We do one thing: we make investigation legible.
BearSignal publishes forensic research on listed companies — the kind that helps investors avoid losses before they arrive, and find value the market has not finished pricing. Our subscribers are the institutions that carry the risk. They are the only people who pay us.
OUR STORY
Founded by analysts who got tired of writing for the wrong people.
Enough years inside the business teaches you the same lesson on repeat: the research that would have helped most never gets written.
BearSignal started as a simple frustration. Our founding team spent years producing institutional research. The work was rigorous. The methodology was sound. The audience was wrong.
Sell-side research is compromised by structure, not by character. The bank earns underwriting fees from the same companies its analysts cover. The conflict disclosure is printed at the bottom of the page, where it belongs to no one. Honest negative coverage is rare, and it arrives after the damage.
We built BearSignal to fix that.
Fixing it turned out to have two halves. The first was obvious: take nothing from the companies we cover, and never trade what we publish. The only client is the subscriber — the investor who pays us directly to be honest.
The second half is the one nobody else does. We gave up the right to choose our own targets.
We could have answered the problem by hiring twenty more analysts. We are in Silicon Valley, so we built a machine instead. It reads the filings, computes the contradictions, and names the company. Coverage is generated, not nominated. Only after that does a person see the name for the first time.
Today the team is investigators on the ground in the United States and Asia, the Silicon Valley engineers who built the machine, and the independent accountants who teach it what a red flag looks like. One principle holds them together: we work for the investor, never for the issuer.
THE OTHER CONFLICT, THE ONE NOBODY NAMES
Most short-side research is written by people who already hold the position. The publication is the exit, not the finding. We don't take that money either. Neither BearSignal nor anyone working for it holds a position in a company we cover, long or short, before publication, during it, or after.
NO ONE HERE CHOOSES THE TARGETS
This is the part that is hardest to fake and easiest to check. No analyst walks in with a name. The detection engine reads filings and produces features — never verdicts. A separate statistical layer turns those features into contradictions worth pursuing.
That ordering removes the largest single source of bias in adversarial research: the analyst who decided first and reasoned afterward.
GRAY FLAG, THEN RED FLAG
A gray flag is a hard contradiction in filed numbers — profit unsupported by cash, receivables outrunning revenue, a gross margin the business model cannot produce. It is not an allegation of fraud. It is arithmetic that does not close, published with its status visible while the work continues.
A red flag is forensic confirmation.
Most gray flags never become red flags. A contradiction that resolves benignly is a result, and it stays on the record.
THE MACHINE IS THE AUTHOR. THE EXPERTS ARE THE CURRICULUM.
Detection is the output of a trained machine. What trains it is credentialed accountants working through filings line by line, recording their reasoning and not only their conclusion. Where they disagree, the disagreement is kept rather than averaged away. No machine-written reasoning has ever entered that corpus, and none will.
Our intent is that judgment sharpens as real outcomes accumulate. We state what the system is built to do, and separately what it has done.
EVIDENCE IS COLLECTED, NOT COMMISSIONED
Field work is dispatched as location, target, action, deliverable. Field units receive no thesis and no analytical instruction, so nothing they bring back is shaped by what we hoped to find. What they capture is hashed on the device and sealed into write-once storage with a custody record.
COVERAGE
US listings on NYSE and NASDAQ, including ADRs, and Hong Kong listings on HKEX.
These two markets are where the method proved out, not where it stops. The contradictions are read out of filed numbers, and every major market requires companies to file them. The same methodology and the same technical capability extend to other major equity markets as the need arises.
WHAT WE WILL NOT CLAIM
We are not a rating agency, and we do not give investment advice. We publish what the evidence supports and mark the rest unresolved. We will be wrong sometimes. When we are, the record stays up: corrections are appended, never substituted.
WHO WE ARE
BearSignal was founded in 2026 and is based in Silicon Valley.
The machine was built by artificial-intelligence engineers from Silicon Valley itself. The Valley is never short of places to go — they could have gone to work on recommendation systems, on chatbots, on the next model that places an advertisement more accurately. They chose something harder and far less celebrated: pointing artificial intelligence at the part of finance that actually resists it — not predicting a price, but reading numbers that have already been carefully written once. They believe that is where this technology most deserves to be sent, and that nobody had seriously tried it before they did.
What teaches this machine to recognize a red flag is a group of independent accountants, invited one at a time. They have names, and credentials anyone can verify. We invite them precisely because they are independent — the only thing they have to defend is their own name, and in this profession a name is a person's entire practice. What they accumulated over a career of judgment now sits, line by line, in this machine's training record.
Our models are built and trained in the United States, in support of this country's long-term ambitions in artificial intelligence.
The research firms worth remembering were all founded by people who gave up better-paying jobs because they could no longer sit inside the conflict. We intend to be the next one.