Why Ambitious Startup Ideas Are Actually Easier To Sell

Predictions and Expectations:

• The speaker expects the current AI harness to evolve beyond its current state, noting the current state is "not good" and "not where it will end."
• The speaker predicts that at some future stage, the scenario of a human needing to manually merge a pull request generated by an AI will become obsolete ("hopefully that isn't a thing anymore in the future").
• The speaker believes that building software products that already have competition is the path forward in the AI world, rather than creating entirely new product categories.
• The speaker anticipates that the role of a product manager will shift to "configuring the software" and setting rules, while the role of a support team will become "supporting the agents to do their job" rather than handling tickets directly.
• The speaker fears that without a robust "intent harness," AI might build a product with "a thousand tiny features" that do not fit together into a "holistic product vision."
• The speaker expects that if a startup does not ship quickly, they are "more likely to screw up" due to the ambition of the idea.
• The speaker predicts that the "US culture" of aiming high will become more normalized, whereas the European tendency to fear failure or over-analyze downside risks is a barrier to be overcome.
• The speaker believes that investors will continue to focus on "what if it all works out" rather than downside risk, noting that 90% of returns in their fund come from just five companies.

Timelines and Milestones:

• The speaker plans to give a talk "later today" about building AI-native companies and recursive loops.
• The basic AI "list" (likely "loop" or "agent system") was in production "as of last week."
• The strategic pivot to become an "AI-first" company began "probably about a year and a half ago."
• The speaker expects the company size to be "more than doubling" within "the same year" as the strategic pivot.
• The speaker expects a product vision to be sustained for "40 more years" to ensure founders do not burn out.

Technology and Product Direction:

• The speaker plans to "ship pull requests" that fix simpler engineering tasks automatically by analyzing logs, errors, and session recordings.
• The speaker expects the system to "feed it back into the world and then sees how it impacts the world and just kind of iterates on that loop."
• The speaker plans to grab "intent data" from various sources, including "support ticket[s]," "video clip[s] of a user," and "what you're talking about in slack."
• The speaker expects to use "trace data for AI first products" to "backward test changes to a product" without waiting for human iteration.
• The speaker plans to build a "harness" that allows the AI to ask clarifying questions (like a human product manager) rather than just generating code immediately.
• The speaker expects the AI to "unblock itself" by using humans as a resource only when necessary.
• The speaker plans to have the AI "instruments your product to collect data" and "send emails" or "do analytics."

Market and Industry Outlook:

• The speaker believes that "the more ambitious" the idea, the "easier" it will be to go to market because it is more "remarkable" in a saturated software landscape.
• The speaker predicts that fundraising will be "easier... for very large ideas" with a "bimodal" outcome where strong investors prefer high-risk/high-reward bets.
• The speaker expects that "everyone is much earlier than we thought they would be" and that the startup landscape is "more chaotic than you would expect."
• The speaker believes that "very strong people" are more likely to want to work for companies "working on something very inspiring" rather than narrow solutions.

Company Plans:

• The speaker plans to hire "really good people" by embracing "very ambitious" goals, noting that "the more ambitious we've gotten... it's been far easier" to hire.
• The speaker plans to shift from building 20 products to building "hundreds" of products that act as tools for an AI agent.
• The speaker expects to "split-brain" as a founder, holding a "long-term intense ambition" while directing the company toward "short-term practicality" to ensure shipping.
• The speaker plans to continue "iterating and improving" the manual and rule book for the support agents, becoming a "philosopher for support."
• The speaker expects the company to grow significantly, noting they are "scaling up quite a lot" and "doubling the size of the company."

Financial Guidance:

• The speaker predicts that investing in a company with a "50 million or 100 million outcome" is "an absolute waste of time" for investors.
• The speaker expects that investing in something with "a one percent chance of being a 10 billion or 100 billion company" is the standard model for generating returns, even if "99% of the time it fails."
• The speaker believes that "90% of our total returns come from five companies" out of 7,000 invested.

Risks and Caveats:

• The speaker warns that if a founder focuses only on the "long-term intense ambition" without "short-term practicality," they risk failure.
• The speaker notes that "it doesn't question" or "talk to you" with current AI tools like Claude, which is a limitation in building a "holistic product vision."
• The speaker cites "imposter syndrome" as a risk for European founders who worry about what "friends or family" will think or if the product is "GDPR compliant" in early stages.
• The speaker hedges on the success of curing disease or cancer, stating "if it did work I mean that would be pretty impactful" but noting "probably it won't work" for such bold claims.
• The speaker admits the first version of PostHog was "super terrible compared to all of our competition" and that they were "actively bad product" founders initially.

Confidence and Disagreement:

• The speaker is "very confident" they will have "product market fit" if they can "accelerate" existing jobs to a "better standard."
• The speaker is "super confident" they can build something "superior to a product manager" regarding understanding customer data across sales calls and meetings.
• The speaker expresses strong belief that "there isn't a ceiling to where you can think" regarding ambition.
• The speaker states they were "probably too much on the side of abandoning things early" during pivots but believes this was necessary to avoid wasting time on "unnatural" ideas.