Raiza Martin remembers the moment clearly. It was late 2024, and NotebookLM—the AI research tool she'd helped build at Google—had just cracked the internet's attention span. Audio Overviews, the feature that transformed dense documents into surprisingly listenable podcast-style conversations, was everywhere. User uploads were spiking. The product had momentum.
And Martin, along with Jason Spielman and Stephen Hughes, decided to leave.
The problem they'd identified wasn't one NotebookLM could solve. Most people, they'd noticed, don't curate document libraries to upload anywhere. They don't organize PDFs or maintain reading lists. What they do maintain—often obsessively—are their inboxes and calendars. That's where the real information lives, the messy scaffolding of daily life. So the three of them walked out of Google's doors and started building something new.
Six months after launching publicly on September 23, 2025, their answer—a startup called Huxe—is still feeling its way forward. But the pitch is straightforward enough: connect your email and calendar, choose a few topics you care about, and each morning Huxe delivers an audio briefing synthesized from your actual schedule, inbox, and the news. No uploading. No curation. Just press play.
Whether anyone actually wants this remains an open question.
Audio, Interrupted
The product revolves around three formats, all of them voice-first. Daily Briefings pull together your calendar appointments, emails (Google or Outlook), and news about topics you've flagged. "Live Stations" function more like topic-specific radio streams—tune in and they refresh with new information as stories develop. DeepCasts are instant, AI-generated podcasts that spin up when you type in a question or subject you want explored.
The twist is interactivity. You can cut in mid-stream to ask follow-ups, request more or less depth, redirect the conversation entirely. It's less podcast, more dialogue with your own data. The app also displays what Huxe calls "generative UI"—contextual information that updates as the audio progresses, complete with sources and timestamps for anyone who wants to verify what they're hearing.
WIRED tested the app in February 2026 and noted playback speeds up to 3×, a feature that suggests users are treating these briefings more like functional utilities than entertainment. Which, perhaps, is the point.
The $4.6 million seed round announced at launch came from Conviction, Genius Ventures, Figma CEO Dylan Field, and Google Chief Scientist Jeff Dean—a roster that signals investor confidence in audio as an interface for personal information. After NotebookLM's viral success, that bet feels less speculative than it might have a year earlier.
Moving Fast, Staying Small
The app's development history shows a team that iterates quickly, maybe out of necessity. CarPlay support arrived in late June 2025, before the public launch even happened. Live Stations shipped the day after the September announcement. By November, Shows and Episodes functionality had appeared. January 2026 brought a redesigned Daily Brief with deeper personalization options.
As of early March 2026, the iOS app carries 274 ratings at 4.6 stars. Android performs slightly better: 4.8 stars from 807 reviews, with over 10,000 downloads. The company claimed "thousands of users" and a Discord community north of 5,000 members in a LinkedIn update three months after launch. Those numbers are unverified, but they track with the app store data.
The team remains notably small. LinkedIn lists 2–10 employees as of March 2026, though that range is self-reported. For a product that requires real-time synthesis of personal data, tight calendar integrations, and interactive audio generation, it's an impressively lean operation. Or a risky one, depending on how you look at it.
The Privacy Calculation
Huxe is testing a specific hypothesis: that people will accept voice as their primary information interface, with text as backup. That's meaningfully different from NotebookLM's model—upload sources, generate discussion—and closer to what Google itself has been experimenting with. In December 2025, the company announced "CC," a morning brief feature for Workspace users that delivers text summaries to Gmail each morning. Huxe delivers audio you can interrupt.
The product is free for now. The terms of service preserve the right to introduce pricing later, which is standard startup language. What's less standard—or at least less comfortable—is the privacy equation. Huxe uses third-party LLM and voice vendors to process your data. The company's September 2025 privacy policy promises not to use identifiable calendar or email content for model training without explicit opt-in, but users are still routing personal information through external systems. WIRED flagged these trade-offs in its February walkthrough, and they're worth sitting with.
There's also structural dependency risk. Huxe relies on Google, Apple, and Microsoft APIs for calendar and email access. The company's terms explicitly acknowledge that third-party providers can change permissions at any time, potentially affecting which features work. It's a vulnerability baked into the architecture.
Everyone Wants Your Morning
The competitive landscape is filling out, though not yet crowded. BriefPod generates daily podcasts from newsletters without touching your inbox. BriefingAM claims similar email and calendar integrations. Google's CC brief, even in text form, signals that the tech giants recognize value in auto-assembled morning summaries. They're watching this space.
Huxe's differentiator, at least for now, is interactivity—the ability to interrupt, redirect, drill deeper in real time. That makes it less like a passive podcast and more like a conversation with your own information, which is either compelling or unsettling depending on your relationship with AI.
Whether that's enough to sustain a business, and whether users will accept the privacy trade-offs involved, depends on execution at scale. The NotebookLM lineage matters here. That team already proved they could make AI audio feel natural and useful. Huxe represents the next iteration of that thesis, applied to the messiest, most personal dataset most people possess: their actual digital lives.
The technology clearly works. The question isn't capability. It's desire—whether people want their mornings narrated back to them, and whether they'll trust a six-month-old startup with the keys to their inbox to make it happen. That's the bet Martin, Spielman, and Hughes are making. Six months in, the answer is still forming.
