Editorial process at Deploy

Deploy tracks the deployment of physical AI. Robots, autonomous vehicles, biometric wearables, and the AI systems that drive them. We track them across companies, models, incidents, regulatory actions, and real-world contexts. This page describes how the work gets made.

Who makes it

One operator runs the publication. Ben Smith, founder. When outside reviewers contribute to a specific narrative or methodology piece, they are named on that article with reviewer attribution. The default is team-of-one. Reviewer attribution is additive when a real collaborator contributes.

What we publish

Deploy operates across three properties.

The registry holds the structured data. Every company, model, deployment, incident, regulatory filing, partnership, acquisition, and patent record lives there.

The news property publishes long-form analysis. Explainers introduce concepts and entities. Framework-in-action narratives apply our methodology to specific real-world events. Methodology essays describe how we think about each axis of the work.

The consumer property publishes pricing and service availability for products consumers can actually purchase.

The three properties cross-reference each other. The registry is the source of truth. The other two surfaces draw from it.

The core discipline

Distinguish what a company claims from what is independently verifiable.

Every entity in the registry, every narrative on the site, every price quote across the consumer property is tagged with the source tier that supports it. Primary government records sit at the top. FDA recall databases. ITC filings. SEC filings. State licensing databases. Maker-official statements sit lower. Aggregator press sits lower still.

We use a 9-tier source-quality rubric to mark every substantive claim. We do not paraphrase aggregator coverage as if it were primary reporting.

When we do not have verification for a field, we leave it empty. Empty fields mean we do not have the data, not that the data does not exist. We would rather publish a thin record than a fabricated one.

Confidence tags

Substantive claims carry an explicit confidence tag in the source layer. We mark whether a figure comes from a regulatory filing, audited financial statement, manufacturer press release, or trade-press aggregation. The same dollar amount means different things depending on its source. We surface that difference rather than collapse it.

When something is unknown

When something is genuinely unknown (a private company's cash runway, a settlement amount, the post-acquisition title of a departed executive), we publish it as unknown rather than guess. Unknown is a valid state. We track it explicitly.

This is the single most important discipline in the publication. Most of what looks like coverage in our space is actually inference dressed up as fact. We try not to do that.

Corrections

When we make a mistake, we correct the page and log the correction on /corrections.

Sources are append-only. Corrections add new sources rather than rewrite the source history. A reader can see what we knew when, and what changed.

If you find an error, write to corrections@deploy.report.

Cross-property coherence

A claim that appears on the registry should appear on the news property in the same form, with the same source attribution, and on the consumer property where relevant. If the registry says a product has a 510(k) clearance, the news property does not say it is "FDA approved" (510(k) clearance is not FDA approval). If the registry has an unverified price, the consumer property does not list a confident price tag.

Internal coherence across surfaces is part of how we test our own work.

What we will not publish

We will not publish a fabricated quote. We will not invent a source. We will not paraphrase a single trade-press article into language that suggests we have independent verification. We will not publish a price we cannot verify. We will not attribute a claim to a regulatory body that did not make it.

If we cannot meet the standard on a specific claim, we leave it out or label it explicitly as the claim of a specific party.

Contact

editorial@deploy.report for general editorial questions.

corrections@deploy.report to submit a correction.

Last updated: 2026-06-05