How your data moves through this site.

When you generate a plan, the objective you type and any files you attach are sent to my server and on to OpenAI’s API, which writes the plan and its milestone-based pilot estimate in the same response. I instruct OpenAI not to store these requests (their store: false API setting). The generator does not persist your objective, attachments, or plan by default. A separate, explicit builder handoff can store the portable plan; that path is described below. Attachments are read in your browser; only their text is sent, and anything that looks like a credential or secret is scrubbed on the server before the model sees it.

If your browser supports on-device summarisation and you choose it, your connected content is processed on your machine and only the summary (which the page shows you verbatim) is sent.

Nothing is handed to a builder automatically. If you press a send button, the site first stores the fee-free portable plan in a private Vercel Blob object under a randomly generated id. It also stores a separate generation record containing that plan id, the selected builder, status and timestamps, and any result or failure state. The pickup URL is unguessable rather than account-authenticated, so treat it like a private link and do not share it with someone who should not read the plan.

The builder you explicitly choose then receives the portable plan: Anthropic for a Claude builder, OpenAI for the Sol builder, or ScaleMe for the ScaleMe draft. The ScaleMe path can mint an anonymous public product preview that remains unclaimed and unpublished until you take the actions shown there. Builders that are not configured are not offered, and choosing one never sends the plan to the others.

The private Vercel plan and generation records do not currently expire automatically; they remain until I remove them. The chosen builder processes and retains its run under the configured provider account and its own retention terms. Email emesdg@gmail.com with the private pickup link if you want the site-held plan and generation record removed.

Connecting a source grants read-only access to exactly what you pick: repositories chosen on GitHub’s own installation screen, or Drive folders you tick here. Content is read only while a plan is being generated, passed through a secret-scrubbing filter, summarised, sent to OpenAI to ground the plan, and then discarded. It is never stored on my servers.

What I keep is the permission, not the content: a GitHub installation id, or a Google refresh token. It lives encrypted (AES-256-GCM) in a cookie in your own browser, not in any database of mine, and expires after 30 days. Clearing your cookies genuinely disconnects you.

Google Drive access is limited to reading the folders you select (Docs, Sheets, Slides, and plain-text files, three levels deep) for the sole purpose of grounding the plan you asked for. No Drive content is used for anything else, shared with anyone beyond the processing described here, or retained after generation.

Revoke any time: the Disconnect button here; GitHub → Settings → Applications → Installed GitHub Apps; or myaccount.google.com → Data & privacy → Third-party access.

What you submit (your name, email, company, and what you write) is processed by Resend and delivered to my inbox. Resend and the email infrastructure may retain it under their operating and privacy terms. I use it to reply to you. If you tick the separate optional checkbox, that choice is kept with the application and may be used to send occasional product or studio updates; leaving it unticked does not affect the application.

This site uses Vercel Web Analytics to measure visits and a small number of product events. Analytics and server logs receive only operational metadata such as the page or control involved, a pseudonymous random UUID, closed first-touch campaign categories, a coarse referrer category, generation counts, token usage, and estimated model cost. They never receive your objective, plan, interview transcript, files, connected-source content, or application answers.

First-party attribution cookies hold that pseudonymous UUID and closed source, medium, campaign and content categories, a coarse referrer category, a closed public landing-page category, and the arrival day. They expire after 90 days. They never hold a raw referrer hostname, URL, query value, private path, or anything you type. You can remove them sooner by clearing this site’s cookies; the next visit starts a new pseudonymous journey. Separate sealed cookies hold connector grants, free usage counters, and a paid-dossier entitlement when applicable. Your browser’s localStorage holds a courtesy generation counter, and the on-device model you picked. If you provide your own OpenAI API key, it is kept in sessionStorage and disappears when the tab session ends.

• Vercel hosts the site and serves requests. Its Web Analytics product receives the limited journey metadata described above, and private Blob storage holds a plan and its generation record only after you explicitly send it to a builder.
• OpenAI receives strategy-generation and text-interview requests (your objective, attachment text, and source summaries) with API storage disabled. It also receives the portable builder prompt if you explicitly choose the Sol builder; that background run must remain retrievable while it executes.
• Anthropic receives the portable builder prompt only if you explicitly choose an available Claude builder.
• ScaleMe receives the strict portable plan only if you explicitly choose the available ScaleMe draft action.
• Resend delivers application-form email.
• GitHub / Google are called read-only, with your grant, for the sources you selected.

Clear this site’s cookies to remove the 90-day pseudonymous attribution record from your browser. If you used a builder handoff, send the private pickup link so I can remove its plan and generation records. If you applied through the form and want that email deleted, or have any question this page does not answer, write to emesdg@gmail.com.