Case study
Supertechstack
A source-backed research platform for investors. It ingests documents, news and SEC EDGAR filings, then turns them into a testable investment thesis with assumptions, risks, and an AI challenger that argues back using only cited evidence.
Overview
Supertechstack is a source-backed research platform for investors. It ingests multi-format sources, including SEC EDGAR filings pulled directly from the regulator, then chunks, embeds and retrieves them to produce evidence-backed answers with citations, saved reports, and a stored trace of every retrieval and generation.
On top of that retrieval layer sits a decision record. An investor states a thesis with entry, target, exit and invalidation conditions; records each assumption alongside its source, confidence and what would falsify it; and maintains a risk register scored separately for likelihood, impact and evidence quality rather than collapsed into a single opaque number. An AI challenger then argues against the thesis using only retrieved evidence, and every review is saved so reasoning can be compared over time.
It is deliberately not a buy or sell signal generator. The aim is reasoning that can be audited.
The underlying system is domain-agnostic: the same workspaces, ingestion and retrieval handle any research topic. Investment research is the first workflow being taken to paid users because the review is recurring and the cost of missing something is concrete.
Roadmap
The next phase builds the portfolio layer around the thesis workflow: a concentration summary derived from user-entered holdings, change comparison between successive thesis challenges so reasoning can be diffed over time, and transparent, versioned valuation scenarios.
After that comes material-change monitoring across company, industry and macro sources, followed by a recurring portfolio review with a decision journal. Monitoring configuration already exists and runs can be triggered manually, but no scheduler executes them yet, so it is not described as automatic until one does.
Broader capability built on the same foundation, including team review, publishing, and research templates for other domains, is intentionally out of scope until the individual investment-research workflow has been validated with paid users.