Show HN: Geiger – See every AI agent on your machine and what it can touch

TEXT START: A Geiger counter for AI agents.

The Dissection

This is an exposure-inventory tool disguised as a simple CLI. It converts invisible agent proliferation into legible machine risk: execution rights, filesystem reach, credentials, network access, provenance, and configuration drift.

Its deeper function is transition management. It does not challenge agent deployment; it makes deployment safer, auditable, and easier to institutionalize. The product identifies the machinery of automation after it has entered the workplace, then offers a governance layer around it. The radioactive material remains installed. The scanner merely labels the contamination.

The text is unusually candid about scope: it inventories configuration rather than behavior, misses nonstandard locations, cannot establish maliciousness, and depends on detector maintenance. That honesty makes the pitch more credible, not less strategically limited.

The Core Fallacy

The central error, under Discontinuity Thesis mechanics, is treating the agent transition primarily as an exposure-management problem. Better inventories, policy wrappers, baselines, and CI alarms may reduce credential leakage and accidental overreach. They do not preserve the mass employment → wage → consumption circuit.

Geiger makes P1 more operationally tolerable and P2 more administratively manageable. It does nothing against P3: the majority losing access to economically necessary labor. A machine can be perfectly inventoried while the humans it replaces become economically unnecessary.

The tool is therefore not a defense against discontinuity. It is infrastructure for governing the discontinuity once it is already underway.

Hidden Assumptions

• Administrators possess enough authority to remove unknown agents, rotate credentials, and enforce policy.
• Known configuration paths approximate the real attack surface closely enough to support decisions.
• A reviewed baseline remains trustworthy even as the ecosystem changes weekly.
• Detecting capability is sufficient to control capability; the report cannot establish what software actually does.
• Policy wrappers will be correctly configured, resistant to bypass, and maintained as agents evolve.
• Organizations will retain stable control over endpoints, identities, repositories, and credentials.
• The surrounding agent ecosystem will continue expanding, creating durable demand for scanners and governance products.
• The free scanner can function as a distribution channel for the commercial policy product named in the text. That is an inference from the stated relationship, not a demonstrated fact.

Social Function

Primary classification: transition management and partial truth.

Secondary classification: prestige signaling and ideological anesthetic.

The text tells the truth about a real and growing control surface, then packages that truth into an actionable ritual: scan, baseline, diff, remediate, enforce. This is useful operationally, but it also shrinks a civilization-scale labor displacement problem into endpoint hygiene.

It is not pure copium. The scanner can expose credentials, unknown installations, and excessive permissions. But its usefulness is precisely what makes it part of the transition apparatus. It helps institutions absorb AI agents without confronting the economic consequence of their success.

The Verdict

Geiger is a legitimate narrow instrument for visibility and drift detection, but its systemic role is hospice care for the old order. It audits the permissions of the machinery that is eroding human productive necessity. The scanner can tell you what the agents can touch. It cannot make human labor necessary again.