Jan 11, 2024 · 1h 27m · lennys-podcast
How to be more innovative | Sam Schillace (Microsoft deputy CTO, creator of Google Docs)
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
Microsoft Deputy CTO and Google Docs creator Sam Schillace joins Lenny Rachitsky to discuss the mechanics of disruptive innovation, the paradigm shifts driven by generative AI, and counterintuitive strategies for engineering and career growth. Schillace shares hard-won lessons from pioneering early web applications to building multi-agent systems, urging builders to leverage first-principles thinking, lower experimentation costs, and lean into their innate strengths.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Lenny holds 17.6% of the talking time here. How this is scored →
Lenny as informed peer 3.6 Guest teaching 5.3 Guest disagreement 0.9 Lenny pushing back 0.0
speaking balance: gold is Lenny, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 1:08:45 Dismissing media cynicism on Gemini and AI
Sam forcefully rejects the tech press commentary on Gemini matching GPT-4, calling it the single dumbest possible interpretation of the ongoing AI revolution.
Hardest push from Lenny ▶ 58:25 Lenny brings up economic reality and career privilegeLenny gently challenges Sam's advice on only doing fun work by pointing out that most people work out of economic necessity and cannot easily abandon unpleasant jobs.
Biggest teaching moment ▶ 1:03:50 The zero-cost pixels paradigm shiftSam reframes the entire AI landscape for Lenny, demonstrating how generative models will commoditize pixel generation in the exact way the internet commoditized distribution.
Lenny holds their own ▶ 57:10 Lenny connects asymmetric bets to Charlie Munger's AlmanacLenny synthesizes Sam's career philosophy with Charlie Munger and Warren Buffett's principle of making massive, concentrated bets when holding an unfair advantage.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Lenny as informed peer | Guest teaching | Guest disagreement | Lenny pushing back | Why |
|---|---|---|---|---|---|---|
| Hook: Embracing What Feels Easy and Enjoyable | 0 | 0 | 0 | 0 | Teaser quote from Sam followed by Lenny's introductory monologue, guest bio, and sponsor reads. There is no live interaction between host and guest. | |
| The First Google Doc and Sneaky Infrastructure Migration | 4 | 4 | 0 | 0 | Lenny kicks off with trivia about the very first Google Doc file. Sam explains the technical architecture migration history and sneaked migration past Sergey Brin. | |
| The Mechanics of Disruptive Innovation and Polarizing Reactions | 4 | 6 | 2 | 0 | Lenny asks how to cultivate disruptive innovation and references Steve Jobs on first principles. Sam explains why disruptive tech polarizes audiences and dismisses pessimistic why not objections. | |
| Applying First Principles Thinking to Generative AI and Aerospace | 5 | 5 | 1 | 0 | Lenny breaks down first principles reasoning with Elon Musk examples. Sam elaborates on framing what if questions before deconstructing technical hurdles. | |
| Cultivating Optimism and Lowering the Cost of Experimentation | 4 | 6 | 1 | 0 | Lenny prompts Sam on why optimism is a strategic advantage. Sam details how lowering the cost of experimentation enables teams to discover non-obvious breakthrough value. | |
| User Laziness as a Metric: Evaluating AI vs. Crypto | 4 | 6 | 3 | 0 | Lenny raises crypto as an example of hype cycles. Sam forcefully critiques crypto for lacking core end-user value and defines user laziness as the ultimate product test. | |
| Reducing Onboarding Friction and the Innovator's Dilemma | 5 | 6 | 1 | 0 | Lenny highlights how Excel feature bloat created an opening for Google Docs. Sam explains the Innovator's Dilemma dynamics and reducing onboarding friction in early Writely. | |
| The Technical Genesis of Rightly and Solving Document Collaboration | 4 | 6 | 1 | 0 | Lenny asks about the timeline to product-market fit. Sam shares the technical engineering hurdle of three-way merging before operational transforms existed in browsers. | |
| Battling Corporate Resistance and Balancing Vision with Feedback | 3 | 6 | 2 | 0 | Sam recounts internal hostility at Google where hardware was denied, requiring threats of escalation to Eric Schmidt, emphasizing balancing mission tenacity with user feedback. | |
| Validating Product-Market Fit and Imagining Interactive Documents | 3 | 7 | 1 | 0 | Sam describes early traction, meeting Michael Arrington, and lays out a bold vision where static documents are replaced by interactive, conversational AI agents. | |
| Building on the Bare Metal: Pre-jQuery Web Engineering | 5 | 6 | 0 | 0 | Lenny references early TechCrunch coverage of Ajax. Sam reminisces about DOM hacking before jQuery existed and circumventing Google JavaScript linting standards. | |
| Guiding Engineering Exploration with Practical North Stars | 4 | 6 | 1 | 0 | Lenny asks how engineers should balance technological curiosity with problem-solving. Sam explains using concrete North Star challenges like writing VI in Python with multi-agent systems. | |
| Overcoming the Fear of Failure and Practicing 'Virtue from Error' | 3 | 6 | 2 | 0 | Sam diagnoses why younger engineers exhibit pessimism and risk aversion, contrasting curated student trajectories with unstructured exploration and framing his motto of virtue from error. | |
| Asymmetric Career Bets and Embracing Natural Strengths | 4 | 5 | 0 | 0 | Lenny connects Sam's career advice on pursuing high-leverage strengths to Charlie Munger's philosophy of betting big on unfair advantages. | |
| Navigating Compromise, Transitioning to Management, and Channeling Intensity | 4 | 5 | 0 | 0 | Lenny raises Seth Godin's principle of giving full effort in every role. Sam discusses transitioning into engineering leadership and learning to channel personal intensity. | |
| AI as a Paradigm Shift: Zero-Cost Pixels and Dynamic Intent | 4 | 7 | 2 | 0 | Sam outlines his thesis that AI makes pixels free just like the internet made information distribution free, mocking tech media dismissals of Gemini. | |
| Actionable Strategies for Mastering AI: Setting Targeted Goals | 4 | 5 | 1 | 0 | Lenny asks how practitioners can master AI without drowning in hype. Sam recommends picking ambitious specific tasks rather than generic open-ended experimentation. | |
| Behind Microsoft's Innovation Engine: Satya Nadella and Cultural Humility | 3 | 6 | 0 | 0 | Lenny asks what makes modern Microsoft innovate so effectively. Sam praises Satya Nadella's energy-raising leadership and Microsoft's cultural humility. | |
| Parting Wisdom on Agency: Building Without Waiting for Permission | 4 | 4 | 0 | 0 | Lightning round covers favorite books, media, and an infamous interview question about counting zeros in factorials. | |
| The Legendary Tale of Selling Blood Sausage to Google | 3 | 5 | 0 | 0 | Lenny brings up the legendary story of selling 200 pounds of blood sausage from Bocalone alongside Writely to Google. | |
| Episode Conclusion: Connecting with Sam and Final Takeaways | 2 | 4 | 1 | 0 | Sam shares how listeners can contact him on Substack and issues a filter that he only wants to review truly disruptive, non-incremental ideas. |
Statements from this episode (21)
Schillace: Writely originally ran on C# on three rented Windows servers
“We started off, in, in C Sharp, which is little known, in, you know, our own, there's, like, pre-cloud, so we, like, had three file servers that we rented that were Windows machines in a data center in Texas with a sysadmin in the Philippines running them. So …”
Schillace: Writely migrated onto Google infrastructure during an unnoticed 8-hour window
“When we migrated in to Google, we were very sneaky about it. And we like put the site into quote maintenance mode for eight hours on a Sunday where everything was just read only. And then we migrated all the data and moved everything and brought the new system…”
Schillace: Google executives wanted to kill Google Docs despite millions of users
“Like in the early days, like we had like, I don't know, a couple million users, five million users. And there were still executives at Google telling me that it was a stupid idea and that it should stop. And like, we shouldn't be doing it.”
Schillace: Writely co-founders initially rejected building Google Docs in browsers
“When I had the idea for Rightly, like, my two co-founders, who were both deep domain experts in both app building and word processors, were like, the browser's never going to support this. It's a bad idea. Let's not do this.”
Schillace: Giving multi-agent AI systems shared whiteboard memory makes them smarter
“We were trying to get multiple versions of them working together, like multi-agents working together, and we gave them whiteboard Working memory, like as a shared working memory thing to fix this problem, and that turns out to make them much smarter.”
Schillace: Crypto lacks meaningful user value even in the best case
“For crypto, I was always like, well, what if it works? Like, then I have to run OPSEC on my personal finances. That sounds dystopian. I don't want that. Like, I can't think of anything as a user that I think actually, like, is valuable here. Even in the best c…”
Schillace: Writely initially let users edit documents without an account or email
“When we did rightly at the beginning, we didn't even ask for an email address because it was such a novel thing. We didn't want any friction at all in the onboarding process. So you could just come in and make a document and start using it without telling us a…”
Schillace: Google Docs won market share, but Microsoft kept the money
“In retrospect, they did fine. Like, we took a bunch of market share, but they kept all the money, basically. So, you know, we being Google, like, you know, so, like, they survived it. Like, you know, they survived. They did a good job surviving the challenge. …”
Schillace: Google Docs collaboration began as an accident before locking was built
“And like, the fact that you could collaborate on them was kind of an accident. Like, they're just like these things on the server. We hadn't built the thing. That would lock somebody out yet. So there was just like, here's a document, like you can edit these t…”
Schillace: Writely might not exist if founders knew collaboration's difficulty first
“Because if we understood how hard the collaboration piece would have been first, Without understanding that value might have been like, eh, it's not worth it. Like, it's gonna be so hard to solve that problem. It's probably not a, you know, useful absence.”
Schillace: Google Docs would have died without legal protections and Schmidt's support
“If I had just been, like, a random Google employee with this idea, and no legal protection, and the CEO wasn't, you know, a fan of the project, it would have died. Like, there's no way it would have made it through that, that negativity, and that pessimism”
Schillace: In five years, software you cannot converse with will feel broken
“I think we're gonna feel the same way about intention and interactivity in our products very soon. Like, if I can't tell something what my intent is and have it configure itself in an intelligent way, have it converse with me, whether that's a device or a piec…”
Schillace: Most Good Product Ideas Originate from Engineering
“I think a lot of the good product ideas, most of the good product ideas actually come up from engineering.”
Schillace: Focus careers on effortless strengths rather than unpleasant grinding
“We tend to undervalue the things we're good at. We tend to think work has to be unpleasant. And so if something is easy and fun, we don't tend to think it's You know, very valuable. So I think lots of people like gravitate in this direction of like, let's go d…”
Schillace: AI will make pixels free like the internet freed distribution
“I think the way that the internet made distribution of information free, I think AI is going to make pixels free.”
Schillace: Believing Gemini's parity with GPT-4 means AI stalled is dumb
“When Gemini came out, all the press take was like, oh, it's not that different from GPT-IV. I guess we're done with AI now. We can go back to bed. I'm like, that is the dumbest possible interpretation of that story.”
Schillace: AI found more use cases in one week than crypto did
“Just to kick cryptos corpse one last time. Like there was some tweet, like at the beginning of the year that I saw that somebody was like, yeah, it took like the AI bros, like a week to come up with as many use cases as you know, crypto came up within that dec…”
Schillace: Generative AI is in the rapid elaboration phase filling white space
“The zero to one is hard. Like understanding that there's something there at all is the really hard part because you have to be lucky and you have to be talented and you have to look in the right place and like, Do some very hard work, but once you understand t…”
Schillace: Testing candidates with seemingly impossible problems reveals builder resilience
“The reason I ask is not to get to the answer, it's because I just want to see how people react when I give them something that seems impossible and unreasonable. And some people just like back off and, you know, refuse to engage with it. And some people are ju…”
Schillace: Google treated candidate controversy and single interview vetoes as positive signals
“I interviewed Dave Bezris. I interviewed him when he came in and I asked him that question and his response was, I don't do math next question. And so I failed him. I was the veto and they have this policy where we're friends now, like, They have this policy w…”