Why this comparison exists
Sensei AI built one of the best-looking landing pages in the category — their design language genuinely competes with Linear or Vercel for premium-minimal craft. Behind the design, the product is competent but unremarkable: standard Electron overlay, Anthropic + OpenAI routing, no STAR-personalization beyond resume parsing.
If you're picking based on landing-page aesthetics, Sensei is a reasonable choice. If you're picking based on what the product does mid-interview, Mirly's structural advantages compound.
Feature matrix
| Capability | Mirly | Sensei AI |
|---|---|---|
| First-token latency | 127ms p50 (open benchmark) | 720ms p50 (measured), "<1s" claimed |
| Pricing | £5 single / £10 pack / £29.99 monthly | $39/mo, no single-interview option |
| Free trial | 7-minute session, full product | 3-minute, model-restricted (no Claude on trial) |
| Renewal | Opt-in monthly | Auto-renew |
| Personalization | Resume + JD + STAR stories + vocabulary fingerprint | Resume parse only |
| STAR-stories support | Native — paste 3–5 stories in your own words, model uses them | Generic STAR-aware prompt, not story-specific |
| Voice fingerprint | Extracted at onboarding from a 60-second sample, baked into prompt | Not supported |
| Stealth APIs | NSWindow.sharingType + WDA_EXCLUDEFROMCAPTURE | Same approach |
| Status page | 24-hourly tested | None |
The personalization gap
Sensei's "personalization" reads the candidate's resume and extracts key phrases. The model uses those phrases generically. The resulting answers are better than nothing but consistently include filler — "leveraged my experience to drive impact" — that anyone reviewing your interview recording would clock as AI-generated.
Mirly's approach is more invasive in the right way:
- Resume + JD baked into the system prompt with cache_control so Anthropic's prompt-cache kicks in (saves ~280ms per call)
- 3–5 STAR stories in your own words — the model has these as exemplars of how you talk
- Voice fingerprint extracted from a 60-second sample at onboarding — captures sentence-length distribution, vocabulary frequency, signature phrases ("the hardest part was...")
The output sounds like you with the structure tightened up, not like a generic Sensei answer with your name inserted.
What Sensei does better
Sensei's onboarding UX is, frankly, more polished than ours. Their setup flow is six steps with animated illustrations and a working sample-call demo. Mirly's onboarding is functional but feels comparatively utilitarian. This will improve on our side over the next few weeks.
Their support response time is also faster — most queries get a human reply within four hours. We're slower at the moment because we're a small team.
Pricing comparison
| Plan | Mirly | Sensei AI |
|---|---|---|
| 1 interview | £5 | Not offered |
| 3 interviews | £10 | Not offered |
| 10 interviews | £20 | Not offered |
| Monthly | £29.99 (opt-in) | $39 (auto-renew) |
| Annual | — | $390 (auto-renew, 17% saving) |
Mirly is meaningfully cheaper at the monthly tier (£29.99 ≈ $37 vs Sensei's $39) and adds the single-interview option Sensei doesn't have.
Switching guide
- Cancel Sensei → Settings → Subscription
- Download Mirly
- Complete the 6-step onboarding (resume, JD, STAR stories, optional voice sample)
- Free 7-minute trial on a real interview