About — Almanex

Almanex is a periodically refreshed research index. Every report is generated by a named large language model, every claim links to a primary source, and every report is timestamped so readers can see exactly when it was produced. The same corpus serves humans on the web and AI agents through API + MCP — when one capable model has done the work to read a 10-K, listen to a transcript, and produce a structured analysis, that artifact gets indexed and re-served instead of regenerated from scratch.

What we're shipping

Volume I covers public-equity earnings. Each report is the output of an institutional-format research workflow: post-earnings update, 8–12 pages, charts, beat/miss table, scenario analysis. Volume II (prediction markets) is shipping next, then arXiv research and policy/regulation. Paid plans get every new volume as it's released — same shape, same source-grounded structure, applied to a different domain.

We're a small team bootstrapping this. If you want early access to a specific volume, want us to cover a name we don't have yet, or want a private vertical of your own, email hello@almanex.ai.

How reports are produced

Each report is generated by a named large language model — currently claude-opus-4-7 via Anthropic's /equity-research:earnings plugin — following an institutional-format research workflow (post-earnings update, 8–12 pages, charts, beat/miss table, scenario analysis).
Every report carries a generation timestamp, the model identifier, and a list of the primary source documents it cites — earnings releases, SEC filings (10-Q / 10-K), and call transcripts.
Reports are not edited or vetted by a human analyst. We treat them as a structured starting point, not as conclusions.
Sources are linked, and where possible cached at our edge so they remain accessible even if the original document moves.

Why provenance matters

Roughly 30–40% of public web text is now AI-generated or AI-edited. When AI models train on AI output without provenance, quality degrades over time — a phenomenon called model collapse. By attaching the underlying primary sources to every Almanex entry and recording the model + timestamp, we give downstream readers — humans and AIs alike — a way to verify rather than blindly ingest.

For AI agents

A machine-readable index is published at /llms.txt. Each entry includes the ticker, report type, generation timestamp, model used, and a link to the source documents. The Builder and Scale API tiers expose the same corpus through a Model Context Protocol (MCP) server, returning structured report objects with per-claim source attribution — useful as a cached, citation-friendly alternative to re-running an LLM over raw filings on every query.

Disclaimer

Not investment advice. Nothing on this site is investment, financial, legal, or tax advice. Almanex is not a registered investment adviser or broker-dealer, and does not make personalized recommendations or solicit transactions in any security.

AI-generated content. All reports are produced by large language models and may contain factual errors, omissions, hallucinations, or out-of-date information. Ratings, price targets, and scenario analyses are illustrative — they do not represent the considered judgment of a human analyst or a registered firm, and are not a forecast of future performance.

Verify before acting. Always read the linked primary sources and consult a qualified professional before making any investment decision.

No warranty. The site is provided on an "as is" basis without warranty of any kind, express or implied. Almanex disclaims all liability for losses arising from use of this information.

Contact & corrections