The supplied material describes a self-hosted process manager for coding agents that keeps sessions persistent in tmux, enables remote access, and can recover or resume unattended work. The strongest snippet concerns Ark0N/Codeman, which supports several coding agents and adds idle detection, re-prompting, context cycling, startup recovery, and orphan-session discovery. However, the evidence does not directly establish the claimed ybouane/rcman repository or its maintainer’s specific claims, so the apparent relationship between rcman and Codeman remains unclear.
Scott already holds and implements this position in `ip:framework.long-running-agents` and `dev:concept.resumable-agent-job-control-plane`, with concrete recovery and remote-status machinery in `dev:project.proposal`. The radar also tracks closely equivalent process-supervision and remote-control releases, especially `radar:supervice-agent-process-supervisor`; given the unclear rcman/Codeman attribution and lack of independent evidence, this is another implementation example rather than a development that changes his view or build direction.
ip:framework.long-running-agentsdev:concept.resumable-agent-job-control-planedev:project.proposalradar:supervice-agent-process-supervisorradar:operator-parallel-coding-agent-orchestrationradar:concept.long-running-orchestration
queries asked of Scott's wikis
- persistent coding-agent session architecture
- tmux-based agent harnesses and recovery
- unattended coding-agent continuation loops
- remote control for long-running agents
- process supervision versus agent orchestration
- coding-agent crash recovery and state persistence
2026-09-20T13:39:46Z
The latest attachment repeats ColliePWA’s existing announcement rather than adding a new implementation or evidence about rcman. With rcman-specific evidence dormant through the polling ladder and no concrete validation pending, this case has faded rather than been disproved.
2026-09-20T13:22:11Z
evidence attached: hn.story.49775454 — shared external link with case evidence
2026-09-17T14:41:23Z
The Windows build-failure report concerns Rocky Surf, not rcman, so it neither contradicts nor validates this case’s recovery claims. Additional adjacent discussion leaves rcman an unverified implementation example rather than a development that changes Scott’s architecture or priorities.
2026-09-13T09:21:37Z
The new Show HN title adds another remote-agent-control offering but supplies no implementation details or recovery results. It reinforces an already familiar pattern without independently validating rcman or changing its value to Scott.
2026-09-13T09:21:20Z
evidence attached: hn.story.49681398 — Independent coding-agent artifact for running and remotely steering local agents bears directly on persistent, recoverable unattended agent workflows.
2026-09-10T19:38:01Z
Orchestrator, Rocky Surf and Runyte broaden the adjacent implementation ecosystem across remote process control, VM lifecycle management and persistent workspaces, but none validates rcman’s recovery semantics or resolves its attribution. These releases remain references for Scott’s existing architecture rather than evidence of a new recovery guarantee or a reason to change build priorities.
2026-09-10T19:22:41Z
evidence attached: hn.story.49648804 — Persistent terminal sessions and coding-agent workflow support bear directly on recoverable, long-running coding-agent execution.
2026-09-10T19:22:41Z
evidence attached: hn.story.49648776 — Open-source control of cloud VMs for coding agents materially contextualizes the emerging long-running and remotely managed agent-workspace pattern.
2026-09-10T15:26:00Z
evidence attached: hn.story.49643645 — Independent implementation corroborates the developing need for persistent, remotely controllable coding-agent sessions.
2026-09-09T18:32:16Z
ColliePWA adds another independently announced mobile interface for terminal agents, broadening the remote-supervision ecosystem without establishing process recovery or validating rcman. The refreshed discussion leaves rcman’s attribution and recovery behavior unresolved; this remains an implementation reference rather than a change to Scott’s architecture decisions.
2026-09-09T18:23:35Z
evidence attached: hn.story.49630842 — A mobile PWA for reaching persistent terminal agents across machines materially supports the case for remotely controllable long-running coding sessions.
2026-09-08T07:34:40Z
The refreshed discussion raises a useful diagnostic distinction between session persistence and boot-time process restoration, citing environment and startup-dependency failures in an adjacent setup. It supplies neither a confirmed diagnosis for crc nor validation of rcman, leaving the product-specific hypothesis and its value to Scott unchanged.
2026-09-08T04:23:46Z
The new comments add anecdotal durability failures and a builder’s pitch for cross-tool session inventory, not operational validation of crc or rcman. They suggest competing approaches to unattended-work reliability without changing the unresolved rcman-specific hypothesis or its value to Scott.
2026-09-08T02:26:44Z
The crc announcement adds a separate implementation aimed specifically at restoring remote coding sessions after reboot, sharpening the distinction between remote access and process recovery. It reinforces the broader workflow need but does not validate rcman’s recovery behavior or resolve its attribution, and remains an implementation reference for Scott’s existing architecture.
2026-09-08T02:22:02Z
evidence attached: reddit.post.1wabfvi — The independently built crc tool reinforces the open case that persistent, reboot-recoverable coding-agent sessions are becoming a practical workflow requirement.
2026-09-07T18:25:55Z
Possess adds a separate session-browsing and transfer tool to the adjacent ecosystem, but its announcement does not establish recovery semantics or validate rcman. The cluster shows continued builder interest in session management, not a new reason for Scott to change his existing recovery architecture.
2026-09-07T18:22:39Z
evidence attached: hn.story.49601208 — A session-browsing and transfer TUI materially contextualizes recoverability and mobility for persistent coding-agent work.
2026-09-07T11:29:00Z
The refreshed adjacent discussion adds no rcman-specific evidence; speculation about the poster’s authenticity is not a basis for judging the tool. Recovery behavior and product attribution remain unresolved, while remote-supervision alternatives are already accounted for.
2026-09-07T07:34:18Z
The refreshed comments remain discussion of native remote access and sandboxed permission handling, not evidence of rcman’s recovery behavior or differentiation. The product-specific hypothesis remains unvalidated; adjacent workflow examples do not supply independent corroboration.
2026-09-07T05:27:11Z
The refreshed discussion repeats existing remote-access alternatives and permission-wait workarounds, adding no distinct capability or operational validation. It neither corroborates rcman’s recovery claims nor establishes that native remote control supersedes process recovery.
2026-09-07T04:25:57Z
The adjacent tool’s builder now acknowledges that Claude’s existing remote control covers most of the cited supervision need, weakening that account as evidence of an unmet gap. This does not settle rcman’s recovery claims or differentiation; those still lack product-specific validation.
2026-09-07T03:23:09Z
The new account highlights an unattended-work bottleneck—sessions waiting on permission prompts or finishing unnoticed—rather than demonstrating crash recovery. It adds demand-side support for remote supervision, but no validation of rcman or material change to Scott’s existing approach.
2026-09-07T03:22:09Z
evidence attached: reddit.post.1w9g9bz — The described browser-accessible persistent coding session materially supports demand for recoverable and remotely controllable agent processes.
2026-09-06T18:29:11Z
This look adds no substantive evidence: Roost remains a separate implementation of the broader persistence pattern, not validation of rcman’s recovery claims. The rcman/Codeman attribution remains unresolved, leaving this a low-relevance implementation example rather than a reason to change Scott’s agent architecture.
2026-09-04T18:24:56Z
Roost provides an independent implementation of interruption-tolerant remote coding sessions, making the broader workflow pattern less isolated. It does not validate rcman’s specific recovery or remote-control claims, so the product-level hypothesis remains unsettled.
2026-09-04T18:22:29Z
evidence attached: hn.story.49567840 — A concrete released tool for persistent, interruption-resistant remote coding sessions materially bears on recoverable long-running agent workflows.
2026-09-02T18:32:29Z
The reobservation adds no evidence beyond the original first-party release claim, leaving both the rcman/Codeman attribution and operational value unverified. It remains a low-relevance implementation example of a pattern Scott already tracks.
2026-09-02T18:28:17Z
grounded: known/low — Scott already holds and implements this position in `ip:framework.long-running-agents` and `dev:concept.resumable-agent-job-control-plane`, with concrete recove
2026-09-02T18:24:44Z
case created — This is a concrete first-party harness artifact addressing recovery and remote operation of persistent coding-agent processes.