Marketing entry · Hero + how-it-works + a live, playable core-feature demo · No sign-up in the prototype
◎ Criterion [See a live interview] [Start practising]
◦ The AI interview coach that shows its work Practice the interview. See exactly how you're scored. Then argue with it. Most interview tools hand you a number. Criterion shows you the exact words behind every score — and if you disagree, you can challenge it and watch it re-evaluate, live. [Start a practice interview →] [Try the core feature ↓] ~25 min · Senior PM and 6 other roles · no sign-up in this prototype
How it works 1 Answer real questions A live interviewer adapts to what you say — typed or spoken. No trick questions, no hidden rules. 2 See the standard Every score links to the exact words you said that earned it. Nothing is scored against something you can't see. 3 Challenge the score Disagree? Make your case and watch the score re-evaluate in the open — up, down, hold, or sent for human review.
This is the whole idea — try it A score you can talk back to Here's a real score from a practice interview. You think it's too low. Challenge it. Handling ambiguity 5/10 Confidence Medium · weight 15% Why this score: you narrowed the scope, but didn't state the assumption you were making or what would change your mind. [⚑ Challenge this score]
↻ Re-evaluating against the transcript & rubric… ◦ Reviewing your case ◦ Re-reading the transcript for missed evidence
✓ Score revised: 5 → 7. You were right — in Q5 you named the assumption and a test to resolve it. That evidence is now matched to this criterion, and confidence rose to High. Roughly 1 in 4 challenges hold or lower the score — it's a real re-evaluation, not a rubber stamp. [Reset demo] [Now try your own interview →]
Every point traces to a quote No black-box numbers. Click any score to read the words behind it. Honest about uncertainty Low confidence means "show me more," not "you failed." Knows when to defer When evidence can't settle it, it asks for human review instead of faking certainty.
Two quick choices, then in · Stepper: Set up · Interview · Score · Debrief · Everything is changeable later
◎ Criterion [1 Set up]—[2 Interview]—[3 Score]—[4 Debrief] [Exit] Let's set up your interview Two quick choices, then you're in. You can change anything later.
1 · Pick a role [Senior PM ✓] Behavioural + product sense [Software Engineer] Behavioural + system design [Product Designer] Craft + critique
2 · How do you want to practice? [Video call ✓] Like a real remote interview. Camera optional. [Voice] Speak your answers, no camera. [Chat] Type at your own pace. Calm, low-pressure. ⛨ Video is for practising presence and nerves — you're scored on what you say, never how you look or sound.
3 · What we'll look for ✓ Adds up to 100% This is the entire basis for your score — shown before you start, not after. Tweak the weights if a real interview would emphasise things differently. Product sense "Frames the problem before the solution" 30% Structured thinking "Answer has a spine you can follow" 25% Communication "Clear, concise, no jargon-as-armour" 20% Handling ambiguity "Moves forward without full information" 15% Self-awareness "Names the tradeoffs and the risks" 10% [↺ Reset to default]
Ready when you are Practising as Maya Osei 5–7 questions · ~25 min · the first is fixed, the rest adapt to your answers. [Start the interview →]
Live adaptive interviewer (video / voice / chat) · Progress + topic · Honesty bar always present
Question 3 of 6 · Behavioural A time you shipped something that failed ● Recording 0:18 [⏸ Pause] [ ▰▰▰▱▱▱ ]
● REC 0:18 ⦿ Live · adaptive interviewer [ Alex · Interviewer — live video ] Captions: Alex · Interviewer …▍ Self view: MO · You · Camera off Controls: [mic] [camera off] [captions] [end call]
⛨ You're scored on what you say — not how you look or sound.
[End & see score] (also: end-call button on the stage)
The product's whole thesis on one screen · Overall score is provisional · every criterion links to the exact words and can be challenged
Your scorecard · Senior PM 6.7 out of 10 provisional · open to challenge Disagree with any of these? Say so. Every score below links to the exact words that earned it. If one feels wrong, challenge it — it'll re-evaluate right here, in the open.
● Handling ambiguity 5.0 Confidence Medium · weight 15% · provisional Why this score: you narrowed the scope, but didn't state the assumption you were making or what would change your mind. (1 quote · Q3) [⌕ See the evidence] [⚑ Challenge this score]
Q3 "I'd narrow it to the riskiest segment and move." ▶ 6:18 Only one short quote matched — that's why confidence is Medium, not High.
● Product sense 7.5 Confidence High · weight 30% · scored Why this score: framed the problem before the solution and updated your diagnosis with data. (3 quotes · Q1, Q3, Q5) [⌕ See the evidence] [⚑ Challenge] Q1 "I'd ask who's actually hurting before touching the feature set." ▶ 0:42 Q3 "The drop was after the aha moment, not before." ▶ 6:18
● Structured thinking 6.0 Confidence Medium · weight 25% · provisional Why: a clear spine early, but the close drifted from the structure you set up. Q2 "First the problem, then options, then the call." ▶ 3:05 ● Communication 8.0 Confidence High · weight 20% · scored Why: concise, plain language, no jargon-as-armour. Your strongest signal. Q4 "In one line: we shipped the wrong half first." ▶ 8:40 ● Self-awareness 6.0 Confidence Low · weight 10% · thin evidence Why: you named tradeoffs when prompted, but rarely volunteered the risk in your own plan. Q6 "In hindsight I'd have flagged the risk sooner." ▶ 13:20
● scored firm, evidence-matched ● provisional matched, but open / medium confidence ● thin evidence low confidence — show me more, not "you failed"
⛨ Roughly 1 in 4 challenges hold or lower the score — this is a real re-evaluation, not a rubber stamp. When evidence genuinely can't settle it, Criterion returns human review ⚑ rather than faking certainty. [Finish & see your debrief →]
Inline within a criterion card · Four steps: make your case → re-evaluate → reasoning → one of four outcomes
Make your case [ Matched too few answers ] [ I covered this elsewhere ✓ ] [ It misread me ] "In Q5 I said I'd ship the smaller version to test the assumption first — that IS naming the assumption and how I'd resolve it." [Re-evaluate this score →] [Cancel]
◠ Re-evaluating against the transcript, rubric & confidence threshold… ◦ Reviewing your case & confidence signals ◦ Re-reading Q5 for evidence that wasn't matched before ◦ Deciding: revise up · down · hold · human review
✓ Score revised: 5 → 7. You were right — in Q5 you named the assumption and a test to resolve it. That evidence is now matched to this criterion, and confidence rose to High.
— Score held at 6. You argued you named tradeoffs; on re-reading, they were all prompted, not volunteered. Here's where each appeared so you can judge for yourself.
▾ Score lowered. Re-reading surfaced a claim the transcript doesn't support — reasons shown. ⚑ Sent for human review. The evidence can't settle this on its own. A person will review your case rather than the system faking certainty.
Modal from any "▶ timestamp" or "See the evidence" link · Silent clip + the matched quote · scored on words, not footage
Replaying this moment of the interview Handling ambiguity [✕]
[ silent replay of your recorded answer ] ⏱ Your answer · 6:18 [↺ Replay segment] Q3 "I'd narrow it to the riskiest segment and move." What it matched: the criterion's test for moving forward under incomplete information.
⛨ Replayed so you can hear your answer again — you're scored on your words, never on the footage.
[↺ Replay segment] [Next moment ›] [⤢ Open full interview] [Close]
Closing summary · One focus area, the final score, strongest signal, full breakdown, and the contest record
✓ Interview complete Solid round. Here's your one thing to work on. Maya Osei · Senior PM · 24 min · 6 questions, 2 adaptive follow-ups · 2 scores challenged
◎ The one thing to work on next Volunteer the risk in your own plan Your weakest area was Self-awareness — and it's the score that held when you challenged it. You named tradeoffs when prompted, but rarely before being asked. Next time, get ahead of it: "Here's where this could go wrong, and what I'd watch." [▶ Replay the 2 transcript moments this showed up]
Final score 7.4 out of 10 · all scores firm & evidence-matched Strongest signal You updated your diagnosis with data in real time (Q3). That's the moment a strong interviewer remembers. [▶ Replay the moment →]
Full breakdown tap to revisit any score Product sense ↑ +0.5 8.0 Structured thinking 7.0 Communication 8.0 Handling ambiguity ⚑ 5→7 7.0 Self-awareness held 6 6.0
What you challenged The contest record travels with your score — a hiring panel can see what was disputed and how it resolved. ↑ Handling ambiguity · revised 5 → 7. You argued it was covered in Q5 — and you were right. — Self-awareness · held at 6. You argued you named tradeoffs; Criterion showed they were all prompted, not volunteered — reasons given.
[⤓ Download scorecard] [↻ Re-run, harder rubric] [⇪ Export transcript]
You were scored in the open: a rubric you set, criteria that filled in while you talked, and a challenge that changed the result. Practice → see the standard → challenge the score — an evaluation you can argue with, not just receive.