Nov 16, 2023 · 1h 18m · in-depth
A masterclass in engineering leadership from Carta, Stripe, and Uber | Will Larson (CTO at Carta)
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 in-depth masterclass, Carta CTO Will Larson shares tactical frameworks for leading modern engineering organizations, covering technical strategy documentation, executive upward communication, productivity measurement, and scaled leadership development. Drawing from his executive experience at Carta, Stripe, Uber, and Calm, Larson provides actionable guidance on navigating organizational context, avoiding management anti-patterns, and fostering data-driven alignment between engineering and business leadership.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Brett holds 13% of the talking time here. How this is scored →
Brett as informed peer 3.8 Guest teaching 4.4 Guest disagreement 0.9 Brett pushing back 0.2
speaking balance: gold is Brett, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 59:10 Rejecting engineer complaints about executive communication
Larson rejects the popular sentiment that executives should simply communicate better, stating bluntly that sitting around demanding better communication solves nothing and pushes unearned blame upward.
Hardest push from Brett ▶ 24:53 Challenging simplistic focus explanations with hidden variablesBerson pushes back on common executive diagnoses like 'lack of focus,' arguing instead that hidden variables such as poor execution or individual contributor instinct drive the vast majority of performance variance.
Biggest teaching moment ▶ 31:00 Educating executive mental models with imperfect metricsLarson educates on the purpose of upward metrics, demonstrating that engineering leaders must use imperfect measures like story points to educate executive intuition rather than holding out for ideological purity.
Brett holds their own ▶ 24:53 Detailed teardown of attribution errors in engineering productivityBerson demonstrates deep operating insight by breaking down attribution mistakes, illustrating how organizations misdiagnose execution issues as focus problems or overlook star performers masking dysfunction.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Brett as informed peer | Guest teaching | Guest disagreement | Brett pushing back | Why |
|---|---|---|---|---|---|---|
| Navigating Company Context vs. Global Frameworks | 4 | 4 | 1 | 1 | The host opens with an incisive inquiry into distinguishing universal engineering frameworks from company-specific context. The guest explains how incoming leaders fail by importing context blindly and illustrates testing ideas for conflict at Calm and Carta. | |
| Lessons from Uber to Stripe: Uncovering Undocumented Context | 3 | 4 | 1 | 0 | The host asks for concrete examples of updating mental models when transitioning across companies. The guest details moving from Uber's polyglot ecosystem to Stripe's Ruby monolith and realizing a pushback engineer had critical undocumented context. | |
| Resolving Values Mismatches and Engineering Motivations | 3 | 3 | 1 | 0 | The host asks about the inverse scenario where the leader had to overrule holdouts. The guest describes halting a services migration at Calm due to an unresolvable values mismatch between personal technical interest and business utility. | |
| The Power of Documenting Engineering Strategy | 3 | 4 | 1 | 0 | The host asks about the mechanics and importance of documenting principles. The guest argues that strategy exists everywhere implicitly, but writing it down is the necessary prerequisite to evolving and challenging it. | |
| Structuring Strategy with Richard Rumelt’s Framework | 4 | 5 | 0 | 1 | The host prompts the guest to move beyond vague strategy advice to define the precise structure of an effective document. The guest references Richard Rumelt's Good Strategy/Bad Strategy framework of diagnosis, guiding policies, and coherent action. | |
| Core Pillars of Engineering Strategy: Allocation and Decision-Making | 3 | 4 | 0 | 0 | The host probes for concrete engineering strategy examples compared to product or business strategies. The guest outlines the dual pillars of technical strategy: headcount/budget resource allocation and standardized decision-making mechanisms. | |
| Strategy Enforcement Models and the Role of Carta's Navigators | 3 | 5 | 1 | 0 | The host asks about strategy enforcement mechanisms across an organization. The guest compares bottoms-up chaos, CTO/architect top-down bottlenecks, and committee paralysis against Carta's localized 'Navigator' model. | |
| Deconstructing Engineering Productivity and the Hidden Variable Problem | 8 | 5 | 1 | 2 | The host demonstrates deep domain expertise by dissecting the 'hidden variable problem' and attribution errors between focus, raw execution, and load-bearing individual intuition. The guest strongly validates this framing and warns against context-free executive aloofness. | |
| Managing Upwards with Velocity Metrics vs. Internal Optimization | 4 | 6 | 1 | 0 | The host asks what actionable advice the guest gives to CTOs under executive pressure to increase engineering velocity. The guest reframes measuring metrics upwards as an educational tool to evolve executive mental models rather than an objective speedometer. | |
| Practical Metrics for Engineering Organizations | 3 | 5 | 1 | 0 | The host asks which concrete metrics Larson relies on day-to-day. The guest discusses uptime SLAs, severity targets, build velocities, and total ship counts, noting that directional metrics outweigh ideological purity. | |
| Leadership Anti-Patterns: The Trap of Bureaucracy vs. Deep Technical Diving | 4 | 5 | 2 | 0 | The host steers the conversation into leadership anti-patterns around micromanagement. The guest challenges common management tropes, warning that staying entirely out of the technical details reduces engineering leaders to disengaged bureaucrats. | |
| Organizational Ratios and Adaptability Anti-Patterns | 3 | 4 | 1 | 0 | The host asks for additional recurring management anti-patterns. The guest breaks down rigid adherence to organizational sizing ratios (1:6-8) and executives who refuse to adapt their playbooks to local corporate reality. | |
| Exercising Individual Judgment and Handling Exceptions | 5 | 4 | 1 | 0 | The host notes the tension between scaling rules across an org and preserving individual curiosity for edge cases. The guest points out that organizations ultimately run on the concentrated judgment of a few specific individuals. | |
| The Triad of Hats in Top Engineering Executives | 4 | 5 | 0 | 0 | The host asks what separates top 1% engineering executives from median performers. The guest breaks down the necessary triad of hats: business executive peer, engineering management administrator, and technical excellence advocate. | |
| How CEOs Can Evaluate and Benchmark CTO Performance | 4 | 5 | 1 | 0 | The host asks how a startup CEO can assess CTO performance. The guest offers a structured diagnostic for debugging perceived shipping velocity issues by testing staffing, tech debt, or cross-functional finger-pointing. | |
| Will Larson's Personal Evolution: Removing the Umbrella | 3 | 4 | 1 | 0 | The host asks how the guest's leadership style evolved over the past decade. The guest explains abandoning the 'umbrella' management mindset in favor of exposing teams to raw, unbuffered business reality. | |
| Mastering Communication and Extracting the Kernel | 4 | 5 | 3 | 0 | The host asks about subtle dynamics in executive communication. The guest forcefully dismisses the common complaint from engineers that executives should communicate better, advocating instead for 'extracting the kernel' from raw feedback. | |
| Finding the Right Presentation Altitude and Calibrating Framing | 4 | 4 | 0 | 0 | The host asks about finding the proper altitude for executive presentations. The guest outlines a 30-second zoom-in structure and suggests pre-testing framing with chiefs of staff or executive assistants. | |
| Avoiding Meeting Derailments and Pacing Executive Reviews | 4 | 4 | 1 | 0 | The host asks for root causes behind executive meeting derailments. The guest explains that executive meetings go wrong when presenters fight derailment or try to control pacing instead of letting executives skim materials quickly. | |
| Live Debates vs. Offline Follow-ups | 4 | 3 | 0 | 0 | The host asks whether executive feedback should be debated live or handled offline. The guest separates shared-context disagreements that benefit from live debate from new-context issues that require offline analysis. | |
| Cultivating Fundamentals and the Apprenticeship Model | 3 | 4 | 0 | 0 | The host asks about building engineering fundamentals and formal training programs. The guest highlights the industry's apprenticeship model and shares high-impact curriculum topics like hiring manager conviction. |
Statements from this episode (36)
New engineering leaders fail by assuming prior company context transfers directly
“The number one way that I see new engineering leaders struggle when they come to a company is they just assume the context from their previous company applies as is.
And if you've got new leaders who come in and really make a mess of things,
It's always becaus…”
Larson canceled Calm's monolith-to-services migration upon joining as CTO
“You know, when I was at Calm, Calm when I joined about 25 engineers, about a hundred when I left,
And when I joined, we were going through a migration from our monolith to a number of services, and I canceled that migration.”
Companies document minor decisions but rarely record rationale for major strategic choices
“And one challenge at companies is like, you usually don't write down why they make the most important decisions.
Like you would
Sort of assume it's the opposite.
But if you go into a lot of companies, the small decisions are documented, but the really importan…”
Stripe's core undocumented technical strategy was maintaining a Ruby monolith
“Undocumented strategies at Stripe was we use a Ruby monolith. And so everything was done in the Ruby programming language with a few minor exceptions.”
Larson: Stubborn holdouts to new ideas often hold the most valuable context
“As an authority figure coming in, it's easy to kind of look at those folks as holdouts, people who are resistance, resistant to change, people who are, you know, just stubborn, but it's usually those folks who have like the most valuable context.”
Larson: Strategy exists everywhere in companies but is rarely written down
“There's this like pervasive belief that there's no strategy anywhere, but I've really come to believe that strategies everywhere. It's just like rarely written.”
Larson: Writing strategy down improves it almost overnight regardless of quality
“If you just write things down, no matter how bad they are, your strategy will improve almost overnight. But if you don't write it down, it's incredibly challenging to improve your strategy.”
Larson: Engineering strategy reduces to resource allocation and decision frameworks
“Engineering strategy comes down to like largely like two kind of core ideas. There's like a resource allocation, like how are you investing the headcount and the, you know, that the budget that you have, and then how do you make how do you make decisions as th…”
Tech companies dodged engineering resource allocation for a decade through overhiring
“One challenge of the last decade, right, is that resource allocation was largely dodged by many companies because they just Decided to spend more and hire more folks. And so oftentimes the real constraint was, like, hiring velocity on a given team and your abi…”
CTOs and chief architects lack context for centralized technical decisions
“Chief architects and CTOs are pretty inefficient in terms of Actually getting enough context in their heads to make a good decision on any given, given point.”
Larson: Architecture review committees rarely make the best technical decisions
“And then committees, like in almost every architecture group turns into a committee where they have to balance their different perspectives, or they're just like slow and kind of inelegant. It's rarely true that a committee gets the best decision.”
Carta appoints business unit navigators to act as local decentralized CTOs
“Something I've been trying at Carta and I increasingly believe in is something that at Carta we're calling navigators, but it's basically Having one person per area and roughly a business unit who is kind of the CTO from a technology perspective for that area …”
Larson: Salespeople are uniquely talented individuals rather than fungible headcount
“Salespeople are not fungible, replaceable individuals. They're like uniquely talented people.”
Believing engineering managers shouldn't get into technical details is an anti-pattern
“I think one of the lessons in management over the last decade has been that You know, managers are people leaders in that it is micromanagement that it is bad to get into the details. I think this is like an anti pattern because it creates disengaged and like …”
Larson: Unburdening top performers to boost output fails to scale
“So in a lot of these cases, the way to drive up productivity is to reduce the burden around like a small number of like really high performers. But if those performers ever leave, your productivity is going to drop massively as well. And so it's sort of a high…”
Larson: Half of metrics' value is educating leadership mental models
“I think people want metrics to show reality. They want to show the truth, but half of metrics is showing the truth. The other half is educating people to inform their mental model about how the truth works.”
Larson: Engineering executives must educate peers instead of shielding teams
“The most valuable thing as an executive you can do is educate like the other execs on how engineering works, not kind of protecting engineering from visibility from anyone else.”
Stripe offered enterprise SLAs with financial penalties below four nines uptime
“So at Stripe, we had a reliability like SLA that we kind of extended to enterprise customers in terms of there was actually like a financial consequence to us if we dipped Below a certain level of uptime. Like I think it was less than four nines of API availab…”
Larson: Ideological purity against imperfect engineering metrics slows organizational learning
“You can be ideologically like pure and say we shouldn't measure something, but all it does is slow down the learning of the organization around you. Doesn't actually help you accomplish your goal.”
Larson: Leaders should dive into details when teams violently disagree
“In cases where there is no clear path forward or where the folks who have the context around the path forward are like violently in disagreement, it pays to get into the details and make the decision yourself.”
Larson: Engineering teams should generally have one manager per six-to-eight engineers
“Something I believe pretty strongly is that teams should generally be one manager to like six to eight engineers. I think this is a really useful sizing mechanism.”
Adapting to company context predicts if an executive lasts over a year
“And this is like one of the, to me, most predictive traits for whether an executive will last more than a year at a company. It's where they come in and they start with, Hey, can I, how can I change myself to fit versus how can I force my entire organization t…”
Will Larson: Large organizations inevitably depend on a few individuals' judgment
“Organizations are really, even, like, very kind of large organizations are the culmination of, like, a few individuals judgment, and there's just no way to escape that, even if you have, like, a huge budget trying to.”
Larson: CTO struggles usually stem from peer friction or poor shipping velocity
“Typically when a CEO of a 500 person company feels their CTO is, is struggling, they're probably running into like one of two different things. Either the individual CTO is like causing friction with other people on the team. For some reason, like their peers,…”
Larson: CEOs need a conviction view, not consensus, on executive performance
“Anything else is going to get like a kind of consensus view, but like you don't need a consensus view on executive performance. You as a CEO need to have like a conviction view on like whether it's actually good or not.”
Larson stopped shielding engineering teams and now exposes them to messy reality
“I used to be a big believer in the umbrella concept where it's okay, we should protect our team from the reality and problems around them so that they can focus. But as you work with more senior teams but also even as you just work in generally, I've come to b…”
Start executive presentations with conclusions and stop talking if they agree
“But when you present to execs, you should start with the conclusion. And then, if they already agree with you, stop presenting. Like, there's no reason to keep presenting.”
Larson: Ask questions to extract insight from bad executive suggestions
“So like almost always when you get these like really like offhand random ideas from executives, there's a kernel of insight there that you can extract by kind of asking more, more detailed questions to them.”
Executive presentations must zoom from high-level goals to specifics in 30 seconds
“I think that the best folks presenting to executives start super zoomed out and then zoom in to their specific problem over the course of like 30 seconds.”
Larson: Calibrate executive presentations by pre-testing with chiefs of staff or EAs
“Testing a little bit with Folks who have context that aren't the exec themselves is the easiest way to help calibrate.”
Larson: High-quality executive meetings require letting executives set the pace
“If you want, like, a high-quality exec meeting, you have to let the execs own the pace.”
Larson: Important disagreements with shared context must be debated live, not asynchronously
“Talking things through live when people do have the shared like context, they just disagree is incredibly valuable and it's almost impossible to kind of like get to the end of these things on like a chat on Slack or something. You really do have to like talk i…”
The tech industry gives opportunities to learn rather than teaching proper execution
“Like we kind of joked earlier, like this isn't an industry that teaches you how to do things. It's an industry that gives you opportunities to learn how to do things.”
Engineers cannot achieve true senior-level mastery in four years at one company
“They can't develop everything you need to be an effective senior engineer in four years working in one company. You need to like see different views on these different problems to really kind of develop mastery.”
Hiring managers should override interview panels when they have strong independent conviction
“Helping managers understand their job is to build conviction, not to kind of take the average of everyone else's conviction. And that they should override their team if they have conviction and they don't think that the concerns that are raised are meaningful.”