Mar 4, 2025 · 22m · we-live-to-build
Founder Market Fit Got Them $4.2M Before a Product Existed
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this interview, Thesis co-founder Rabbi Guha joins host Sean Weisbrot to reveal how deep domain expertise enabled his team to raise a $4.2 million pre-product seed round, while detailing their vision for replacing static text chat with dynamic generative UI across enterprise software.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Sean holds 30.4% of the talking time here. How this is scored →
Sean as informed peer 4.3 Guest teaching 5.4 Guest disagreement 1.7 Sean pushing back 2.7
speaking balance: gold is Sean, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 16:29 Pushback against voice-only UI prediction
Rabbi directly challenges Sean's assertion that voice will strip away all visual UI, using Jira scrum boards and complex decision-making as concrete counterexamples.
Hardest push from Sean ▶ 19:06 Host challenges generative UI practicalitySean pushes back on Rabbi's thesis, stating that dynamically generating unique UI on the fly for each user feels funky and difficult to picture in production.
Biggest teaching moment ▶ 10:00 VC power law and portfolio probability breakdownRabbi educates the host on why VCs reject revenue-generating businesses by contrasting a 10% shot at a billion-dollar outcome against a guaranteed hundred-million-dollar outcome.
Sean holds their own ▶ 9:02 Host demonstrates personal fundraising backgroundSean demonstrates domain knowledge by detailing his personal experience raising VC funding, angel investing, and hitting investor revenue requirements.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Sean as informed peer | Guest teaching | Guest disagreement | Sean pushing back | Why |
|---|---|---|---|---|---|---|
| The Core Fundamentals of Founder-Market Fit | 3 | 4 | 1 | 1 | Sean asks standard interview questions regarding Rabbi's background and how Thesis raised $4.2M pre-product. Rabbi provides a clear explanation of founder-market fit and the importance of warm network introductions. | |
| Leveraging Deep Domain Expertise and Industry Experience | 4 | 5 | 1 | 2 | Sean cites his observations from Reddit and LinkedIn regarding early-stage founder struggles to validate why domain trust matters. Rabbi expands by advising junior founders to gain deep vertical domain experience in corporates or startups before launching. | |
| Building Cross-Border Operations Between India and San Francisco | 4 | 5 | 2 | 3 | Sean brings up perceived Silicon Valley bias against non-white or Indian founders and questions if opening an SF office was forced by VCs. Rabbi gently reframes the issue, explaining cross-border R&D dynamics and clarifying that moving to SF was a proactive strategic decision. | |
| Navigating the Venture Capital Mindset and Math | 5 | 6 | 1 | 2 | Sean references his personal experience raising VC funds and angel investing to contrast early-stage VC expectations with founder realities. Rabbi details the mathematical mindset of venture capital, explaining the preference for a 10% chance at a billion-dollar outcome over safe smaller returns. | |
| Mid-Show Call to Action and Channel Support | 6 | 6 | 4 | 5 | Sean presents an argument based on recent CRM investments and tools like Bordy that voice interfaces will replace graphical UI. Rabbi pushes back directly, arguing multimodal UI and structured visual data remain essential for workflows like shopping and task tracking. | |
| Generative UI Architecture and Thesis Product Roadmap | 4 | 6 | 2 | 4 | Sean expresses skepticism about on-the-fly generative UI, labeling it funky and hard to visualize. Rabbi explains Thesis's generative canvas and how LLMs can dynamically output domain-appropriate data visualizations rather than plain text. | |
| Strategies for Reducing AI Compute Costs and Model Distillation | 4 | 6 | 1 | 2 | Sean raises practical concerns regarding high server and GPU compute costs for generative UI models. Rabbi outlines strategies for developers, emphasizing task-specific evaluation benchmarks and model distillation to maintain quality while reducing inference expense. |
Statements from this episode (10)
Guha: Thesis raised a $4.2M seed round without an existing product
“Especially at our stage, the seed stage, when we raised, we didn't have a product. We just had an idea and we wanted to bring this in front of people.”
Guha: VCs hesitate on India-based founders, not Indian identity
“There is obviously, I'll be honest that there is a hint of hesitation among the Silicon Valley VCs for founders who are based out of India, but not so much like Indians.”
Guha: VCs prefer a 10% chance at $1B over guaranteed $100M
“One of them being that you being able to convince them that this is a VC business. It means that this is, this has the potential to become a billion dollar company at some point. Even if it is like it's a 10% chance, that's what they are probably looking for. …”
Guha: VCs withhold critical feedback from founders to avoid burning bridges
“From what I understand talking to VCs is that they are also not trying to burn bridges. So even if they have some feedback for you or they don't think you will be able to take it positively, they'll hold it back.”
Guha: Thesis splits its seed budget evenly between R&D and go-to-market
“Since we are the seed stage right now, that is split almost 50 50 between R&D and go to market.”
Guha: No foundational AI models are built specifically for UX design
“There are some good examples out there, for example, Claude and other stuff, but there are no fundamental models that are just built on specifically doing UX and UX well.”
Guha: AI apps must match traditional SaaS usability to drive adoption
“You really have to think about the UX. If you want your customers to love it and switch from their old SaaS based workflows, you have to bring to them the same level of engagement, the same level of, you know, user friendliness that they were used to.”
Weisbrot: The AI agent Bordy autonomously raised an $8M funding round
“The AI ended up raising eight million dollars for itself without the human's feedback until it got time to like, get serious about legal. And then it was like, oh, by the way, I've raised money for you. Like you need to sign the papers.”
Guha: Thesis trains LLMs to dynamically generate UI on the fly
“What we do is basically train the LLM to understand these design patterns. And instead of you having to manually encode them into your application or manually hard code them into your application, You can just use RLLM to generate it on the fly and just engage…”