Company facts
| Identity | Yankı is an operating name used by its founder, Mevlana Yalçın, for an emotional check-in companion. Contractual and contact point: merhaba@mevlanayalcin.com.tr · Ankara, Türkiye. Not a registered legal entity. |
|---|---|
| Domain | mevlanayalcin.com.tr is a personal domain the founder has held since September 2023. Product work on Yankı began in October 2026 and the two dates are kept separate on purpose. |
| Team | One person: founder, engineering, support. No agency, no subcontracted build. |
| Claude | Reflections are produced through Anthropic's Messages API with claude-haiku-4-5-20251001 under a reflection-constrained prompt that must return JSON. The output is schema-checked server side; a malformed answer is discarded rather than shown. Crisis wording is decided by a rule list before the model is called. |
| Status | Live: this site, the /api/reflect endpoint, the waitlist. In development: the account, history and reminders. Planned: a clinician-reviewed protocol and a second language set. |
Roadmap
| Reflection endpoint with schema checks | Live | Runs on Cloudflare Workers; sample answers on this page were captured from it. |
| Waitlist with real deletion path | Live | One KV row per address, removed on request. |
| Accounts, check-in history, reminders | In development | Local-first storage is the default; sync will be opt-in. |
| Crisis word list in Turkish and English | Live | Our own conservative list; a false positive costs one unnecessary helpline banner, not a wrong reflection. |
| Clinician review of the reflection protocol | Planned | Not started, no partner signed yet. |
| Mobile apps | Planned | No date committed. |
Verify this independently
Questions a reviewer usually asks
Why a language model at all?
Because naming a feeling from prose is not something a rules engine can do. The parts that must not be probabilistic — crisis detection, output shape, refusal to quote text that is not there — are handled by code around the model.
What is Claude load-bearing for?
The reflection itself. Remove Claude and the product has no centre; there is no rules-based fallback pretending to be one, the box just says it is unavailable.
Is this a wrapper?
The model is the core call; the value is in the constraint around it: JSON-only output, quoted-source-of-truth being the user's own text, rule-based crisis handling, no memory of your sessions in the beta.
Who is it for?
People who want a daily check-in habit and are not in active treatment. It is not for crisis care and says so in the interface.