The full interview is very much worth a watch/listen:.

Looking forward: Rebecca sees voice AI naturally moving beyond call centers and becoming embedded directly into core product experiences. This vision aligns perfectly with what we're seeing at @uselayercode.

Today, even the best voice agents still struggle with:. 1. Long conversations (15-20 min is today's practical ceiling).2. Sensitive discussions where human touch is the whole point.3. Multi-party calls with several people talking at once.

• The best voice agents excel at conversations with clear boundaries and objectives (e.g. Appointment scheduling, lead qualification & payments collection). • Agents often outperform humans in uncomfortable tasks (like collections) because they lack shame or fatigue. •.

Today's best voice AI agents focus on specific, high-value use cases in industries where calls are revenue drivers — not cost centers. Enjoyed @davitb's recent conversation with Rebecca Greene on this and other @regal_ai insights from building customer-facing voice agents. Some.

Today's @ycombinator startups = tomorrow's unicorns.

So what's next? The cutting edge has already moved beyond phone agents. Recently, I've seen demos of voice agents that negotiate calendar conflicts, book complex travel, and manage entire projects through conversation. If this is happening now, imagine where we'll be in.

That's why we're building @uselayercode:. To handle the plumbing so any developer can ship natural, conversational voice AI that actually works in production. No specialized team or audio expertise required.

Building production-ready voice AI means wrestling with complex orchestration, edge cases, and latency optimization. Most devs don't want to become experts in conversational voice pipelines — they just want to ship code.

Devs can now leverage powerful APIs for transcription, text-to-speech, and audio processing to add voice to their existing text agents. But there's still a gap. .

Second: The dev toolkit is maturing fast. While audio-native models (OpenAI's advanced voice, etc) remain expensive, restrictive, and unpredictable, text-powered voice pipelines are more robust in production.

First: Infrastructure costs dropped while voice quality advanced dramatically. Foundation LLMs and modular voice infrastructure can now deliver conversations that feel genuinely human. What was prohibitively expensive has become affordable and now makes economic sense.

Voice AI is crossing the chasm. A year ago, responsive voice agents were hackathon demos. Today, most businesses have no idea they can hire agents to handle calls, book appointments, and guide customers 24/7. Two breakthroughs have made this possible:

RT @levie: You can code far faster with AI, but the work of building the right thing, scaling a service, and creating a business still rema….

RT @dctanner: Never trust your LLM on 1st April. Have we inadvertently trained them all to play April fools pranks based on training data….

Ensuring none of us are ever stuck listening to hold music again is a multi-billion dollar industry.

Great work by @amix3k and the @todoist team. Watch this to see an excellent use of voice AI + responsive visual UI.

You can check it out out for yourself at For the most fun: try booking a fantasy trip to your favourite fictional location.

We're building @uselayercode to make it easy for your voice agent to update the state of a visual interface alongside your natural conversation. We built this demo to show how voice + UI can help reduce friction in an area we're all familiar with.

Most voice demos feature open-ended conversations with a voice agent, kind of like a phone call. These are fun, but only show a slice of what's possible with voice AI today. As agents becom smarter and more capable, I'm more excited about the huge potential of voice + UI to.