- Education with a patient AI tutor: promise and risks
Infinite patience and personalisation, but also shortcuts and errors.
4 - Emotion-aware replies: calm voice when I sound stressed
Prosody carries a lot of information. Use it, gently.
12 - Open-source models vs closed models: who wins voice?
Closed leads today. Open is closing the gap and respects privacy.
12 - What happens to call centres in the next ten years?
I think work moves up the value chain, but the transition will be rough.
6 - AI agents negotiating with AI agents: customer service of 2030
Your assistant haggles with their assistant. Do humans even join?
45 - Will voice become the main interface to computers?
Keyboards are 150 years old. Will we look back and wonder why we typed so much?
5 - Latency under 300 ms: why it matters more than IQ
A slightly dumber assistant that answers instantly feels smarter.
6 - AI companions and loneliness: where is the line?
Comfort is real, but so is dependence. Thoughts?
6 - Will on-device models catch up with the cloud giants?
Small models improve fast. For most daily tasks they may be enough.
39 - From assistants to agents: what changes when AI can act?
Answering questions is easy to forgive. Taking actions needs trust, permissions and undo.
9 - Ambient AI: the assistant that listens all day (and the ethics)
Always-on is useful and uncomfortable. Where should the lines be?
5 - Multilingual by default: the next billion users speak local languages
The next wave is not English-first.
6 - Emotion and prosody: will assistants finally sound human?
Rhythm and breath matter more than a perfect accent.
6 - Let me interrupt the assistant mid-sentence (barge-in)
Talking over the answer should stop it instantly, like a person would.
4 - Predictions thread: what is true about voice AI by 2030?
Write one bold prediction and one boring one. We will check in 2030.
11 - A personal AI that remembers your whole life: wonderful or terrifying?
Total recall is a superpower. Who else may ever see it?
5 - What jobs does voice AI create?
Prompt designers, voice curators, evaluators, accessibility specialists...
5 - Earbuds as the new smart speaker
The assistant that is always with you will beat the one on the shelf.
4 - Voice-cloning rules: will laws kill the good uses?
Consent and labelling make sense. Blanket bans do not.
4 - Speech-to-speech models: the end of the text middleman?
Skip transcription and generation as separate steps and latency drops dramatically.
6