Gokul Rajaram is one of the most prolific product builders of the last 20 years, having built core ads and product businesses at Google, Facebook, Square, and DoorDash during their most formative scaling periods. Alongside his operating career, Gokul has invested in more than 700 companies through his new firm Marathon, giving him an unusually broad view into how products are built and scaled.

The conversation explores how product building is fundamentally changing in the AI era, from the collapse of traditional PM/designer/engineer roles to the emergence of non-deterministic software requiring constant evaluation. Gokul shares insights on building durable AI applications, the three ways ad businesses make money, and lessons from working closely with Larry Page, Sergey Brin, Mark Zuckerberg, Jack Dorsey, and Tony Xu as they built generational companies.

The Death of Traditional Product Development Roles

Product development roles are merging as AI capabilities advance rapidly - product managers now 'sit with engineers and researchers and write code, do prototypes' rather than just defining requirements

Companies are choosing engineers over designers, with the ratio shifting from 1:3 or 1:10 to 1:20 as 'AI can leverage the design language to create designs'

Non-deterministic software requires constant evaluation - 'you could do X, Y happens, but if you do slight variation of X, something completely different happens' - Gokul

All companies now include explicit 'prototyping interviews' in their hiring process, forcing product managers to demonstrate hands-on building capabilities

Building Durable AI Applications in the Foundation Model Era

Target high-value workflows requiring custom data because 'if the CIO of your target customer can build what you're building' using foundation model tools, startups become obsolete

Five sources of durability: ownership of scarce assets, control points for money/data interaction, hardware dependencies, essential workflow integration, and network effects

AI agent companies must 'replace the entire system' because legacy platforms like Slack are cutting off API access to prevent value extraction from incumbent systems

Migration tools require 1-2 years of engineering investment - one portfolio company 'hired engineers in Eastern European country for two years to build this migration thing'

The Three Laws of Ad Business Success

First way: Own coveted user inventory with interaction surfaces - Google Search provides high intent, Facebook provides precise identity targeting capabilities

Second way: Drive specific outcomes without owning inventory - AppLovin built a $100+ billion company by mastering mobile app installs across the entire ecosystem

Third way: Become exclusive provider for premium inventory - Trade Desk handles display budgets that don't go to Google or Facebook

ChatGPT represents the 'dream of any advertising person' by combining Google's intent data with Facebook's identity data in natural language queries

Which Software Companies Should Fear AI Disruption

Seat-based pricing models are most vulnerable - Zendesk charges per seat while 'AI agents can sit right next to Zendesk' and reduce seat requirements over time

Data-based systems are more protected - NetSuite as ERP is 'career-limiting to suddenly take out' because it runs entire businesses with timeless data

Companies must shift from utility-based to outcome-based pricing, changing from '$20-30 per seat to maybe 50 cents or 20 cents per ticket result'

Many legacy software companies 'probably need to go private because they have to make this business model transformation' away from public market scrutiny

Leadership Lessons from Tech's Greatest CEOs

Larry Page and Sergey Brin focused on technological superiority and scale - Larry asked about AdSense: 'what percentage of all ads on the internet are you? Less than 1%'

Eric Schmidt's image-only strategy presentations forced teams to distill complex ideas - 'people don't remember words. They remember how things made them feel'

Mark Zuckerberg invented Facebook's custom audiences by connecting disparate domains - asking 'why can't they just upload their whales into our system?' for gaming advertisers

Jack Dorsey called product managers 'product editors' because 'the role is not to add more features' but to edit down to what truly matters

The Self-Serve Product Philosophy

Larry Page mandated that everything built for large customers 'is also available to small customers' after discovering internal tools weren't accessible to self-serve users

Self-serve definition: customers can 'onboard and use the product without ever talking to or engaging with a single member of the employee base'

Small customers often exploit advanced features faster than large customers, teaching companies about their product's true capabilities through creative usage

Self-serve opens aperture from reaching '10,000 customers with 100 salespeople' to 'millions of customers' through word-of-mouth and bottom-up adoption

Hiring and Career Strategy in the AI Era

Work projects are essential because 'everywhere else, you can just BS your way without doing stuff' - Square's CorpDev candidates had to recommend acquisition targets

Best candidates 'rejected the premise completely' and conducted independent customer research rather than following instructions blindly

Span of control must exceed 10 people - 'if you're managing even 15 people, maybe you meet with them once a week. That's 15 hours. What do you do for the other 25, 30, 40 hours?'

Job hopping for 12-18 months is 'immediate red flag' because 'you can't achieve anything of value' or 'have any impact on a company' in that timeframe