How Firebase co-founder Andrew Lee pivoted his way to a $175M valuation and why Big Tech threats shouldn't scare you.

Building a massive venture-scale business requires finding acute customer pain, even if it means pivoting away from years of work. Serial founder Andrew Lee shares how he built and sold Firebase to Google, ran into the realities of corporate inertia, and returned to Y Combinator to build Tasklet, an AI command center burning millions on inference.

We cover why investor hype follows extreme growth rates, why incumbents move slower than you think, and how to stay ahead of model providers like OpenAI and Anthropic.

What you will learn:

• How a customer development call transformed a struggling chat app into Firebase
• Why Google's threats to build competing products are usually empty
• The exact metrics Tasklet used to raise at a $175M valuation before Demo Day
• How to avoid getting wiped out by frontier models from Anthropic and OpenAI
• Why 100% month-over-month growth rate trumps total run rate every single time

📌 Watch now and subscribe for more smart, fast, founder-first interviews.

Chapters:

00:00 $30M Inference Teaser

00:14 Meet Andrew Lee

02:25 From Chat to Firebase

05:31 Demo Day Misstep

08:09 Early Checks and Runway

10:45 Why Return to YC

18:34 Tasklet Explained

21:47 Beyond Chat Interfaces

23:05 Future of Software Companies

24:11 YC Roaster Ad Break

24:51 Next Big Question

24:56 Competing With Claude

25:51 Why Tasklet Wins

26:55 Model Marketplace Bet

28:30 Traction Growth Advice

29:18 Painkiller Problems

31:41 Idea Maze Grind

32:46 Timing The Pivot

34:45 Raising Early In YC

38:21 Bigco Threat Reality

39:15 Google Acquisition Lessons

42:24 Carving Niches To Win

45:37 Founder Focus Exception

47:12 Stay Out Of Their Way

48:35 Wrap Up And Links

🎧 Listen on the go:

Spotify: https://open.spotify.com/show/1u1JyRKH8JYFjhBkvTUOAP

Apple: https://apple.co/4cZ8RMm

My other YouTube Channel: https://www.youtube.com/@lobstercap

This Lobster Talks episode is brought to you by YC Roaster, a free tool helping founders improve their YC applications.

🔗 https://ycroaster.com

AI-native startups are no longer selling software to incumbents; they are replacing them completely to capture full margin.

Vincent, CEO of Penta, breaks down how he abandoned traditional SaaS to build an AI-native commercial insurance brokerage.

We explore why legacy brokers are anti-efficiency, how AI collapses back-office overhead by 90%, and why lean teams of two founders can comfortably scale past $10M ARR.

What you will learn:

- Why traditional SaaS sales fail in legacy industries and how full-stack replacement wins.

- How AI natively shrinks back-office headcount to 1/10th of traditional insurance operations.

- The exact unit economics behind 3x productivity gains per headcount.

- Why legacy carriers are structurally disincentivized from adopting efficiency tools.

- How to leverage AI loops to replace manual human workflows across professional services.

📌 Watch now and subscribe for more smart, fast, founder-first interviews.

Chapters:

00:00 AI Replaces Workers

00:29 Meet Penta Founder

01:16 Why Brokers Reject SaaS

02:52 What AI Native Means

05:17 Margins and Distribution

07:27 Legacy Company Playbook

09:48 Self Improving AI Loops

10:50 Next Industries to Disrupt

13:11 Corgi and Market Hype

14:38 How Brokers Work Today

16:43 Scaling Back Office with AI

18:35 Can Carriers Go AI Native

21:02 Regulation Blocks Efficiency

21:39 Insurance Pricing Reality

21:58 Middleman Speed Problem

22:34 AI Fixes Broker Errors

23:31 Hiring and Getting Covered

23:49 Tiny Teams Big Revenue

26:28 Broker Model vs SaaS

27:15 AI Native Unit Economics

29:58 Jobs Displaced by AI

30:53 Humans as Bot Workarounds

32:44 Abundance vs Job Loss

34:10 Hacker House to YC

38:17 Finding Penta Online

38:35 Final Thanks and Wrap

🎧 Listen on the go:

Spotify: https://open.spotify.com/show/1u1JyRKH8JYFjhBkvTUOAP

Apple: https://apple.co/4cZ8RMm

My other YouTube Channel: https://www.youtube.com/@lobstercap

This Lobster Talks episode is brought to you by YC Roaster, a free tool helping founders improve their YC applications.

🔗 https://ycroaster.com

This YC biotech startup built a genetic kill switch to end cancer. Here is how 3 Oxford PhDs cracked the platform.

FinalDose is taking on a $5 trillion market by shifting from traditional search-and-edit gene therapies to search-and-destroy target architectures. By combining CRISPR shape-change mechanics with frontier AI models, they prove how hardware constraints dissolve when software timelines hit deeptech.

We sit down with the founders to breakdown their foundational patent, YC batch execution, and the timeline to commercialization.

What you will learn:

- Why 60 years of targeted protein pharma missed the underlying DNA driver signature

- How FinalDose engineered a Cas9 conformational shape change into a cellular kill switch

- The exact framework used to validate their platform in just 40 days during YC

- How AI models collapse drug discovery costs by 10X and accelerate timelines by 3X

- The delivery vectors needed to clear targeted organs like the liver, lung, and skin

- How mutation burden per year dictates the biological ceiling for human longevity

📌 Watch now and subscribe for more smart, fast, founder-first interviews.

00:00 Intro 00:18 The $5 Trillion Thesis: Ending Cancer 02:11 Why 60 Years of Pharma Targeted the Wrong Mechanism 04:08 The Eureka Moment: Turning CRISPR into a Kill Switch 06:12 The AI Unlock: Harnessing Conformational Changes 08:04 How YC Forced a Bio Startup onto Software Timelines 10:14 Defending a Foundational Patent Against Big Pharma 13:16 Solving the Specificity and Delivery Bottlenecks 16:11 Can We Delete DNA Damage to Live 300 Years? 19:15 10x Cheaper, 3x Faster: The New Bio Playbook

🎧 Listen on the go:

Spotify: https://open.spotify.com/show/1u1JyRKH8JYFjhBkvTUOAP

Apple: https://apple.co/4cZ8RMm

My other YouTube Channel: https://www.youtube.com/@lobstercap

This Lobster Talks episode is brought to you by YC Roaster, a free tool helping founders improve their YC applications.

🔗 https://ycroaster.com