Forensic AI for Thailand's capital markets
Not an AI summariser. We open the file on every listed Thai company — management's own discussion, the risk factors, the auditor's report, and what was said at Opportunity Day — search it across companies, compare it across years, and point to where the words and the numbers don't match. No statement reaches a user before it has been proved against the source document.
918
listed companies in the corpus
512,593
findings on the axes we publish
98.86%
verified back to the source filing
5,834
withheld — never shown as a verified result
Measured over Results & financial position, Risk factors, Business & competition, Governance and Auditor opinion (FY2021 onward) · as of 20 Nov 2026 · See these figures live on our own instrumentation · Read the platform due-diligence dossier
The earnings number is free now. The edge moved into the prose.
What everyone already has
- Net profit · financial ratios · price history
- Available in milliseconds, free, and identical for everyone who looks
Data everybody holds equally confers an edge on nobody.
What still has to be read
- How management explained the result — and the quarter the explanation changed
- Which risk factor appeared for the first time this year
- What the auditor chose to write up as a Key Audit Matter
- The related-party transactions and legal disputes a company is obliged to disclose about itself
All of it is prose, not numbers — which is why nobody has ever extracted it and laid the whole market side by side.
Profit is an accounting opinion. Management's discussion is a statement somebody signed their name to. Our job is to make every signed statement in the market searchable in one pass — so you can see who raised an issue first, and the exact filing where someone's story changed.
Money that leaves a company has to surface somewhere, and Thai SEC rules oblige the company to write the record itself. Our job is to make those lines findable, instead of leaving them buried on page 174 of a 200-page document nobody finishes.
Read the case study: 216 disclosures the companies wrote themselvesThree rules every line in the system has to clear
Strict Source Verification
No line reaches the screen unproved
HowClose
Search results are not written by a model. We pull the real text out of the document and force it to match, word for word, against the source file before it renders. A line that cannot be matched is withheld from the results, and the count and the reason are reported to you. Where Boom Leverage AI does write a summary, every sentence has to cite a source the search returned, and it is checked against that source before you see it.
506,759 of 512,593 findings verified (98.86%) · 5,834 withheld · Results & financial position, Risk factors, Business & competition, Governance and Auditor opinion, as of 20 Nov 2026
Symptom-First Semantic Search
Search the symptom, not the keyword
HowClose
Type the symptom you suspect — “receivables ageing abnormally” — and the engine finds the companies that described the same thing in an entirely different vocabulary, and in the other language. The embedding model is the same multilingual one at index time and at query time, so an English query lands inside a Thai paragraph, and the reverse.
One query across 918 companies, from FY2021 onward
Bank-Grade Risk Lens
Designed by someone who had to sign off on models
HowClose
What counts as a signal here was not guessed from what investors might like. It comes from the Model Risk Management framework that governs credit-risk models inside a bank — the same discipline whose first rule is that an output you cannot explain the provenance of must not be used to make a decision.
The same discipline used to own and validate a production monthly TFRS 9 ECL dashboard at a Thai bank
Founder
Varanchai Yingkhamnueng (Boom) — former Senior Data Scientist / Model Risk Manager in Thai banking. Building an NLP early-warning system to catch credit-risk signals in prose, and owning and validating production TFRS 9 ECL models, is where this platform's logic for separating signal from noise actually comes from.
Analysts and deep-research investors
Search the whole market's One Reports by meaning, with the source filing linked on every result. Free to start, no card.
Try the Terminal freeInstitutions, quant desks, corporate research
Multiple seats · Excel/CSV export, 100 rows per search · API into your internal systems · onboarding run with your team.
Talk about desk deploymentBoom Leverage is a search and analysis tool over public disclosure documents, built to accelerate research. It is not a recommendation to buy or sell any security and it does not forecast prices. The decision is yours, and the source document is one click away — check it every time.