How This Is Made

Laurenceo publishes research and Laurenceo Intelligence, a continuously updated picture of where AI, and the business and risk environment around it, is heading. This page sets out the principles all of it is held to, how the intelligence system and its forecasts work, which sources are accepted, how errors are corrected and the terms on which everything is offered.

Our Principles

• Non-commercial. Nothing is sponsored, and no one pays to be covered, ranked or forecast.
• Sourced, or not published. Every development links to the public sources it relies on, and those sources must meet the standards on this page.
• Fact and interpretation kept apart. What a source reports and what Laurenceo concludes from it are presented separately and labelled.
• Estimates stated as estimates. A forecast is a probability with a resolution rule, a deadline and a public record. It is never presented as a certainty.
• A permanent record. Editions and forecast histories are preserved as published. Nothing is silently rewritten.
• Open about automation. Laurenceo Intelligence is produced by an automated system, and this page says plainly where AI is used and where it is not.
• Corrected in public. Confirmed errors are fixed, marked and logged, and anyone named in coverage can ask for a correction or a reply.
• Information, not advice. Everything here is offered for general information on the terms set out below.

Who Is Responsible

Laurenceo is a non-commercial research and intelligence organisation based in Bermuda. It publishes research articles on this site and Laurenceo Intelligence at laurenceo.com/intelligence.

Jorai Laurenceo is Laurenceo's Founder & Executive Director and the Editor of Laurenceo Intelligence. He sets the editorial standards on this page and the direction the publication follows, and he is the named author of Laurenceo's research articles.

What Is Published

Research articles are long-form work on AI governance, compliance and operating models, including responses to public consultations. Laurenceo Intelligence is a connected set of live views:

• Overview. Headline readings of AI capability, adoption and risk, what is expected next and what has changed.
• Forecasts. A board of open questions about AI, technology and regulation, each with a probability, an evidence rating and its full history.
• Evidence. Every measured series behind the readings, grouped by theme, with its method.
• Today. The day's most consequential developments from public reporting, each with its sources, and a Bermuda lens where one genuinely applies.
• Earlier briefings. The archive of every daily briefing as it was published.

Research Articles

Research articles are developed with frontier AI models that gather and cross-check primary sources, test the argument against the evidence and produce working drafts. Every article is approved by Jorai Laurenceo before it is published, and carries his name, its publication date and the date it was last reviewed.

How Intelligence Works

Laurenceo Intelligence is produced by a proprietary automated system that combines frontier AI models with rule-based verification and quantitative analysis. It runs every day and works in five stages:

• Research. The system surveys public reporting worldwide and identifies the most consequential developments.
• Verify. Before anything is written, every source is checked by rule-based systems, separately from the AI: that it is real, specific, current and within the source standards on this page.
• Analyse. The verified material is analysed against the open forecast board and long-run trend data.
• Validate. Each edition must pass structural and sourcing checks. An edition that fails is not published.
• Publish and preserve. The edition is saved as a permanent dated record, and the overview, forecasts and evidence are recalculated from it.

The system publishes without edition-by-edition human review. Its safeguards are the standards on this page, automated verification at every stage, a public corrections process and the ability to correct or withdraw anything it publishes. If an edition is late or a day is missed, the site says so.

The Daily Briefing

Each development links to its sources so that any reader can check them. The account of what happened summarises those sources. "Why it matters", the Bermuda lens and Laurenceo's wider reading are interpretation. A Bermuda lens appears only where the sources show a real Bermuda connection.

Automated verification confirms that sources are real, current and specific. It cannot guarantee that every statement is accurate or complete, and individual editions are not reviewed or verified by Jorai Laurenceo or any other person before they publish. Statements about organisations and people reflect what the cited sources reported at the time.

Forecasts

Laurenceo runs a quantitative forecasting system. Every question on the board has a fixed resolution rule and a deadline, and carries a probability that is updated as the evidence changes. Each estimate is informed by AI analysis of verified evidence, by statistical trend models fitted to measured data and, where a comparable market exists, by prediction-market prices.

Every change passes automated controls before it is recorded:

• Bounded. No estimate is ever stated as a certainty.
• Evidenced. A material change needs a verified source and a recorded reason.
• Rate-limited. The size of an automated daily move is capped, so a single report cannot swing a forecast.
• Settled before extreme. Evidence that is still unverified or provisional cannot carry an estimate past very likely or very unlikely.
• Permanent. Every revision is added to an append-only public history, so earlier estimates cannot be quietly rewritten.

Open questions are checked against public reporting on a rolling schedule, most often as their deadline approaches, and are resolved against their stated rule. Laurenceo may also correct or resolve a forecast on editorial review, with the source cited, and that change is recorded in the same history.

Resolved forecasts are scored with proper scoring rules, the Brier score and the log score, and calibration is reported so readers can see whether stated probabilities match outcomes. Accuracy is reported from resolved questions only and is shown as pending until there is a record to report. For measured series, the system compares several growth models and shows the best-supported trend beside Laurenceo's own estimate.

Analysis and Editorial

Alongside the daily briefing, Laurenceo publishes four kinds of original analysis, drawn only from its own sourced coverage and forecast data. Each is listed at Analysis and Editorial.

• The Laurenceo View. The daily editorial: the publication's position on the lead story, produced under standing editorial direction set by Jorai Laurenceo, who can change that direction or withdraw any day's View.
• Forecast Moves. When a forecast moves sharply in a day, a short story reports the change and the evidence behind it.
• Laurenceo Analysis. Every week, a long-form answer to the most important question the week's coverage raises for Bermuda.
• In Numbers. Every month, the figures from Laurenceo's own coverage and forecast board.

These pieces are produced by the same system to the same standards, and are likewise published without piece-by-piece human review. Numbered links in each piece open the story or source behind the point.

Reading the Labels

Three separate things are shown for every forecast, because they answer different questions.

• Probability and likelihood. How likely the outcome is. The label follows fixed bands: Very unlikely at 10 percent or below, Unlikely up to 35, Toss-up up to 65, Likely up to 90, and Very likely above that.
• Evidence strength. Thin, Moderate or Strong, reflecting how much recent, sourced evidence from separate publishers supports the question and whether the measured trends agree.
• Confidence. How much weight the number can bear. It is derived from the evidence alone and never from the probability itself, and an extreme call resting on low confidence is flagged.

Evidence and Readings

The overview and evidence pages are computed directly from stored data. Each tracked measure belongs to a pillar: capability momentum, adoption and investment, or risk pressure. A pillar's reading is stated in plain language and backed by a count of the measures behind it, never an invented index. Measures that are out of date or close to a hard ceiling are shown but not counted. The method for each series is set out on the Evidence page.

Source Standards

Primary sources come first: regulators, courts, government agencies, standards bodies, company announcements and filings, and published research.

Next come established newsrooms that have their own editorial standards and corrections policies.

Laurenceo does not cite partisan or hyperpartisan outlets, personal commentary blogs, investing or crypto influencer sites, newsletter roundups, or aggregators that republish other outlets' reporting.

Every development links to the sources it relies on, and a briefing item is linked to a forecast only when that forecast cites one of the item's own source URLs.

Integrity of the Record

• Preserved as published. Every edition is saved as its own dated record and kept in the archive as it ran, apart from marked corrections.
• Append-only forecasts. Forecast histories only ever grow, and a permanent daily ledger records every change.
• Honest status. If the data is out of date or part of the service is unavailable, the site says so instead of presenting old figures as current.

Where AI Is Used

AI models research, analyse and draft. Rule-based systems verify, validate, score and calculate. People set the standards: Jorai Laurenceo sets the rules the system runs on and approves every research article. The only thing a visitor does that reaches an AI model is typing a question into Ask Laurenceo, as the Privacy Policy explains.

Corrections and Right of Reply

To report an error, or to respond to coverage that names you or your organisation, email Jorai@JoraiLaurenceo.com with the page or edition date and what needs correcting. Every request is reviewed.

Confirmed errors are corrected on the page or edition, marked with a correction note, and recorded in the corrections log with the date and what changed. Published material is not silently rewritten.

Terms of Use

General information only. Laurenceo research, briefings, analysis and forecasts are published for general information. Nothing on this site is legal, regulatory, financial, investment or other professional advice, and nothing is an offer or recommendation to buy or sell any security or asset. For decisions that depend on BMA, PIPA or other requirements, consult a qualified adviser or the relevant regulator.

Forecasts. A forecast is a probability estimate, not a prediction that something will happen. Estimates can change at any time as the evidence changes, and past accuracy does not guarantee future accuracy.

Accuracy. Laurenceo Intelligence is produced by an automated system and may contain errors or omissions. It is provided as it is, with no warranty that it is accurate, complete or current. Check the linked sources before relying on any statement.

Coverage of organisations and people. Coverage summarises what the cited public sources reported at the time of publication. Laurenceo's own interpretation is labelled as such and is opinion, not a statement of fact about any person or organisation. Laurenceo is not responsible for the content of the sites it links to.

Liability. To the fullest extent the law allows, Laurenceo and Jorai Laurenceo accept no liability for any loss arising from the use of this site or from reliance on anything published on it.

Privacy

Details of how the site handles information are in the Privacy Policy.

Image Credits

Photographs on laurenceo.com come from Wikimedia Commons and are used under their free licences. They are cropped, tinted and compressed for the site.

• The Pentagon (aerial) and a US Navy carrier strike group. US Navy photographs, public domain (source, source).
• Two F-22 Raptors over Alaska. US Air Force photograph, public domain (source).
• Pleiades supercomputer. Marco Librero, NASA Ames Research Center, public domain (source).
• North Atlantic Council chamber, NATO headquarters, Brussels. Haim Zach, courtesy of the Spokesperson Unit of the President of Israel, CC BY-SA 3.0 (source). The version on this site is cropped, levelled, perspective-corrected and resized, and shared under the same licence.
• Earth at Night (Black Marble 2016). NASA Earth Observatory, Joshua Stevens, using Suomi NPP VIIRS data, public domain (source).
• Frontier supercomputer, Oak Ridge National Laboratory. OLCF at ORNL, CC BY 2.0 (source). The version on this site is cropped, colour-graded, retouched to remove the logos on the cabinets, and resized.
• Data centre roof. Rsparks3, CC0 (source).
• Hamilton Harbour, Hamilton Yacht Harbour, Crossing Hamilton Harbour and House on Hamilton Harbour, Bermuda. David Broad, CC BY 3.0 (source).
• Sessions House, Hamilton. JoeyBagODonuts, CC BY-SA 3.0 (source). The version on this site is adapted and shared under the same licence.
• Main Reading Room, Library of Congress, and courthouse columns, Scranton. Carol M. Highsmith, public domain (source).
• Low Memorial Library, Columbia University. Magpieturtle, CC BY 4.0 (source).
• Radcliffe Camera, Oxford. Tejvan Pettinger, CC BY 2.0 (source).
• Boston skyline at dusk. 4300streetcar, CC BY 4.0 (source).
• Research Room, New York Public Library. Diliff, CC BY 2.5 (source).
• Suzzallo Reading Room, University of Washington. Guywelch2000, CC BY 4.0 (source).
• Parliament Hill, Ottawa. Asif A. Ali, public domain (source).
• Supreme Court of Canada. Dwong33, CC0 (source).