Summary
Highlights
Many technical people are interested in AI but lack a compelling startup idea. The hosts warn against 'hackathon ideas' or jumping on bandwagons, which are often superficial and easy to build but lack depth and real-world applicability. Truly impactful ideas are often harder to build initially but address significant problems.
Founders should look within themselves, examining their unique history, skills, and prior work experiences. Examples include Salient AI, whose founder leveraged his experience at Tesla to address auto debt collection, and Diode Computer, whose founders combined electrical and software engineering expertise to create an AI circuit board co-pilot. This approach leads to 'founder-market fit,' identifying problems only they are uniquely suited to solve.
Internships, especially at leading-edge companies, can be a rich source of startup ideas. The example of Data Curve's founder, who leveraged her Cohere internship experience in data tools for LLMs, illustrates how previous roles can reveal unmet needs in rapidly evolving fields. Working at companies pushing technological boundaries helps identify high-quality, future-oriented problems.
Some ideas stem from a founder's desire to see a specific change in the world, something they are passionate about working on long-term. Gabriel, founder of Can of Soup, was encouraged to pursue an idea that 'captures the human imagination,' leading to an ambitious AI-powered social network. Happenstance, which uses AI to improve search for connecting people based on fuzzy queries, is another example of building something truly desired.
If personal expertise doesn't yield ideas, founders must actively seek them out in the real world. This involves treating themselves as researchers, getting outside their comfort zones, and observing industries firsthand. For example, the founder of ESS Health gained insights into dental office administration by shadowing a dentist (his mother), leading to an AI solution for insurance processing.
A more extreme method involves taking an 'undercover' job in an industry to deeply understand its problems. One founder took a job as a medical biller to identify opportunities for AI automation. This direct immersion reveals unique pain points, especially in manual, knowledge-work-intensive roles, which are ripe for AI disruption.
Job listings (like on Indeed.com) for remote analyst or clerical roles, especially those in low-wage outsourced environments, can signal areas where AI can create significant value. Lilac Labs, for instance, found its niche automating drive-thru order taking after realizing these roles were often outsourced. Similarly, problems with existing enterprise software often lead to consultancies, indicating a gap for a better, AI-driven product.
Consistently engaging with the latest technology and developing on bleeding-edge platforms allows founders to be among the first to realize new possibilities. Examples include Automate, whose founders leveraged early access to Google's Bard (now Gemini) to build a superior RPA solution, and Pre DB, whose founders, immersed in the world of technical builders, discovered a need for better Postgres-Vector database integration.
Building and shipping products, even if they aren't immediate successes, develops expertise and connections that can lead to better ideas. Juicebox (now PeopleGPT) started as a freelancer marketplace but their work with users revealed a need for AI-powered recruiter search. Founders are encouraged to trust their direct experience and the market's response, rather than being swayed by external opinions or perceived competition.
It's normal for founders to take time to find the right idea. The rapidly changing landscape of AI means new opportunities are constantly emerging, offering a 'golden age' for ambitious entrepreneurs. Founders are encouraged to persist, leverage their deep insights from intrinsic or external exploration, and build solutions that address real human problems.