AI in Finance
8 research pieces on “AI in Finance” — quoted word-for-word from the filing, with a link back to the source
Our AI Deletes Its Own Sentences — Here Is the Validator That Decides Which, and the Mistakes It Still Lets Through
Our AI Deletes Its Own Sentences — Here Is the Validator That Decides Which, and the Mistakes It Still Lets Through
Boom Leverage AI answers questions from Thai Form 56-1, MD&A and Opportunity Day sources, and every line it writes goes through a citation validator before a reader sees it. This is the model-validation write-up of that control: the six checks it runs on each line, the production incidents of 10–11 October 2026 that shaped it, what happens to a line that fails, when you are charged and when you are not — and, from a test run against the real validator code, the specific errors it does not catch. Written so you know exactly how far to trust an answer and where to keep checking yourself.
Same Baht, Opposite Sign: How AI Summary Turns a Whole-Market Search Into Who Gains, Who Loses and Who Hedged — With Every Row Cited
Same Baht, Opposite Sign: How AI Summary Turns a Whole-Market Search Into Who Gains, Who Loses and Who Hedged — With Every Row Cited
When the baht moves, a market-wide search returns dozens of passages from dozens of companies, and the hard part is no longer finding them — it is sorting them. AI Summary on Boom Leverage Terminal, live since 11 October 2026, reads the result set you already searched and arranges it by impact angle: who gains, who loses, who hedged, who only lists it as a risk. Every row cites the passage it came from, a validator checks the numbers before you see them, and you pay only when the checked answer is shown. This walks through the format, what it costs, a worked example on the stronger baht in 2025–2026 built from real Form 56-1 and MD&A passages, a workflow you can copy, and the traps a model-validation reader would look for first.
Type the Theme, Not the Year: How to Ask AI About Thai Stocks So the Answer Comes From the Filings, Not From Your Question
Type the Theme, Not the Year: How to Ask AI About Thai Stocks So the Answer Comes From the Filings, Not From Your Question
Most people ask a filings AI the way they ask a search engine: a company, a year and a direction in one sentence — 'which companies were hurt by energy costs in 2568?'. On Boom Leverage AI that question quietly works against you. The year in the sentence does not filter anything, and the direction narrows the evidence before the AI has read a page. This is the workflow I use instead: type the theme, set years and scope in the filters, read the cited passage before the summary, and open the page before you rely on it. Three worked rewrites with the real passages each one returns, what every step costs in credits, and the traps that make a confident answer wrong. Figures and passages measured 11 October 2026.
Ask the Whole Thai Market a Question and Get an Answer You Can Audit — Boom Leverage AI, Where Every Line Ends in a Source You Can Open
Ask the Whole Thai Market a Question and Get an Answer You Can Audit — Boom Leverage AI, Where Every Line Ends in a Source You Can Open
Since 11 October 2026 every signed-in Boom Leverage Terminal account can ask a plain-language question across Form 56-1 filings, MD&A, auditor reports and Opportunity Day transcripts, and get a written answer in which every factual line ends in a numbered source — a highlighted PDF page, the original at the SEC, or the minute in the Opp Day recording. Lines that cannot be traced back are cut before you see them and the cut is counted on screen; if nothing survives, the answer is withheld and you are not charged for it. This is what it is, what it deliberately is not, how one question travels from scope to checked answer, what it costs, the five traps I would warn a model-validation committee about, and where it is still weak.
One Company, Every Lens, Six Years Deep: How the Health Check Reads a Thai Stock Before You Do — and Where the AI Summary Is Not Allowed to Go
One Company, Every Lens, Six Years Deep: How the Health Check Reads a Thai Stock Before You Do — and Where the AI Summary Is Not Allowed to Go
Most research tools answer 'which companies talked about X'. The question a holder actually has is the opposite: 'what has this one company disclosed about itself, across every part of its filings, over the years I can see?' This walks through the Health Check on Boom Leverage Terminal — eight fixed clusters from red flags to the auditor's key audit matters, year chips that rewind one cluster's timeline, and the AI summary that writes over it under rules the code enforces: fixed cluster order, a citation on every bullet, no invented sections, the auditor's own words, and no 'fraud' vocabulary the source did not use. Worked example: CP ALL (CPALL), every quote linked to the SEC filing and page. Figures measured 11 October 2026.
You Threw a Form 56-1 Into ChatGPT and Trusted the Summary — the Bank of Thailand Flatly Forbids Financial Institutions From Doing That
You Threw a Form 56-1 Into ChatGPT and Trusted the Summary — the Bank of Thailand Flatly Forbids Financial Institutions From Doing That
Retail investors drag a 56-1 into an LLM, ask whether the financials look good and whether to buy, get a polished summary that cannot point to its own sources, and trade on it. Meanwhile the Bank of Thailand has issued AI risk management guidelines covering both generative and agentic AI, anchored on the four FEAT principles, requiring a human in the loop wherever AI touches loan approval. This piece translates the regulator's standard into four hard rules a retail investor can apply immediately when talking to AI about their own money.