Can you build?
Give every candidate the same interactive challenge and an AI to direct — the way the job is actually done now. You get a scored, evidence-backed report and make the call. No trivia, no take-home theatre.
We're running this as a paid pilot with a small number of organisations, and publishing what we learn. It isn't for sale online — if it sounds useful, talk to us.
The CV and the take-home stopped telling you anything.
When every candidate writes and codes with the same AI, the polished artefact tells you about the tool, not the person. Hand them a real problem and an AI to direct instead, and the signal comes right back: what they ask first, what they verify, and what they're willing to ship.
The whole thing, in three pages.
The concept, what candidates experience, and the identity separation and oversight built in — end to end.
Read →A rubric written for the role, scores that cite the transcript, and a verdict only a person makes.
Read →Short, honest field notes on what changes about hiring once candidates work with AI.
Read →No names. No emails. No CVs. We never learn who your candidates are.
Every candidate is an opaque link token — there is no field anywhere in the product for a name, an email, or a CV. The mapping from link to person stays in your systems. No webcam, no microphone, no screen recording, no keystroke logging: the assessment is text, and the record is the work.
It is still personal data, and we'll sign a DPA saying so. It's just a very short list — here is all of it.
The questionnaire, answered up front.
Most assessment tools set off a long review because of what they collect — faces, screens, keystrokes, CVs. We collect none of it, so the answers are short. Here they are before you ask.
| What reviewers ask | Our answer |
|---|---|
| Do you store candidate names, emails, or CVs? | No. There is no field for them. |
| Webcam, microphone, screen recording, keystroke logging, biometrics? | None. Not now, not optionally, not ever. |
| Then what do you hold about a candidate? | An opaque link token, their conversation with the AI, timing (including prompt-count and pacing figures the judge derives from that transcript), per-criterion scores and reasons, your reviewer's decision and notes, and your own private label for them if you set one. That is the complete list. |
| Automated decisions? | No. The AI scores the work; a named person decides, and the trail exports. |
| Do you train AI models on our data? | No. |
| Where does it live, and who else touches it? | The EU. Sub-processors: Anthropic (scoring), Cloudflare (CDN/TLS), Hetzner (hosting), Sentry (errors). That's all of them. |
| Deletion? | Any assessment, on request. Everything from a pilot is deleted at pilot end + 90 days. |
| Is this still personal data? | Yes — pseudonymous. You're the controller, we're the processor, and we'll sign a DPA. We'd rather tell you that than have you find it. |
Imagined by builders. Built for hiring.
Can You Build? is an experiment by paddoq, a small group of engineers exploring how work — and how we hire for it — changes now that everyone builds with AI. Every assessment is evidence: what someone can actually do, judged on the work.
Try it on one real role.
We run Can You Build? as a paid pilot with a small number of organisations — fixed scope, fixed fee, a defined end date, and your data deleted when it's over. Tell us what you hire for and what you'd want to measure, and we'll tell you honestly whether this is any use to you.
We reply personally, usually within a day. There's no sign-up form and no way to buy access online — a pilot starts with a conversation.