Skip to content

Stand up the forensics blog — nothing exists yet, and content is now the primary channel #78

Description

@robbyrob42

Short answer to "is the blog ready to have articles?" — no. None of it exists. There is no content collection, no src/content/, no article route, no RSS feed, and no Article/BlogPosting schema anywhere on the property. The eight routes in the sitemap are the whole site.

That matters more than it did last week. The channel direction set on 2026-09-04 makes targeted outbound plus content the primary play and demotes directory listings to a capped experiment, which reverses the 2026-08-26 ordering that had directories first. Content is now the thing everything else waits on, and it has no floor to stand on.

The blog goes on the forensics host, not Taproot — decided earlier, unchanged. Topical authority accrues per host and the businesses are distinct.

Why this is a real gap, measured

The 2026-09-04 GEO audit put topic authority at 77/100 with a single cluster of 3 pages, and the tool's own recommendation was blunt: "Strongest topic 'Root System' has only 3 pages — add 2 supporting pages to build entity authority." Eight marketing pages is a brochure. Entity authority is built by covering a topic from several angles, and there is currently no vehicle for that.

The audit also flagged no RSS/Atom feed as a discovery gap, which only becomes fixable once there is something to syndicate.

Infrastructure required

  1. Content collectionsrc/content/ with a typed schema, following the pattern the copy deck already sets: content as validated data, with a publish gate. The deck's assertPublishable exists because draft copy reaching a litigator audience is worse than a blocked deploy; articles deserve the same gate.
  2. Routes/writing/ index and /writing/[slug]/. Prefer /writing/ over /blog/: the audience is retaining counsel, and "blog" signals consumer content marketing to that reader. Open to being overruled.
  3. BlogPosting schema in the per-page graph Structured data on every route, and a dateModified derived from git #76 introduced, with author as an @id reference to the Person node on the expert profile. That reference is the whole E-E-A-T mechanism — it attaches an article to a credentialed human rather than to a brand.
  4. datePublished and dateModified — extend scripts/content-modified.mjs to cover article source files. The git-derived mechanism from Structured data on every route, and a dateModified derived from git #76 already does exactly this job; articles are the case it was built for.
  5. RSS feed at /rss.xml via @astrojs/rss, linked in <head> with <link rel="alternate">.
  6. Sitemap and llms.txt — both are generated, so both pick articles up automatically once the collection exists. Verify rather than assume.
  7. The footer link from rootsystem.com, tracked as outbound_to_forensics. Agreed long ago, never built. Today nothing on rootsystem.com links to forensics at all, which is a free internal link from an established host to a new one going unspent.

Article concepts

Ordered by how much of the work is already done. The house rule holds throughout: describe methodology and its limits, state conclusions with the qualifications the discipline supports. Overclaiming is impeachment material, and sophisticated buyers read it as inexperience.

Already written, needs expanding:

  1. The timestamp distribution that was not mass copying. The representative matter, at full length. An opposing expert read clustered timestamps as evidence of bulk copying; the discriminator was the shape of the distribution. This is the strongest single asset on the property — it is a real finding, it is reproducible, and it demonstrates the method rather than asserting it. It currently exists as a summary on /matters.
  2. Where the method comes from: defect metrology to expert testimony. Separating true detections from false ones, characterizing what a measurement actually says. The profile makes this argument in three paragraphs and it can carry a full piece, anchored to the IEEE/SEMI publication.

Case-type explainers, written for attorneys:

  1. What a source-code comparison can and cannot establish. Code provenance, authorship attribution, and the limits of similarity metrics.
  2. What a timestamp actually records. File-system semantics, clock skew, timezone handling, and how metadata is altered by ordinary operations — the failure modes that produce confident wrong conclusions.
  3. Establishing what data trained a model. Provenance for machine-learning systems, and what is and is not recoverable after the fact.
  4. Reconstructing a software failure. Logs, traces and the difference between correlation and causation in an outage.

Admissibility and process:

  1. What the 2023 amendment to FRE 702 changed for technical experts. The proponent must now demonstrate reliability is met more likely than not, and the amendment specifically targets testimony overstating certainty beyond what the methodology supports. Directly relevant, genuinely useful, and links to primary sources.
  2. Frye and Daubert for software testimony. Some state courts still apply general acceptance; what that means for how a methodology is described.
  3. Prompt logs are discoverable work product. Already the strongest FAQ answer on the site and it deserves a full piece. AI disclosure has become a standing vetting axis for retained experts.
  4. What a reproducible expert report looks like. Structure, with a redacted example. The property's whole argument is that a finding should be a query the other side can run; showing the artifact is more persuasive than describing it.

For the retaining attorney:

  1. How to vet a technical expert witness. A checklist, written to be used — including the questions that would disqualify this practice. That is the vetting-first positioning applied to itself, and it is the piece most likely to be cited by someone who is not a client.
  2. What a fixed-scope assessment produces. Overlaps Content depth for retrieval: expand the FAQ and thicken the thin pages #77's FAQ item; the article is the long form of that answer.

Sequencing

Infrastructure first, then #1 and #7 as the launch pair. One is proof of method, the other is proof of currency, and they argue for each other.

Do not publish on a cadence for its own sake. A stale blog is worse than no blog on a property whose argument is rigour — the same reason a stale directory CV is worse than none.

Open questions

  • /writing/ or /blog/?
  • Bylines: articles under Rob's name, under the practice, or per-author as the bench publishes? Affects whether author points at a Person node or the Organization, and the Person node is the stronger signal.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions