From: gregisenberg
Dan Chipper, an incubator of AI startup ideas, shares his insights on the future of AI, particularly focusing on voice-first technologies and new media formats. He emphasizes the opportunity to build products by thinking about how AI can make previously expensive tasks cheaper and create new user experiences [00:00:02].
Advanced Voice Notes and Interactive Media
Dan Chipper highlights the potential of advanced voice notes, inspired by ChatGPT’s voice mode, for enhanced communication and information sharing [00:01:01]. He envisions a system where a user can have a conversational session with an AI to think through issues, like business strategy, and then share a summary with colleagues [00:01:18]. The key innovation is the ability for recipients to interact with the transcript of the original conversation, asking questions and getting deeper context beyond the summary [00:01:53]. This approach aims to reduce repetitive questioning and provide richer context [00:02:18].
This concept extends to interactive media. An internal incubation at Chipper’s company, called “tldr,” converts meetings into podcasts that summarize key takeaways in minutes, allowing users to catch up on missed discussions [00:03:01]. A prototype further allows listeners to interrupt a podcast and have a conversation with the host to clarify points [00:03:30]. The broader idea is to transform any media into a two-way, interactive experience with AI [00:04:29].
Monetization of New Media Formats
Monetization for these advanced voice notes and interactive media can be considered based on value delivered, such as the number of downloads, plays, or listens, which scales with the organization’s usage [00:07:33]. A trend in AI pricing is shifting away from monthly subscriptions towards charging per successfully completed task [00:08:26].
Voice-First Interfaces
Voice is emerging as a natural interface, especially for younger generations [00:28:28]. This is observed in children’s use of Apple Watches and Siri, where they master voice commands due to limited screen interaction [00:28:46]. Advanced voice mode allows even pre-reading children to interact with AI for hours [00:29:07].
This shift suggests a potential return to an “oral culture” where interactions with computers and the environment are primarily voice-based [00:30:01]. Startup opportunities exist in developing “voice-first interfaces” for existing services, such as a voice-first version of Expedia [00:30:35]. Products with a conversational dynamic, like an improved phone system, are particularly well-suited for voice-first development compared to those requiring complex information retention [00:31:21].
“big companies are really Limited in the risk of the amount of risk that they can take to like make their actual interface good for that specific thing because they can’t risk like pissing off their regular users and that’s a tremendous Advantage for startups to like actually figure out what the future of this is like and start with the early adopters of people who really want to use it that way and then slowly just get into the mass market after that” [00:33:56]
This highlights the advantage of startups in innovating with voice-first interfaces, as larger companies may be hesitant to disrupt existing user experiences [00:33:56].
Other AI Startup Ideas
N of One: Personalized Data and Predictions
Inspired by Kaggle, a platform for data scientists to post datasets and bounties, the “N of One” idea proposes a platform where individuals can post their own personal or business datasets and offer bounties for useful predictions [00:09:03]. With the rise of AI tools like GPT-4, “everyone’s a data scientist,” making it easier for non-experts to analyze data [00:09:40].
A personal example given is using biometric data (e.g., from Whoop) combined with voice and facial movement data to predict the onset of OCD symptoms [00:10:12]. This concept suggests a new approach to science focused on building predictive models for individuals (“n of one”) rather than generalized explanations [00:11:56].
Database as a Service (DaaS)
Public datasets, like those found on Kaggle, offer a “gold mine for coming up with startup ideas” [00:12:54]. The easiest startup idea from these datasets is “Database as a Service” (DaaS) [00:15:17]. DaaS involves curating existing data, putting it behind a paywall, and selling it, often to social audiences or through paid advertisements [00:15:31]. An example is leveraging data on startup votes and cities to identify regional interests or market gaps [00:14:47].
AI-Generated New Great Books
AI enables the creation of new media formats for classic works, such as the “Great Books” [00:16:11]. Many find these canonical texts difficult to read [00:16:36]. AI can generate new versions that are translated and tailored for specific audiences (e.g., a 30-year-old tech-savvy individual), making them highly engaging [00:17:09]. These could include multimodal formats like audio and video, even AI-generated Plato movies or interactive elements with advanced voice mode [00:17:56].
Monetization could involve one-off access, subscriptions, or creating luxury physical versions with digital companions [00:19:20]. This aligns with the idea that as technology makes things cheaper, the original forms can become luxury status goods [00:23:02] (e.g., Broadway plays became luxury experiences after TV and phones made entertainment cheap [00:23:12]). The opportunity lies in making non-luxury items luxurious through AI-assisted craftsmanship and limited editions [00:23:35].
AI and Luxury
As AI makes things cheaper, look for what can become more expensive, human-made, luxury status goods for a smaller percentage of the population [00:23:43].
Distribution Strategies for AI Apps
For those building AI apps, an effective distribution strategy is crucial [00:34:33]. Dan Chipper’s strategy involves leveraging an existing media business and newsletter that has cultivated an engaged audience [00:34:51]. By building products for themselves and their shared interests, they can attract an initial user base of hundreds to thousands of people on the first day [00:36:16]. This “distribution first” approach is key in the current AI wave, which offers many “low-hanging fruit” opportunities [00:35:56].
For new ventures in 2025, starting with a YouTube channel and then funneling subscribers into a newsletter is recommended [00:38:18]. This allows for audience segmentation to ensure app invitations are sent to the most relevant users, fostering community and gathering valuable feedback [00:39:39].
The Anti-AI Trend
An interesting trend noted is that many college students are anti-AI, partly due to concerns about job market competition and initial negative experiences with AI tools [00:26:42]. This opens up an opportunity to build “anti-AI products” or products that help users disconnect from digital overload, similar to tools like “The Brick” or “Light Phone” [00:27:35].
Further Resources
- Spiral: spiral.computer (Automates repetitive creative work like social media posts and YouTube descriptions) [00:40:25]
- Sparkle: makeitsparkle.co (Automatically organizes file systems with AI) [00:40:37]