architecture Scott already uses and tracks. The zero-to
- 08-28 18:34alert_routeThe repository establishes an open-source implementation of scheduled Claude Code jobs with a deterministic pre-model guard, but this repeats an architecture Scott already uses and tracks. The zero-to
- 08-28 18:34groundKnown via `Linux router with autonomous AI operations`, where Scott already runs scheduled Claude Code agents behind deterministic health checks, and the Heartbeat Supervisory Program already specifie
- 08-28 18:32promote_anchororigin walk conf 0.97
- 08-28 18:29createThe repository is a concrete reusable artifact for scheduled coding-agent operation rather than a report about one autonomous-agent experiment.