The research engine Engine active

Six agents.
One signature.

The system reads everything. Only survivors reach human review. Cheap research labor means more ideas killed early.

Agent activity / session replayrunning▮
UNIVERSEglobal universe refreshed · 4,280 candidates
FILINGparsing 3316.HK FY25 annual report
SKEPTICcandidate killed · customer concentration 61%
CHECKLIST82/82 questions scored · verdict: reject
ANALYSTreport signed · published free · no gates

The pipeline

01

Universe Agent

Output: candidate queue
02

Filing Agent

Output: source pack
03

Extraction Agent

Output: evidence table
04

Skeptic Agent

Output: kill memo
05

Valuation Agent

Output: valuation deck
06

Human Analyst

Output: signed decision
Primary sources

Every number must survive a source check.

Annual reportsreadcash flow, margins, segment notes
Investor decksparsedbacklog, capex, production, guidance
Filingscheckedownership, dilution, related parties
Transcriptsindexedmanagement claims, risk language
Local languagetranslatedforeign small-cap evidence

The 82 questions

Same filter · Every company · No exceptions
ICircle of competence6 questions
IIMoat and competitive advantage7 questions
IIIIndustry structure8 questions
IVEarnings quality and cash flow10 questions
VBalance sheet strength7 questions
VICapital allocation7 questions
VIIManagement and governance9 questions
VIIISpecific and tail risks7 questions
IXValuation and margin of safety7 questions
XFinal decision8 questions
XIPost-investment review6 questions
Bear

What breaks?

Base

What is already visible?

Bull

What is not in the price?

Rejection philosophy

Rejection is not a by-product. It is the product.

+More filings read
−More ideas killed
=Fewer, better reports

See what survived.

Open the archive → Open Quant Lab