TL;DR - We developed an AI based methodology to create 10K high-quality startup ideas. Find 3 that could be your next company: Get your next idea - Lev AI

Coming up with any idea is hard but coming up with an idea that has a high likelihood of being successful is tremendously difficult. We all know the stats - 60%+ of tech companies fail in 5 years and 90% in ten years. A lot of the failure rate is driven by execution after coming up with an idea - is there a real market need, can you outcompete other players, and so on. But we wondered, could we use AI to frontload the work so the ideas we started off with already had a high likelihood of success?

As a baseline, we began asking the models to generate good venture scale ideas. The result? Pretty generic. No matter if you ask for 5, 10, or even 50 ideas, we saw that they were all based on the same concept, endless clones of “AI for X”.

Unsurprisingly, we needed to get more sophisticated. Treating the models like a junior PSL team member, we came up with a highly structured diverge and converge methodology. With this approach, we saw a huge uplift in quality of ideas - matching and exceeding our own analysis internally.

Our Methodology

Forcing a wide variety of ideas or concepts and then landing on a few (diverge & converge) is a generally widely accepted framework in design thinking. We just wanted to do it at scale. We developed a systematic method to diverging - we took 150+ industry verticals, 200+ job titles, and 50+ workflows with 10 advanced combination templates to come up with ideas. Our templates encoded proven patterns of how problems manifest in real businesses, but with a few safeguards against typical or generic ideas. For example, we’ve hard-blocked cliché outputs like “CSV converter” or “URL shortener”.

With the volume we were working on, there was a high chance the models continued to produce clones that would be low value. To combat this, we ran automated quality control. We used Levenshtein distance combined with similarity ratio checks to flag any ideas that are “too close” to each other. This isn’t just exact-match deduplication; it’s fuzzy similarity detection that can catch when two outputs are functionally the same despite superficial differences. For example, “AI that drafts sales emails” vs. “AI that writes outbound prospecting messages” would be treated as near-duplicates, with only the better-scoring variant carried forward. This guaranteed that every surviving idea represents a distinct opportunity. Using our structured divergence, we generated 160K potential ideas.

Now we needed to converge. Many times in ideation sessions, teams converge by voting on ideas. Inherent in each person’s vote is their criteria for what a “good idea” looks like. To enable our system to evaluate consistently at scale, we decided to encode our own criteria into our system - informed by our experience vetting 500+ ideas and launching 40+ venture scale start-ups. That includes everything from how acute the customer pain is to the unit economics to the experience of the founding team. While we started with the venture scale rubric, we know that not every founder is targeting a venture scale outcome. So, we have customized our rubric for different use cases, including a bootstrap rubric that focuses on ideas that are immediately executable and cash flow positive for a solo founder of small teams.

We implemented convergence by having Sonnet take each idea it generated, identify the relevant rubric (i.e. venture scale or bootstrapped), and continue to think on each idea until it was best in class across every single criterion. We selected the top 10K for additional analysis. We then used Opus, a much more advanced albeit expensive model, to go deep into strengths, weaknesses and risks. We redid the scoring and ranked our ideas. Because of our sophisticated rubrics, our system was able to conduct more intelligent ranking, surfacing better quality ideas compared to the single evaluation lens that chatGPT or other idea generators use. At the end of this process, our system generated 10K diverse & high quality and investable ideas.

Making it available

You don't just have to take our word for it! There are far more strong opportunities here than any single team could pursue, so we are opening up the ideas to anyone. You can go to this platform and get access to 3 tailored ideas based on your experience and/or interests. Check out your matched ideas and tell us how we did!

Our takeaways

AI automating our process - like sizing the market or doing competitive analysis - was table stakes. We have been using the models to do that for a while. But, to have AI lean into the creative side, to take what we thought was part of our secret sauce and churn out nearly 100x the volume for less than 10% of the cost, is incredible. We spent $5K to get to 10K high quality ideas, but that was because we started with a massive funnel. Individual founders who are starting with a smaller possibility set based on their background, can do this for much cheaper. As we see more and more individuals become founders with the support of AI, we are incredibly excited for the potential increase in velocity and successful outcomes from methodologies like ours that focus on upfront ideation and validation.

Ready to take your matched idea further? Here's how we can help:

For venture-scale ambitions with full support: PSL knows what it takes to get an idea from concept to Series A. When we work with founders, we bring our network, operational expertise, funding, and hands-on support through the 0-to-1 journey. If you want PSL as your partner, let’s talk! Reach out to shilpa@psl.com.

For founders who want to move fast independently: We’re launching Lev, an AI co-founder platform. For founders looking to make progress on their own timeline or who already have a team in place, Lev will help you rapidly test assumptions, build key assets, and go from 0 to 1 faster. We’re in alpha now and waitlisting for access. If you’re interested in trying out Lev, sign up here.

‍

About Us

Pioneer Square Labs (PSL) is a Seattle-based startup studio and venture capital fund. We partner with exceptional founders to build the next generation of world-changing companies, combining innovative ideas, expert guidance, and investment capital. PSL operates through two primary arms: PSL Studio, which focuses on creating new startups from scratch, and PSL Ventures, which invests in early-stage companies. Our mission is to drive innovation and growth by providing the necessary resources and support to turn big ideas into successful, impactful businesses. If you have a groundbreaking vision, connect with us hello@psl.com, and let’s build something extraordinary.