Analysis for the people who carry the exposure.

Swiss, working between Las Vegas, Los Angeles and Zurich. Her background runs from the US Army through major corporate insurance accounts in London and Hong Kong, then Oracle and Apple, with an art degree and twenty-five years of design work underneath all of it.

She is committed to technology that is ethical, transparent and decentralised — three commitments that are easy to state and awkward to live under, which is why this publication exists in the form it does.

This blog is a governance model

NeuralPulse is not a place where those ideas are described. It is a place where they are run. Every claim made here about how autonomous systems should be governed is one the publication itself operates under, in public, where the failures are visible.

The arrangement is bounded autonomy with a human at the review point. The AI researches, sources and composes. A deterministic gate — rules, not a model — checks the draft before anyone sees it. Nothing approves its own output, including the system that wrote it.

What the editor does

Approves or disapproves, in one sitting each Sunday, on whatever the system has queued. The article is read against its own sources, which arrive alongside it.

There is no editing path in the software. Approval means the piece publishes exactly as drafted, or it does not publish. A rejected piece is re-run from a fresh search rather than patched.

Subjects are sometimes supplied by the editor. The composition is never.

Where a subject comes from someone else, they are credited by name on the piece. What is contributed is attributed as it was contributed — writing as writing, a proposal as a proposal.

Comments

Every comment is read before it appears. The editor screens for profanity and abuse and does not censor.

Disagreeing with an article is not grounds for removal, and neither is disagreeing with the editor. Adversarial comments are welcome and publish unedited, exactly as written — the software has no facility to alter them, so that is a property of the system rather than a promise about anyone's intentions.

The complete list of grounds for rejection is abuse aimed at a person rather than an argument, profanity, spam, and anything unlawful to publish. Those four are enforced in code.

Write to lara@loveconcursall.com.

The pipeline

1. Research. The system searches current sources and assembles a set of claims. Each claim carries the URL and date of the source it came from.
2. Drafting. The article is generated from those claims and nothing else. The engine may state as fact only what the research supplied — it is barred from filling gaps from its own training data, which is where invented dates and figures come from.
3. Automatic checks. A deterministic gate runs before any publication: prohibited framing, links in the body, length and structure. It runs on every piece, however the piece was made.
4. Human approval. The editor reviews the draft against the sourced claims and either approves or rejects it. Rejection discards the piece and starts a fresh search. Nothing publishes without an approval on record.

What the checks cannot do

The automatic gate catches rule breaks. It cannot catch a wrong fact — an incorrect date reads as clean prose and passes every check we can automate. That is precisely why a person reads each piece against its sources before it goes out, and why the sources sit above the draft on the approval screen rather than below it.

Attribution

Articles here are generated by artificial intelligence, using Claude, a model built by Anthropic. The disclosure is made at publication level rather than repeated on each piece: everything published under the current pipeline is machine-written, so a per-article label would be noise rather than information.

The editor approves every article and occasionally supplies the subject. The composition is the AI's throughout — a supplied subject is a starting point, not a draft, and the research and the writing follow the same path either way.

The editor is named on each piece and holds editorial responsibility for publication.

Outside contributions are welcome. What is contributed is attributed as it was contributed, by name: a piece written by a person carries that person's byline, and a subject proposed by someone else is credited to whoever proposed it. The absence of any such credit means the pipeline described above produced the piece end to end.

Corrections

If something here is wrong, write to lara@loveconcursall.com with the article and the claim. Corrections are made to the article and noted at the foot of it. Posts are unpublished rather than deleted, so the record of what was said survives the correction.

If you cite us

Cite the article, the publication, and the date. Note the authorship — it matters for what the citation is worth:

Quote what you need. Do not present it as human-written work.

What we cover

Regulation as it lands, not as it is announced. Market structure and where consolidation leaves the organisations that built on top of it. Provenance, oversight, and the mechanisms that decide whether a system can be governed at all.

One argument per piece. What happened matters less here than what it means for an organisation with something at stake.

Who publishes it

NeuralPulse is a publication of loveconcursall, a technology company working on AI governance infrastructure. That is a conflict worth naming: we write about a field we also build in. Our own products are barred from these pages — not mentioned, not alluded to, and blocked automatically before publication.

How it is written

By an AI system, from sourced claims, approved by a human before publication. The full method is on the How this is made page, including what the automated checks catch and what they cannot.

What we don't do

No paid tiers. No advertising. No cookies, no tracking, no data collected or sold. The subscriber list exists to send you the articles and nothing else.

Contact

lara@loveconcursall.com

LinkedIn · RSS