OpenBrief AI: Open-Source Intelligence Briefs for WordPress

OpenBrief AI is an open-source WordPress plugin that turns guided web research into structured, human-reviewed intelligence briefs.

Research rarely fails because information is unavailable. It fails because relevant information is fragmented, duplicated, contradictory, and difficult to turn into a coherent account of what actually changed.

A single development may appear across dozens of publications. Early reports may later be corrected. Several articles may cite the same original source while appearing to offer independent confirmation. Important evidence can sit beside speculation, routine commentary, recycled news, or material published outside the period being studied.

Generative AI makes it easier to search and summarize this information. It does not automatically make the resulting analysis reliable.

At Hormuz Group, we encountered this problem while developing research workflows for market intelligence. Existing tools could search the web, summarize articles, generate reports, or manage editorial content. Few combined those functions into a single workflow that preserved human judgment throughout the process.

That led us to build OpenBrief AI.

More than an AI summarizer

OpenBrief AI is not designed to collect a large number of links and compress them into a longer piece of generated text.

Its workflow is closer to a lightweight intelligence desk:

Research framework
→ Guided web discovery
→ Evidence collection
→ Deduplication
→ Event extraction
→ Human review
→ Scenario development
→ Editorial publication

The distinction matters.

A conventional summarizer treats every article as a separate input. OpenBrief AI attempts to identify the underlying events represented by those articles. Multiple reports about the same development can therefore be reviewed as evidence for one evolving story rather than reproduced as separate news items.

This is particularly useful when a situation changes throughout the reporting period. An initial announcement, a subsequent denial, and a later implementation should not necessarily appear as three independent developments. They may be different stages of the same event.

The objective is not to reproduce the information environment. It is to reduce it into a smaller, clearer account of what matters.

Guided discovery rather than a fixed source list

OpenBrief AI does not require users to maintain a permanent list of RSS feeds or predefined publications.

Instead, users define a research framework:

• the subject being monitored;
• the reporting period;
• the languages to search;
• the main research tracks;
• the forms of evidence the system should seek;
• the domains or source types that should be excluded;
• and the limits governing evidence volume and publisher concentration.

The system then searches for relevant material within that framework.

This approach allows the source base to change with the subject and the week. A regulatory development may require official documents and legal commentary. An energy disruption may require industry publications, company statements, infrastructure reporting, and market analysis. A fixed list of general news outlets would not provide the same coverage in both cases.

The model is allowed to discover sources, but not to treat every discovered page as valid evidence.

Search-result snippets are not accepted as evidence. The original page must be retrieved and inspected. Publication dates and event dates are handled separately. Repeated publication does not count as independent corroboration. Material claims can be traced back to primary evidence or genuinely separate source chains.

The governing principle is simple:

Search discovers. Evidence rules decide.

Human review is part of the architecture

AI-assisted research often presents human review as an optional step added after generation. OpenBrief AI treats it as part of the system itself.

Researchers can review extracted events, reject irrelevant or duplicated items, retain leading indicators, examine forecasts, and assess the quality of the generated report before publication.

Quality checks cover issues such as:

• source adequacy;
• numerical accuracy;
• date consistency;
• unsupported causal claims;
• excessive certainty;
• and unresolved disagreement between sources.

The software can flag these issues, but it does not claim authority over the final editorial decision.

A reviewer may decide that the report needs revision. A warning may also be overridden when the reviewer has stronger evidence or contextual knowledge, with the decision recorded for later audit.

This reflects the philosophy behind the project:

AI should reduce mechanical research work without replacing editorial responsibility.

From events to a readable brief

One of the most difficult parts of intelligence production is not finding information. It is deciding what deserves to survive into the final document.

OpenBrief AI separates evidence processing from editorial presentation.

The final output is designed as a short, structured article rather than a database dump. It can include:

• a concise analytical introduction;
• a limited number of major narratives;
• an explanation of why each narrative matters;
• signals to monitor;
• a thirty-day scenario outlook;
• a concluding assessment;
• and a compact evidence section.

Sources remain available for traceability, but they do not need to interrupt every paragraph. Repeated links and syndicated copies can be consolidated into a smaller evidence ledger at the end of the report.

The same reviewed material can be delivered as:

• a WordPress article;
• an HTML email;
• or a PDF brief.

Scenario analysis without pretending to predict the future

OpenBrief AI can generate several thirty-day scenarios, normally including:

• a base case;
• an upside case;
• a downside case;
• and, where useful, a lower-probability shock scenario.

Each scenario can contain an estimated probability, supporting evidence, assumptions, confirmation signals, and invalidation conditions.

This is not intended to turn uncertain events into confident predictions.

The purpose is to make uncertainty explicit. A good scenario should explain not only what may happen, but what evidence would make that path more or less plausible.

Forecasts can also be retained for later evaluation. This allows teams to compare past expectations with actual outcomes rather than allowing old predictions to disappear once their horizon has passed.

Built for more than one field

The original research problem emerged from market-intelligence work, but the underlying workflow is not specific to one country or industry.

OpenBrief AI can be configured for subjects such as:

• competitor monitoring;
• technology and AI developments;
• energy markets;
• policy and regulatory change;
• supply-chain risk;
• regional security;
• public-sector monitoring;
• scientific or industry news;
• and executive weekly briefings.

Research tracks, coverage expectations, report language, model selection, evidence limits, and editorial rules can all be configured inside WordPress.

The plugin is intended for research teams, analysts, specialist publications, consultancies, industry associations, policy organizations, and other groups that need a repeatable process for turning open-web information into reviewed briefings.

Why we made it open source

We initially built the system to solve an internal problem. We decided to release its general-purpose core because the problem is not unique to Hormuz Group.

Research teams across many fields face the same structural issues:

• too many sources;
• too little source independence;
• duplicated reporting;
• weak traceability;
• generated text without editorial control;
• and fragmented tools that do not fit existing publication workflows.

Making the project open source allows developers and research teams to inspect the workflow, adapt it to different domains, identify weaknesses, and contribute improvements.

It also makes the project less dependent on a closed hosted platform. Users can run the plugin in their own WordPress environment and retain control over their settings, research records, reviews, and reports.

OpenBrief AI is released under the GPL-2.0-or-later license.

Current status

OpenBrief AI is currently a public beta.

It should be tested on a staging site before production use. Running it requires a compatible WordPress environment, PHP 8.1 or later, and an OpenAI API key. Web research and model processing create API costs, which vary according to the number and length of evidence pages processed.

The software does not eliminate the need for domain knowledge. It cannot guarantee that public information is complete, that sources are truthful, or that generated scenarios will prove correct.

Its role is narrower and more practical: to make the research process more structured, reviewable, and efficient.

Explore the project

The source code, documentation, issue tracker, and beta release are available on GitHub:

Explore OpenBrief AI on GitHub

Researchers and developers can test the plugin, report issues, propose improvements, or adapt its research framework to their own fields.

OpenBrief AI is not intended to replace analysts or editors. It is designed to reduce the repetitive work surrounding research so that human attention can remain focused on interpretation, uncertainty, and decisions.