Hey ! 👋 I built Synvix because I was terrified of technical interviews. Not the coding part — the talking part. You know that moment when the interviewer is explaining their system architecture and you're desperately trying to figure out if they want an answer or if they're just setting context? Every existing "AI interview helper" I tried would fire off answers mid-sentence. The interviewer says "So we use Kafka for event streaming—" and the AI starts generating a paragraph about Kafka. Now you're reading the answer while the interviewer is still talking. Awkward. Synvix v1.2.0 fixes this with a question classifier. Before answering anything, it runs a fast check: is this actually a question? If the interviewer is just talking, explaining, or sharing context — Synvix stays quiet. No more reading answers nobody asked for. Other stuff that matters: Stealth mode — Windows Display Affinity makes Synvix invisible to Zoom, Teams, Meet, OBS during screen share. Not a minimized window. Literally not capturable. 26 languages — transcription and answers work in your language Custom base URLs — use OpenRouter, self-hosted models, or any OpenAI-compatible API Local-first — your API keys never leave your machine, no accounts, no telemetry It's free, open-source, and runs on Windows/Mac/Linux. Would love feedback on the classifier logic — right now it's a single fast LLM call but I'm exploring whether a smaller fine-tuned model could do it even faster. What do you think? synvix-admin@proton.me or DM on X @DeanBlunn

Major update and bug fixes

The interview assistant that knows when to speak

The interview assistant that knows when to speak