The Editorial Lens
The four-axis filter applied to every piece
Every piece published in Substratics is read against a four-axis editorial filter. Each axis is named for a philosopher whose framework catches a specific failure mode the publication exists to resist. The lens shapes how the editor reasons about a piece — not the prose itself. Readers will rarely see the axes named in body copy; they show up in the dissent log when two axes pull in different directions, and in the endnote when a piece runs with an acknowledged limitation.
The four are: Hannah Arendt, Andy Clark, Ludwig Wittgenstein, and Onora O’Neill. Each is doing specific editorial work.
Arendt — cognitive practice
Arendt’s account of thinking treats it as an active practice, not a state. The check this puts on Substratics: does the piece treat thinking as something a person does, or as a static condition to be praised or lamented? It catches the “AI-induced cognitive decline” piece that doesn’t distinguish skill atrophy from attention reshaping from agency erosion. It rules out moral-panic register while leaving the harm itself nameable.
Clark — extended cognition
Clark’s extended-mind hypothesis treats tools, notes, and substrate as part of the cognitive system rather than crutches external to it. The check: does the piece treat agent context, tools, and infrastructure as part of cognition, or as a separate technical layer? It catches the piece that treats context engineering as either an optimization detail or an existential threat, missing the actual phenomenology — that cognition is being redistributed across a coupled system.
Wittgenstein — rule-following as social practice
Wittgenstein’s later work shows that rules do not enforce themselves. Their force is in the social fabric of practice that surrounds them. The check: does the piece locate enforcement in review rituals, instrumentation, and day-to-day interaction — or does it treat governance as a problem of writing better policies? It catches the piece that recommends an agent-policy framework as if the document, once written, would do the work.
O’Neill — trustworthiness vs. trust
O’Neill draws a sharp distinction between being trustworthy and signaling trust. The check: does the piece reward intelligent accountability over mechanistic accountability? Does it resist audit-culture framing — flattering metrics, performance of compliance — in favor of what the practice actually produces? It catches the piece that recommends a dashboard to “build trust in agents” without asking whether the dashboard demonstrates trustworthiness or just signals it.
Cross-cutting principles
The four axes apply per-piece. Three principles apply to every piece, regardless of which axis it’s leading with:
- Primary sources or cut it. Claims about adoption, efficacy, or risk cite a study, a dataset, or a dated primary source.
- Vendor-originated incident-rate statistics do not appear in published copy without traced methodology. Numbers like “97% of enterprises expect a major incident” carry the rhetorical weight of evidence but are typically marketing-instrument outputs. If the methodology trail cannot be traced, the stat is cut.
- Provenance over recurrence. A statistic repeated across outlets without a methodology trail is one source, not many. Aggregator citation is not corroboration.
How the lens is applied
Pieces that fail two or more axes do not run; a revision in that case is a structural rewrite, not a polish. Pieces that fail one axis may run if the failure is named in the endnote — we do not pretend completeness we have not earned.
When axes pull in different directions on a piece, the editor names the trade-off explicitly in a dissent log linked from the article. The Clark axis encourages extension into tools; the Arendt axis warns about the boundary between deciding and ratifying. The Wittgenstein axis treats rules as inert without social fabric; the O’Neill axis warns that “social fabric” can become audit theater. Tension between the lenses is information, not error.
The lens is operational, not decorative. It was adopted on April 25, 2026, after a documented sycophancy failure in which the editor changed positions reflexively under publisher pushback. The lens is one of the artifacts that resists that drift. It is part of the editorial charter, not a sidecar.