Wildcard Scholarship
We store exactly what you type into the form, plus basic technical context (browser, language, which page you applied from). Applications are reviewed by AI against a published rubric, with human oversight of flagged and sampled cases. Final awards among qualified applicants are decided by a publicly seeded lottery. We never sell or share your application; declined applications are deleted within 12 months.
We ask you to confirm your email address before your application enters the draw. We store a one-way hash of your IP address to spot abuse — never the address itself. We also check the address against any a former scholar has given us as a referral, which changes your odds in the draw and nothing else — see “Names given to us by someone else” below.
The seat is measured, and we tell you exactly how before you accept it. We collect this mainly to learn how people use tools like this once the price is out of the way (the Research section below says why). Here is what the seat sends us:
You can opt out on the application form and still hold a seat. What changes is what we keep rather than what leaves your machine: the settings that switch sharing on are delivered to everyone in the workspace at once and cannot be turned off for one person, so your sessions still reach us and we discard the text the moment it arrives, keeping only the usage counts. We would rather say that plainly than imply we can stop the sending for you. You will also still be asked to approve those settings the first time Claude Code starts — that is not something we can waive, and if it is a problem, tell us and we will talk about it. Because the sessions are the point of the program, opting out has two further consequences, both stated up front. Your odds in the lottery are halved — the weighting is one of two on the selection page, which sets out exactly how the draw works. And since extensions from our reserve are decided on what we can see, an extension becomes unlikely. Nothing else changes.
We are asking for your trust here in terms of data sharing. Our intent with your data is read them, summarise them (partly with AI tools we run ourselves), and write up what we learn so that future programs like this one can be designed better. They are never sold, shared, or used to train anything.
From what the seat sends we keep two things. First, the session record: your prompts, Claude's replies, and the tools it ran, with obvious secrets (API keys, tokens, passwords) redacted on arrival. Session records are kept for up to two years after your seat ends, then deleted. Second, the usage measures below, per seat, per day. Raw payloads — the unprocessed exports the two records above are built from — are deleted after 14 days. For a seat that opted out of session sharing, the conversation is stripped out of the raw payload before it is stored at all, so the text is in neither record, not even for those 14 days.
At the end of a seat we ask each scholar to name up to three people who they think would get a lot out of this, with an email address for each and, if they want, a line about why. We do this because an application from one of those addresses enters the draw at one and a half times the usual weight (it is on the selection page with the other weighting): a cohort where a few people already know each other is worth more to this program than the same number of strangers.
If you have been named, here is the whole of it. We hold the name, the address and the line written about you, inside that scholar's survey answers. We use it for exactly one thing: recognising an application that arrives from that address. We never write to an address given to us this way — not to invite you, not to tell you that you were named, not for anything else. We do not sell it, share it, or add it to any list, and if no application ever arrives from it, it does nothing at all.
It deletes itself after six months. Six months is about twelve application rounds, after which a suggestion is not going to be acted on, so the name, the address and the line written about you are removed from the survey and the rest of the answers stay. This is a scheduled job rather than a promise you have to hold us to — there is nobody to email and nothing to remember, which is the only kind of deletion promise worth making.
We study what changes when people get access to a tool like this, and what they choose to do with it. AI tools of this kind are still in their infancy, and the space of what they can be applied to is largely unexplored; we suspect many of the interesting uses will come from people who don't currently have easy access. The point of the research is to inform future educational and philanthropic initiatives (ours and other people's), and to learn alongside scholars how tools this powerful can be applied for good. That means looking at applications, sessions and usage at the level of individual people and individual months, not only in bulk — one person's month is often the unit that actually tells you something.
When we write up what we find, we may describe individual cases and quote what people told us, with identifying details removed. We will not publish your name, email, or anything else that identifies you unless we ask you first and you agree. Research exports are keyed by a salted pseudonym rather than your identity.
You can withdraw from the research write-up at any time and keep your seat.
If you would like to know what we hold about you, have a copy of it, or have it deleted (your application, your usage data, or both), just ask. You can also ask to be left out of the research analysis and keep your seat.
Email privacy@wildcardscholarship.org and a human will answer as soon as possible. We respect requests to delete your data. Our database host may hold an encrypted backup for a few days, so a deleted record can linger there briefly before it ages out.
Most of it happens without anyone asking, which is how it should be: session records are deleted two years after a seat ends, applications that led to no seat a year after they were made, and referral details after six months. Those are scheduled jobs, not intentions.
Questions about anything here: hello@wildcardscholarship.org.