About

Why this site exists, and how the work behind it is done

UO Solutions is a one-editor research operation run by UO Solutions, LLC in Miami, Florida. It exists because the gap between what AI research tools promise and what they reliably do is wide, poorly documented, and mostly described by the companies selling the tools.

How it started

UO Solutions did not begin with a launch plan. It began as a working habit: saving the raw output of every AI tool used during filing research, because the interesting part was never the summary itself but where the summary quietly went wrong. That habit became a test library, the test library became a method, and the method is now this site.

Two facts anchor the operation today. It is run from Miami, Florida, by UO Solutions, LLC. And its test library is built on US-listed company filings retrieved from SEC EDGAR, so every published claim traces back to a document the reader can open for free. The first documented test published here compared four widely used AI tools on a single annual report.

What we do and what we do not do

We publish documented tests of AI tools used in stock research, workflow guides for reading filings with AI assistance, and we offer paid consultations on research process. We do not recommend securities, publish price targets, manage money, sell signals, or cover options, crypto, forex, or penny stocks. The last sentence is not a legal formality. It is the editorial boundary that decides which article ideas get accepted and which get thrown out.

The working method

Every article starts from a question a reader can actually act on, not from a keyword. "Can a chatbot summarize a 10-K" is a keyword question. "Where exactly does a chatbot's 10-K summary go wrong, and how do I catch it in under five minutes" is a reader question. The second kind survives our selection process; the first kind does not.

Once a topic is selected, sources are gathered in a fixed order. Primary documents come first: the filing itself, retrieved from EDGAR, plus any directly relevant regulator material. Second, our own reproducible tests: prompts, tool versions, dates, and full outputs are saved to a test library so claims can be re-checked later. Third, academic research on language model behavior with financial text. Vendor documentation is read, but only to record what a company claims about its product, never as evidence that the claim holds.

Drafting separates facts from judgment explicitly. A sentence like "tool X returned the wrong fiscal year in test 3" is a fact with a saved output behind it. A sentence like "we think tool X is still worth using for section location" is judgment, and it is phrased as judgment. When two credible sources conflict, the conflict is shown rather than resolved by preference. When we make an error, and with dated tool versions we will, the correction is made on the page with a note, not silently.

Articles are re-reviewed when the underlying tool changes materially or when a reader reports a discrepancy, whichever comes first. Each page shows its publication date and its last review date. Pages we can no longer stand behind are withdrawn rather than left to rot.

How a test is run

1. Step 1 — Fix the task One document, a small set of questions with verifiable answers, and a written record of what a correct answer looks like before any tool is touched.
2. Step 2 — Run and save Identical prompts across tools. Tool name, version or access date, and full raw output are saved to the test library.
3. Step 3 — Check against the filing Every figure and claim in the output is traced back to the source document. Errors are classified: wrong period, dropped qualifier, invented statement, or correct.
4. Step 4 — Publish with the method The article includes enough of the prompts and error log for a reader to repeat the test on their own setup.

Who writes and who is accountable

Donald Ramirez, Editor. Donald writes every article on this site, runs every test, and answers reader mail personally at donald_ramirez117@gmail.com. His byline is a statement of accountability rather than a credential: the test files behind each article are kept, and readers may ask to see the prompts and outputs behind any published claim.

The operating entity is UO Solutions, LLC, 2875 NE 191st STE 801, Miami, FL 33180. The LLC is the contracting party for all paid consultations and is responsible for the content on this site.

I would rather publish one test a month that a reader can re-run than four summaries a week that read well and prove nothing. That choice costs traffic. It is still the right choice for a site whose only real asset is that its claims can be checked.

Corrections and updates

If you find an error, whether a wrong figure, a broken method, or a tool behavior that no longer reproduces, write to donald_ramirez117@gmail.com with the page URL and what you observed. Confirmed errors are corrected on the page with a dated note. Substantive corrections are also recorded in the update log of the affected article.