Lessons from Bob McGrew

Bob McGrew is the former Chief Research Officer at OpenAI and an early engineer at Palantir, McGrew's perspectives are highly sought after.

On the Future of AI and AGI

Quotes:

"From the outside, it feels like everything's accelerating. From the inside, it looks different." [1]
"We might ask an AI to build a 'cool product,' but we still have to define what 'cool' means. That part is still very human." [1]
"We solved a big chunk of reasoning. Now the challenge is scaling it, which is very hard—but it's the path forward." [1]
"Keep working on it. Progress won't slow—it'll just change direction." [1]
"2025 is going to be the year of reasoning." [2]
"I think we're going to continue to see capabilities increase. It's going to continue to feel like it's felt super fast, super exciting over the last even five years, and I think it's going to continue feeling like that. There's not a wall here." [2]
"If you asked people what AGI was [in 2018], they would say it's a model that you can actually interact with, it passes the Turing test, it can look at things, it can write code, it can even draw an image for you... and and and none of that is happening [in terms of mass job loss]." [3]
"The techniques that give that to you in terms of thinking harder, also apply to taking action you know in the real world in the virtual world... what we're going to see out of reasoning out of long thinking is that it's really going to unlock the possibility of agents to do actions on your behalf." [3]
"I think we're going to look back at this conversation... and science is only going like 30% faster than it was, why isn't it 300 times faster? And we'll have to figure it out... it'd be a great problem to have." [3]
"I think there's going to be two roles [in the future]: one will be something like lone genius... and the other role is manager. That you know you will be the CEO of your own firm and that firm will mostly be AI." [4]

Learnings:

AI Progress is Shifting, Not Slowing: McGrew predicts a shift from releasing bigger models every few months to focusing on new form factors, reliability, and deeper workflow integrations. [1]
Human Agency Remains Crucial: As AI handles more tasks, the distinctly human ability to define goals and what is valuable becomes even more important. [1]
AGI is a Continuum: He views Artificial General Intelligence not as a singular event but as a gradual and continuous process of development and scaling. [1]
Reasoning is the Next Frontier: The ability for AI models to "think" for longer periods (test-time compute) is a major area of advancement, unlocking more complex problem-solving. [1][5]
The "ChatGPT Moment" for Robotics is Coming: McGrew anticipates a breakthrough in robotics similar to what ChatGPT was for language models, though scaling physical hardware presents unique challenges. [3]
AI Adoption is Slower Than Expected: Despite the rapid advancement of AI capabilities, their integration into the broader economy and daily life is happening more slowly than many predicted. [3]
AI Agents Will Be Transformative: The development of AI agents that can take actions on our behalf is a key future application of reasoning models. [5][6]
The Future of Work is Collaboration with AI: McGrew envisions a future where humans work alongside AI, augmenting their capabilities rather than being wholly replaced. New jobs will emerge as others become obsolete. [5][6]
Scaling Laws are Foundational: The principle that increasing data and compute power leads to better model performance has been a driving force behind OpenAI's successes. [5]
From Scaling to Reasoning: The focus in AI development is moving beyond just making models bigger to incorporating more sophisticated reasoning capabilities. [5]

On Startups and Building Products

Quotes:

"If you're a founder the the right approach is to start with the very best model you can because you know your your startup is only going to be successful if it exploits some something about AI that realistically is going to be on the frontier." [3]
"The most important thing in a startup is actually your time... you don't want to be unless you have to you don't want to be like Palantir taking three years to get to market you want to be able to to build that product as quickly as possible." [3]
"A forward deployed engineer is someone, typically technical and engineer, who sits at the customer site and fills the gap between what the product does and what the customer needs." [5]
"The FD model effectively is doing things that don't scale at scale." [7]
"When we got started, the focus of our company was to build software for the intelligence community, specifically software for spies. And one of the challenges in building software for spies is that I don't know any spies." [5]
"The standard thing that you expect is that you spend a lot of time early on doing things that don't scale... and then you discover product market fit. Once you discover product market fit you do something entirely different... you want to embrace distance from your customer." [7]
"[With the FDE model] you're going to get pushed towards larger and larger contracts." [7]
"Early on it makes sense for the startup to just take on all the risk... you pay us if it works." [7]
"You're going to have to figure out a way past [the IT department]... this is part of why it matters that you're working on one of the CEO's top five problems." [7]
"The FD goes and builds like a a gravel road to where the product needs to go. And then the role of my team of the the product and engineering team was to look at that and basically figure out how that should generalize... and then turn that you know gravel road into like a paved superhighway." [7]

Learnings:

Start with the Best AI Model: For AI startups, leveraging the most advanced models available is crucial to finding a competitive edge. Optimization and cost-reduction can come later. [3]
The Forward Deployed Engineer (FDE) Model: This Palantir-pioneered strategy involves embedding engineers with customers to deeply understand their needs and build custom solutions, which then inform product development. [5][7]
FDEs as Product Discovery: The FDE model turns the traditional sales and product development process on its head by using engineers to find product-market fit directly with the customer. [7]
"Doing Things That Don't Scale" at Scale: The FDE model is a way to institutionalize the "do things that don't scale" mantra, allowing for deep customer integration even as the company grows. [7]
The Value of Taking on Risk: For startups selling to large enterprises, taking on the initial risk of a project can be a powerful way to win contracts, as large companies are often skeptical of new technologies. [7]
Solving Top-Level Problems: To bypass internal bureaucracy in large organizations, startups should focus on solving problems that are a top priority for the CEO. [7]
The "Gravel Road" to a "Superhighway": The FDE creates a quick, custom solution (the gravel road), which the core product team then generalizes and refines into a scalable feature (the superhighway). [7]
The FDE Model is Not Consulting: While it involves custom work, the goal is to feed insights back into a scalable software product, leading to increasing profit margins over time with each customer. [7]
AI Startups are Reviving the FDE Model: Because AI agents are a new category of product, there is a significant need for product discovery within enterprises, making the FDE model highly relevant today. [7]
Pricing Based on Outcomes: The FDE model lends itself to pricing based on the value delivered to the customer, rather than traditional SaaS metrics like seats or usage. [7]

On Management and People

Quotes:

"The core thing that you have to do as a manager is you have to really care about the people you're managing." [8]
"There comes a time as a manager when you have to ask someone to do something hard... if they know that you are doing what's best for them, then when you tell them to do that thing that is super hard and extremely scary for them, sometimes you can help them across the chasm." [8]
"Very talented people have superpowers, but they also have debilitating weaknesses... and for for people who are at the very edge of these capabilities they often don't even understand what their weaknesses are but it's it's it's extremely apparent to everyone around them." [8]
"When people fail... it's almost always a form of self-destruction." [8]
"Even if you're firing someone... if they're not going to succeed in this role... then it is in their own best interests for me to tell them that they're not succeeding and give them the opportunity to find somewhere else." [8]
"Loyalty in the end is the thing that I think unlocks all of the other things that you want in management." [8]

Learnings:

Genuine Care is the Foundation of Management: The most critical aspect of being a manager is truly caring about the well-being and success of your team members. [8]
Loyalty Enables Difficult Conversations: When your team trusts that you have their best interests at heart, you can guide them through challenging but necessary decisions. [8]
Managing Superstars Requires Understanding Their Weaknesses: Highly talented individuals often have significant blind spots that are obvious to others. A manager's role is to help them see and navigate these weaknesses. [8]
Failure is Often Self-Inflicted: At the highest levels of capability, career-altering failures often stem from an unwillingness to confront a difficult personal or professional challenge. A good manager can help prevent this. [8]

Learn more:

Lessons from Bob McGrew

On the Future of AI and AGI

On Startups and Building Products

On Management and People

Written by Antoine Buteau

Lessons from Shyam Sankar

Lessons from David Holtz