Privacy Notice

Alexis Gilbert · Last updated 6 November 2026 · hypothesisnext.app and biotechandbanks.vercel.app

The short version

  • No accounts, no adverts and no tracking cookies. Your settings stay in your browser.
  • Press Release Tool: what you submit is saved to a public archive with a share link. Please do not paste personal or confidential text.
  • Stock Research AI: your questions are sent to outside AI companies.
  • Feedback is saved in a private code repository.
  • Your IP address is handled for hosting, and held in server memory for rate limits.
  • Vercel Web Analytics and Speed Insights are on. Fonts are served from this site, so your browser does not connect to Google Fonts.

1. Who runs this site

This site is the M&A Hunter community dashboard, at hypothesisnext.app and biotechandbanks.vercel.app. It is built and run by Alexis Gilbert in the UK, for The M&A Hunter Substack community. It is not run for profit and has no adverts.

Alexis Gilbert is the data controller under the UK General Data Protection Regulation (UK GDPR) and the Data Protection Act 2018. Contact: alexis.cfsme@gmail.com.

2. What is handled and why

WhatWhere it goesWhyLegal basis
IP address, page requested, time and browser typeVercel, our host, in its request logsTo serve the site and keep it secureLegitimate interests
IP address, for rate limitsServer memory only (section 5)To slow down abuse and protect the sign-inLegitimate interests
Page visits and speed measurementsVercel Web Analytics and Speed Insights (section 7)To see how the site is used and how fast it isLegitimate interests
Press releases you submitAI providers, then the public archive (section 3)To produce and share the analysisConsent: you choose to submit
Stock Research AI questionsAI providers (section 4)To answer your questionConsent: you choose to ask
Feedback you writePrivate GitHub repository (section 6)To read and act on itConsent: you choose to send
Emails you send usThe email account you write toTo replyLegitimate interests

3. Press Release Tool

Treat anything you submit as public. The tool has no sign-in, and we do not record who submitted what. Do not paste personal or confidential text, or a web address that contains it.

What happens when you submit

  • If you give a web address, our server fetches the page. If it cannot read enough from the page, it sends the address to the Jina AI page reader (r.jina.ai), which reads the page for us. Pasted text does not go to Jina.
  • Up to 16,000 characters of the text go to an AI provider for analysis, and to the next provider if that one fails (section 4).

What is saved

  • A text excerpt: the first 400 characters.
  • The web address, if you gave one, and its site name.
  • The AI’s analysis, which may quote your text, and the name of the AI model.
  • The company ticker if one is found, a price and score snapshot, and the date and time.
  • A scrambled fingerprint (a hash) of the address or text, used as the key in the share link.

The full pasted text is not saved. Your IP address and name are not part of the saved entry, unless you put your name in the text.

Who can see it

Saved entries go into a public archive, stored as a file in our private GitHub repository. Anyone can open an entry from its share link, see it in the list of recent analyses, or see it on that company’s ticker page, labelled as member submitted. The site asks search engines not to index its pages, but cannot guarantee that they will not.

The phone version of the tool also keeps a list in your browser, of up to 100 entries. It is saved when you start another analysis or reset the tool, and it can include other people’s entries from the public archive as well as your own. An entry an admin has removed can stay in that list, and on your device, until you clear your browser data.

Limits and removal

  • A public submission may not be saved. Saving stops once the archive holds 180 entries or reaches its size limit, and for up to a day once a server has saved 100 public entries. It also fails if the result is too large or the archive cannot be reached. Only an admin’s own runs can go past 180, up to 400. You still get your result, and the text has still gone to an AI provider, but there is no share link.
  • Nothing is deleted automatically to make room. Entries stay until an admin removes them.
  • An admin can remove an entry. Its share link then stops working and it cannot be run again by the public.
  • Removal does not erase repository history. Earlier versions of the file stay in the private repository’s history. That history is not public.
  • To ask for a removal, email alexis.cfsme@gmail.com with the share link or web address.

4. AI providers

The Stock Research AI sends your question, the ticker and data about that company from the page to one AI provider at a time, moving to the next if one fails. The desktop version also sends your last few messages in that chat. Our site does not save your question or the answer. Please do not include personal data. The Press Release Tool also uses one provider at a time and moves to the next if one fails, so your text can reach more than one of them.

FeatureProviders it can use
Stock Research AI, desktopGoogle Gemini, Groq, OpenRouter
Stock Research AI, phoneGoogle Gemini, Groq, OpenRouter, Mistral
Press Release ToolAnthropic (Claude) first, on a paid key with a spending limit. If that fails, including when the limit is reached: Google Gemini, OpenRouter, Mistral, then Groq
Admin and scheduled jobs: reading M&A Hunter posts, checking tickers and earnings dates, finding company websitesGoogle Gemini, Groq, OpenRouter, Mistral, Anthropic, OpenAI

The admin and scheduled jobs work on posts and company information, not on what visitors type. To read paid M&A Hunter posts, they use the site operator’s own Substack subscriber login cookie. It belongs to the operator, not to visitors, and is never set in your browser. The text of those posts, including paid ones, goes to the AI providers above.

OpenRouter passes requests on to the company hosting the chosen model, and this site asks it for free models, which change over time. Each provider handles what it receives under its own terms and privacy policy. These may let it keep what you send, or use it to improve its services, especially on free plans.

5. Rate limits and your IP address

To slow down abuse, the site counts requests per IP address, using the address Vercel passes to our code. The count is held only in the memory of the server handling your request. Our code does not write it to disk, to the repository or to a log.

FeatureLimit per IP address
Each AI tool (Stock Research AI, Press Release Tool)10 a minute and 60 a day
Feedback form1 a minute
Price requests300 a minute
Admin sign-in5 failed attempts in 15 minutes

These counts are per server, so they are approximate. Your address can stay in memory after its time window ends, until that server restarts or its list gets full. There is no fixed period, because Vercel decides when servers restart.

6. Feedback

The feedback form saves your text (up to 3,000 characters) and the time to a file in our private GitHub repository. On the site, only someone signed in as admin can read it. Anyone with access to the private repository can also read it. It does not save your name or IP address, so please do not include anything you would not want kept. Deleting an entry later does not erase it from the repository’s history.

7. Analytics

Vercel Web Analytics and Vercel Speed Insights are on. Both load from this site’s own address and report to Vercel. According to Vercel’s documentation:

  • Web Analytics records the page address, referring site, approximate location, browser, operating system and device type.
  • Speed Insights records the page address, loading-speed measurements, network speed, browser, operating system, device type and country.
  • Web Analytics uses no third-party cookies. It tells visits apart using a hash made from the request, and discards visitor sessions after 24 hours.
  • Speed Insights is not tied to any individual visitor or IP address.

We see totals, not individuals. Page addresses show which ticker pages are visited. See Vercel’s analytics privacy page and its Speed Insights privacy page. The site has no switch to turn these off; a content blocker in your browser can stop them.

8. Cookies and browser storage

Our code sets two cookies in browsers. Neither is a tracking or advertising cookie.

CookieWhat it doesHow long
prefer_desktopSet only if you tap "Desktop view" on a phone. Keeps you on the desktop layout.1 hour
_adminSet only when the admin signs in. Keeps that session signed in. Ordinary visitors never get it.30 days, or until sign-out

The site also uses your browser’s own storage. These are all the items its code can save:

PurposeItems
Appearancemah_v4_tweaks, mah_theme, mobile_theme
Admin page tab (only in the admin’s browser)admin-tab
Layout, last view and start pagemah_v4_view, mah_v4_selected, mah_v4_start_view, mah_v4_nav_collapsed, mah_v4_nav_expanded, mobile_start_page
Filters and columnsmah_filters, mah_columns, mah_col_order, mah_presets, mah_v5
Watchlist, pinned and hidden tickersmah_v4_pinned, mah_starred, mah_starred_migrated, mobile_watchlist, mobile_hidden
Phone list of press results (section 3)press_archive
Your acceptance of the popup asking you to read these policies: the version and the dateterms_acceptance
Older features: sentiment votes and recent searches. The code remains, and anything saved stays in your browsermah_polls, mah_search_history
This browser session only: whether the phone start page has already opened_mnav
This browser session only, if the site owner turns on the optional access PIN: that you entered itmah_authed

The acceptance record stays in your browser, and no server sees it.

Our code does not upload this storage. When you view your watchlist, your browser does ask our server for prices for those tickers, so the ticker names can appear in Vercel’s request logs. You can clear the storage at any time through your browser’s site data settings.

A watchlist link holds the list after the # in the address, which browsers don’t send to the site and the site’s analytics don’t record, but anyone with the link can read the list, and it stays in browser history and any app it’s sent through.

9. Fonts and other connections from your browser

Fonts are served from this site, so your browser does not connect to Google Fonts. In our tests, loading the main pages made no requests to any other company’s servers. You only reach another site if you follow a link to it, such as Substack, SEC or Yahoo Finance, and that site’s own policy then applies.

10. Other services we use

Our servers contact these, not your browser. Apart from Vercel, GitHub and Jina as described, they receive only requests for public data, such as a ticker or company name, and not your IP address.

ServiceUsed for
VercelHosting, request logs, analytics
GitHubStores site data, the press archive and feedback in a private repository, and runs scheduled data jobs
Jina AI reader (r.jina.ai)Reads a submitted web page when our server cannot. Receives that web address
Yahoo FinanceStored prices, price history, earnings dates
Twelve DataFilling in missing prices, price history
StooqPrice history, as a fallback
Alpha VantageEarnings dates
ClinicalTrials.govClinical trial information
SEC (EDGAR)Company filings and financial data
Parallel SearchUsed only when the admin runs a ticker check. Receives ticker symbols and company names
SubstackPosts and feeds from M&A Hunter and related newsletters. Paid posts are read with the operator’s own subscriber login (section 4)
Google SheetsOptional export of scores that the operator can run by hand from the data scripts. No scheduled job runs it, and it holds no visitor data
News and regulator feedsFierceBiotech, BioPharma Dive, STAT, Endpoints News, FDA press releases

When you submit a web address to the Press Release Tool, that website sees a request from our server, not from you.

11. International transfers

Your data can leave the UK. Vercel, GitHub and most of the AI companies listed in section 4 are based outside the UK, many in the United States, and the site’s server functions have been observed running in a US data centre. Where a provider relies on particular transfer safeguards, its privacy policy explains them.

12. How long things are kept

WhatHow long
Press archive entriesUntil an admin removes them. Older versions stay in repository history
FeedbackNo fixed period and no automatic deletion. Kept until removed from the repository. Older versions stay in repository history
IP addresses in rate-limit memoryUntil that server restarts or its list gets full (section 5)
Vercel request logs and analyticsAs set by Vercel’s own retention periods
AI provider copiesAs set by each provider’s own terms
Browser storage and cookiesUntil you clear them, or as shown in section 8
Emails to usNo fixed period. Kept until deleted from that email account

13. Your rights

Under UK GDPR you can ask to see the personal data we hold about you, correct it, delete it, restrict or object to its use, or receive a copy of it. Where we rely on consent, you can withdraw it. Email alexis.cfsme@gmail.com. We aim to answer within one month. The site has no accounts and its forms do not ask for your name, so please tell us which entry or message you mean. If you email us, we hold your email address and whatever your message contains.

We do not make decisions about you by automated means. AI is used to analyse companies and press releases, not people.

You can complain to the UK regulator, the Information Commissioner’s Office, at ico.org.uk or on 0303 123 1113. We would appreciate the chance to put things right first.

14. Changes

If this notice changes, the date at the top changes too. See also the Terms of Use and the Investment Disclaimer.

Questions: alexis.cfsme@gmail.com