Y Combinator Co-founder Paul Graham: Most Standard Startup Advice Still Applies in the AI Era
Y Combinator co-founder Paul Graham stated that founders often ask how much standard startup advice is still applicable in the AI era, and that almost all of it is still relevant. The essence of entrepreneurship remains the same as before, but there is no guarantee that this will be the case in 10 years.
Graham pointed out that AI makes standard advice more valid in some respects. For example, YC often tells startups not to be conservative because they have little to lose; since they can act quickly, they should do so, and this principle is now more important than ever.
He further advised not to worry about having a moat, but to focus on growth. While lacking a moat is slightly disadvantageous, giving up growth opportunities due to a lack of a moat is far worse, as moats often form naturally in later stages.
Graham noted that one change is that founders no longer always need co-founders to share the actual workload, as AI can handle many tasks. However, the main role of a co-founder has never been just to help with work, but to share the pressure, which AI cannot replace.
These views appeared in his series of posts responding to questions raised by founders during YC talks. Graham has previously emphasized that a founder's "formidable" qualities (whether they can get what they want) are more important than specific ideas or industries, and AI has not changed this core principle.
Buyers are early founders seeking to adjust decisions in the AI environment; sellers are those providing experience within the YC system. The flow of capital and attention is shifting from a pure reliance on AI tools to a renewed focus on speed, growth, and psychological resilience, benefiting teams that can execute quickly and handle pressure, while disadvantaging projects that overly focus on early moats or completely rely on AI to replace interpersonal support.
Graham also mentioned in recent YC discussions that even among AI batches, the most impressive companies may not all be working on AI, and the founders themselves remain the strongest predictive indicators.
Source: Public Information
ABAB AI Insight
Paul Graham co-founded Y Combinator in 2005 with Jessica Livingston and others, standardizing early-stage startups through a batch incubation model and writing extensively about founder selection, speed, and growth. He has repeatedly emphasized that excellent founders are more important than specific ideas, and continues to apply this framework to new batches in the wake of AI, noting that even during the AI boom, the standout companies may not necessarily be AI projects.
In terms of capital pathways, Graham positions AI as a tool for accelerating execution rather than disrupting fundamental rules. YC directs resources toward "quick action" and "shared pressure" through office hours and public posts, rather than encouraging founders to reduce co-founders or overly pursue early moats due to AI. The motivation is to maintain the effectiveness of YC's selection and mentoring system, avoiding founders misinterpreting technological changes as rule obsolescence.
Drawing parallels to the early internet, Graham advocated for "doing first and thinking later" and "living in the future," when technology reduced building costs but did not eliminate the need for founder resilience. The current AI era is in a similar phase: tool costs are decreasing, but the structure of pressure and growth decision-making remains unchanged, and moats still depend on later scale rather than early design.
The essence is a localized occurrence of technological substitution: AI has replaced some execution work but has not replaced the core mechanisms of pressure sharing and decision-making speed. The mechanism is that as building costs decrease, marginal advantages shift toward execution speed and psychological resilience rather than static barriers; prematurely pursuing moats sacrifices growth windows, while lacking pressure-sharers amplifies individual decision volatility.
ABAB News · Cognitive Laws
- Tools reduce execution costs but do not reduce pressure costs.
- Growth opportunities take precedence over early moats.
- Standard advice often holds more validity under technological changes.