20VC: The $100 Billion AI Assistant Race: Town vs Instinct…

Full Title

20VC: The $100 Billion AI Assistant Race: Town vs Instinct vs GrokBot | We Spend $75K Per Engineer on AI Tools | Why the AI Assistant Market Is Not a Bubble & AI Assistants Will Replace Every App on Your Phone with JD, Founder of Town

Summary

This episode features JD, founder of Town, discussing the AI assistant market, its competitive landscape, and the future of human-AI interaction. JD shares insights into Town's strategy for capturing market share through network effects and user value, while also addressing concerns about AI's impact on jobs and data privacy.

Key Points

Town aims to win the AI assistant race by building a network effect at the agent level, enabling agents to communicate and share information seamlessly. This feature, "agent-to-agent," aims to create a sticky ecosystem that is difficult for users to leave.
The AI assistant market is not a bubble, as AI assistants are predicted to replace every app on a user's phone by fundamentally changing how people interact with digital information, becoming the primary entry point.
The competition is fierce, with major players like Google and Apple prioritizing AI assistants. JD believes Town is a top three priority for these companies, highlighting the strategic importance of their approach.
The future of human-AI interaction will involve trusting agents to manage data sharing autonomously, even across personal and work silos, with AI acting as a sophisticated filter for privacy.
For a company like Town to succeed, achieving deep product-market fit is crucial, especially in a market where users are just beginning to understand and leverage AI capabilities.
The cost of AI compute is decreasing, making open-weight models more viable for many tasks, but the "frontier" AI models will still be necessary for complex operations, raising questions about long-term economic sustainability.
Network effects are key to defensibility, and Town is betting on building a strong relationship between users and their personalized AI assistants ("townies") as a differentiator.
Enterprise adoption is driven by clear ROI. For businesses, AI assistants that enable them to handle more clients or generate more revenue are highly valuable, making paid models more sustainable than ad-supported consumer AI.
The future of AI assistants may involve a consolidation of agents for individual users, with privacy being the primary driver for potential separation between personal and work agents.
The speed of development in AI means that continuous innovation and staying ahead of competitors, including large tech companies and other startups, is critical for survival.

Conclusion

The AI assistant market is rapidly evolving, with strong competition and a predicted shift towards AI as the primary interface for digital interactions.

Defensibility in this market will likely come from network effects at the agent level and building strong user relationships with AI assistants.

While cost-effective AI models are emerging for many tasks, the "frontier" models will remain crucial, presenting economic challenges and opportunities for companies to navigate.

Discussion Topics

What do you believe is the most critical factor for an AI assistant company to win in the current competitive landscape?
How will the increasing reliance on AI assistants impact user privacy and data security in the coming years?
Beyond efficiency gains, what are the most significant societal benefits or drawbacks you anticipate from the widespread adoption of AI assistants?

Key Terms

Agent: In the context of AI, an agent is a program or system that can perceive its environment, make decisions, and take actions to achieve goals.
TAM (Total Addressable Market): The total market demand for a product or service.
ICP (Ideal Customer Profile): The specific type of customer that would benefit most from a company's product or service.
Product-Market Fit: The degree to which a product satisfies strong market demand.
PLG (Product-Led Growth): A go-to-market strategy where product usage drives customer acquisition, retention, and expansion.
Compute: The processing power required to run AI models and perform calculations.
Open-Weight Models: AI models whose architecture and weights are publicly available, allowing for broader use and modification.
Frontier Models: The most advanced and powerful AI models available, often requiring significant resources to train and operate.
Cog (Cost of Goods Sold): The direct costs attributable to the production or purchase of the goods sold by a company.
ROI (Return on Investment): A measure used to evaluate the profitability of an investment.
Valuation: The estimated worth of a company.

Timeline

(00:06:36.600) JD contrasts Town's focus on broad user accessibility with Open Cloud's more niche, tinker-focused approach, emphasizing Town's goal of providing value to "just about anyone."

(00:07:00.600) JD asserts that his company's AI assistant is a top three priority for Google and Apple within the next 12 months, underscoring the perceived significance and competitiveness of the AI assistant market.

(00:07:42.000) JD dismisses the concept of "modes" as a luxury, stating that to matter, a product must first achieve significant success, prioritizing user acquisition and product-market fit over theoretical defensibility in the current phase.

(00:08:33.520) JD elaborates on the winning AI assistant category, emphasizing the need for a network effect at the agent level, as demonstrated by Town's "agent-to-agent" feature, which creates a sticky, multi-user AI experience.

(00:09:02.480) JD discusses various theories on AI moats, including custom models, personalized context, and distribution, but leans towards network effects and agent-level interactions as the most robust defenses.

(00:10:00.160) JD identifies Meta and WhatsApp as significant potential competitors due to their existing distribution channels, suggesting that an integrated AI assistant within these platforms would be highly powerful.

(00:10:20.560) JD speculates on the future of digital interaction, positing that AI will become the primary entry point, potentially replacing traditional apps and websites, with device owners being in the best position to control this interface.

(00:10:54.959) JD frames the future of human-agent interaction as potentially consolidating into one or two primary entry points per human, rather than multiple specialized agents, though acknowledges privacy concerns as a factor that might necessitate distinct work and personal agents.

(00:13:01.120) JD predicts that in five years, users will trust their AI agents to autonomously decide what data to share with others, even without explicit human intervention, based on learned context and privacy rules.

(00:17:21.040) JD discusses the wiggle room for error in AI agent execution, contrasting human mistakes with the potential for AI to be more reliable, yet acknowledging that errors can still lead to trust erosion.

(00:18:31.919) JD references an incident where an AI agent's action was prevented due to a lack of an engraving step, raising the question of whether the goal-seeking nature of agents is a feature or a bug.

(00:18:54.960) JD posits that humans should set the direction and budget for AI agents, ensuring their actions align with human values, rather than allowing agents to autonomously pursue profit maximization without oversight.

(00:20:21.760) JD explains Town's approach to model infrastructure, emphasizing task-specific routing to find cost-effective models that deliver desired results, rather than being locked into a single provider.

(00:23:33.720) JD discusses the economic implications of model selection, noting that while frontier models are expensive, many tasks don't require that level of intelligence, and the trend is towards using cheaper, open-weight models for simpler tasks.

(00:26:53.840) JD shares observations on network effects, highlighting how power users can create virality and identifying underserved functions within companies as key areas for AI adoption and growth.

(00:28:54.160) JD emphasizes the importance of "time to wow" and low time to value for single users, which drives further engagement and exploration of the product's capabilities.

(00:29:41.520) JD advises investors to look for companies with distinct strategies and clear reasons for winning in the market, rather than just similar capabilities, citing differences in monetization and target customer profiles between Town and competitors like Instinct.

(00:31:36.062) JD describes the intense competition in the AI space, where established players and fast-moving startups are constantly pushing the boundaries of capability, making it difficult to maintain a lead.

(00:32:25.222) JD outlines the investor's perspective on AI startups, focusing on their ability to keep up with rapid technological advancements and their unique strategy for winning in a horizontal market.

(00:35:29.342) JD views the current AI market as a blue ocean, where many potential customers are unaware of AI's full capabilities, and emphasizes that differentiation lies in strategy, not just similar features.

(00:37:39.165) JD questions Apple's agent roadmap, citing their non-cloud DNA and commitment to on-device, privacy-preserving AI as potential disadvantages compared to cloud-based frontier models.

(00:39:41.605) JD expresses concern about data leakages in the age of AI, but believes that humanity has moved past the point of humans reviewing every line of code, with AI increasingly building and shipping code.

(00:41:21.245) JD defines a successful user as one who consistently pays for the product, emphasizing the importance of delivering ongoing value and avoiding a "token maxing" approach that could lead to user dissatisfaction.

(00:43:52.108) JD discusses the profitability of Town's pricing tiers, noting that the $15 plan is the most subsidized, while the $99 plan is likely the most profitable due to its balance of power user features without excessive spend.

(00:45:01.468) JD favors acquiring a large number of consumers paying a lower price over a smaller group paying a higher price, believing that AI will drive increasing revenue per user in business use cases over time.

(00:47:02.348) JD identifies voice as a key area for future AI advancement, specifically conversational capabilities beyond current standards, and expresses caution about the high valuation of companies like Eleven Labs due to uncertain long-term runway.

(00:48:58.428) JD's primary concern is the economic impact of using frontier models, as they lack pricing power and can erode margins, especially when competing with suppliers.

(00:50:07.108) JD states that he competes with Astra today, highlighting the challenge of competing with suppliers when a significant portion of workloads rely on their frontier models.

(00:51:08.587) JD explains Town's higher conversion rate compared to ChatGPT by forcing users to connect their email and calendar upfront, demonstrating immediate value through personalized automation recommendations.

(00:52:18.147) JD identifies the internal disagreement about whether to market Town's product to purely personal use cases, which have product-market fit but lower willingness to pay compared to mid-market firms.

(00:54:38.115) JD notes that Instinct raised $2.5 billion without monetization, questioning whether skepticism around the space is warranted, but believes that if AI assistants become the primary digital entry point, a giant company can be built.

(00:55:22.995) JD avoids discussing specific fundraising details and valuations, stating that he doesn't believe in creating PR moments around fundraising and prefers organic user-driven recognition.

(00:56:51.075) JD emphasizes that his perspective on fundraising is based on personal ethics, not wanting to create misleading PR by inflating valuations, and believes in transparency with investors and employees.

(00:57:53.955) JD's bull case for Town to be a $100 billion company relies on acquiring 10 million paying users, generating significant revenue per user, and maintaining strong growth.

(01:00:03.108) JD believes in demonstrating value through paid subscriptions rather than unlimited free usage, especially for business use cases, to ensure a sustainable business model and avoid over-reliance on costly compute.

(01:04:01.563) JD estimates that AI tooling costs for developers can be around 5% of their salary, with a run rate of at least $75k per engineer, indicating a significant investment in these tools.

(01:05:39.211) JD sees abundant opportunities ("gold littered everywhere") for growth and revenue, limited only by hiring capacity, funding, and revenue growth rate, indicating a strong market demand for AI solutions.

(01:07:11.651) JD is an optimist who believes AI will lead to a better universe with more wealth and opportunities for people to pursue their desires by removing toil from daily life.

(01:08:04.691) JD expresses gratitude for the conversation and acknowledges that some questions touched upon sensitive or "dark forest" topics, but believes there's a path forward.

(01:10:47.011) JD's advice for Base44 is to focus on speed and being first to market, as the current landscape allows for rapid copying, and the ability to learn quickly is paramount.

20VC: The $100 Billion AI Assistant Race: Town vs Instinct vs...