TAPE 33
the source of truth is the tape.
TAPE 33 is an invite-only research terminal for US-listed equities, built on a single discipline: every datum has a source, and every claim cites it. Bloomberg costs twenty-four thousand dollars a year. Refinitiv costs more. They serve institutions. This serves the 33rd degree.
▸ WHY
the influencers are paid.
the research is a podcast.
Public-market research has been hollowed out. The Bloomberg terminal is a private utility, gated by price. The discount brokerage research desk has been replaced by a content marketing team. The independent newsletter is a sponsorship vehicle. None of these are calibrated for someone who has to make a sized decision before the close.
TAPE 33 is the answer if you happen to be that someone. It is a cassette built for one user at a time. It does not scale to millions. It scales to the few who are willing to read the tape instead of the headline.
▸ WHO
Solo build. The 13F warehouse, the FINRA pipeline, the SEC concept-tie-break logic that fixes BABA-class ADR market caps, the RAG retrieval with intent-based boosts — all of it one human. Access is granted directly, by the operator, to the 33rd degree.
▸ THE 33RD DEGREE
The 33rd degree is the highest tier of esoteric mastery — the level at which one stops following rules and starts writing them. In freemasonry it is the Inspector General Honorary. In the trading floor it is the desk that does not need permission from compliance to size a trade.
For our purposes: it is the discipline of refusing to repeat a claim that has not been cited. Of treating provenance like a margin call. Of reading the tape instead of the news. TAPE 33 is built for that audience.
▸ HANDSHAKE / FAQ
Q.01What is TAPE 33?▾
TAPE 33 is an invite-only research terminal for US-listed equities. It aggregates SEC EDGAR filings, institutional 13F ownership, congressional trade disclosures, FINRA short interest, biotech catalysts, insider activity, and an AI synthesis layer — every datum cited and clickable to its source. There is no public sign-up.
Q.02Who is it for?▾
Independent investors, biotech catalyst traders, and small allocators who want Bloomberg-class data without the $24,000/year terminal. Specifically: people who refuse to take hot takes seriously and want every claim sourced.
Q.03How is it different from Bloomberg?▾
Bloomberg costs $24k+/year and is owned by a public company. TAPE 33 is self-hosted, free for invitees, source-grounded, and built around the discipline that every datum must be cited. The AI synthesis is a 3-5 sentence read on every security page that links inline to the filing or article that supports each claim.
Q.04Is the data real-time?▾
Prices are real-time IEX via Tiingo Power with Yahoo and Stooq cascading fallbacks. SEC filings appear with a sub-second EDGAR latency. FINRA short interest publishes T+2 settled, twice per month. 13F is 45 days post-quarter-end (statutory). Congressional trades are FMP's hourly free-tier poll.
Q.05Where do you get your data?▾
Ten primary streams — SEC EDGAR, Tiingo, Finnhub, FMP, FINRA, ClinicalTrials.gov, Yahoo Finance, GlobeNewswire + PRNewswire, Motley Fool, OpenAI + Anthropic. The full provenance reference is at /sources, with each provider's license, temporal coverage, and update cadence.
Q.06How do I get access?▾
Invite only. The allowlist is enforced in code (Clerk auth + ALLOWED_EMAILS env var checked on every request, not just at the dashboard layer). There is no application form and no waitlist — access is granted by the operator directly to the 33rd degree.
Q.07Is my data private?▾
Yes. Watchlists, screener saves, and chat history are stored in a per-user-scoped Neon Postgres table keyed on Clerk user ID. The Python sidecar's DuckDB cache holds only ticker-keyed market data (no user data ever). No analytics, no third-party trackers, no advertising.
Q.08What does "33rd degree" mean?▾
The 33rd degree is the highest tier of esoteric mastery — the level at which one stops following rules and starts writing them. TAPE 33 is built for that audience: the trader who treats provenance as discipline, not decoration. The tagline is the doctrine: the source of truth is the tape.
Q.09Can the AI synthesis hallucinate?▾
The synthesis layer is grounded in retrieval-augmented generation (RAG): every claim is conditioned on chunks pulled from filings, transcripts, and articles via OpenAI text-embedding-3-small + DuckDB VSS, then rendered by Anthropic Claude with explicit instructions to cite. The system prompt rejects assertions that aren't grounded in retrieved evidence. If a model wants to fabricate, the prompt is configured to surface a 'no signal' response instead.
Q.10Is this open source?▾
The architecture, gotchas, and operational discipline are documented publicly at the project repository on GitHub (https://github.com/jst213/tape-us). The terminal itself runs against the operator's own data stack — provisioning it for someone else is a manual handshake.