Aaron B.

Here's the leadership strategy that propelled me to Principal Engineer at Google: "I push rocks down hills." This is the pithiest summary of Principal Engineer-ing that I've come up with, and it's a major part of how I approach exec-level engineering leadership. Moving up the career ladder as an engineer is about finding leverage: ways to produce more value per unit of time. You can grow your skills so you can handle harder projects and finish them faster. You can improve your judgment so you identify opportunities that produce more value for the effort. You can lead teams, so your skills and judgment get magnified across more contributors. One way to magnify your output as a leader is by turning a hard problem into an easy problem. Something I've observed over time is that it's a lot more challenging to start a project than to keep a project moving in the same direction. The first few commits, the project outline, the high-level architecture: these are the hardest and the most consequential challenges. Once the skeleton is in place, it's harder for a project to run off the rails. Starting projects is hard; maintaining that momentum is comparatively easy. This leads to the discipline of pushing rocks down hills. Your goal isn't to drive a project to completion; it's to impart enough momentum to the project that it will cross the finish line even if you stop paying attention to it. The goal is to build teams that don't need you. Don't build teams with yourself at the center: build teams with yourself on the outside. Once a problem is moving under its own momentum, it's no longer a high-leverage place to spend your time. You can move on. It's time to push a different rock down the hill. One consequence is that you will start more projects than you finish. Not because the projects *don't* finish, but because *you're* not the one driving them over the line. You need to have the humility to step back and give credit to the folks who finish the job. Scaling yourself means taking a smaller share of credit for a larger number of projects.

1,538

56 Comments

Do we have metrics on how often LLMs generate similar blocks of code across the industry? Most software being developed isn't unique, and it seems we'd be better served by new frameworks, and higher level abstractions. But then we're back to the discovery problem, something LLMs seem to help with.

106

14 Comments

Meta just announced another AI reorg - again. While everyone debates strategy and spending, I'm thinking about the employees living through this constant change. Here's the reality: In AI today, rapid org shifts are inevitable. The technology evolves so fast that yesterday's structure becomes tomorrow's bottleneck. Meta's pattern isn't unique. Google, OpenAI, Anthropic - they're all constantly restructuring. But we're not talking enough about the psychological toll on the people building these systems. I've watched brilliant engineers go through three team changes in a year. They start questioning everything: Am I working on the right thing? Will this project even exist in six months? Should I even bother forming relationships with my new teammates? The human cost is real: ➡️ Senior engineers hoard knowledge because they don't know who they'll work with next quarter. ➡️ Mid-level engineers burn out not from technical challenges, but from repeatedly rebuilding trust and restarting projects. ➡️ Junior engineers never get the stability to develop deep expertise, or they never even get hired in the first place. Each restructure makes the next one feel more inevitable, more exhausting. People operate in permanent transition mode where creativity goes to die. As engineering leaders, we can't stop the pace of change in AI. But we can be intentional about how we shepherd our people through it: 👉 Acknowledge the emotional labor of constant change. 👉 Create stability where you can - rituals, standards, communication patterns. 👉 Invest in relationships, not just hiring. 👉 Be transparent about uncertainty. The race to superintelligence will be won by teams that maintain human connection and psychological safety while everything shifts around them. Here's the article: https://lnkd.in/eqy4VnkM

148

28 Comments

One of my smartest friends left a cushy software engineering FAANG role to join a startup. 18 months later, they were burnt out, underpaid, and back on the market. They took a risk, and the risk didn’t pay off. In hindsight, the writing was on the wall: The market was crowded. The business model wasn’t solid. A year in, growth the “fast-growing” startup had stalled to a halt. The founders were scrambling. Priorities changed weekly. Looking back, my friend didn’t ignore red flags, they just didn’t know what to look for. Joining a startup isn’t a vibe check. It’s a bet. A big one. You’re not investing money, you’re investing something harder to get back: • Time • Focus • Energy • Reputation Startups can be incredible places to learn, grow, and thrive. But only if you pick them with care so treat the decision with the rigor it deserves. Ask the same questions a VC would: 1. Is this a market worth betting on? 2. Can this team actually win? 3. Is there a business behind the product? 4. What’s the actual deal you’re getting? 5. Is this the right move for you?     I break these down with none other than Oana O. , a seasoned VC who graciously shared her wisdom with us in my newsletter article. Read it here: https://lnkd.in/gxEewSru ♻️ Please repost if you found this useful

71

7 Comments

How trading platforms display stock prices update every millisecond? Imagine users refreshing the pages manually, missing critical updates causing frustration Imagine app asking the server "Is there a new price?" every single second. This constant polling would swamp your servers and create a sluggish, costly, and poor user experience. Solution: WebSockets Here’s how it works - It begins with a single HTTP request from the client. This request includes a special "Upgrade" header, asking the server to switch protocols - If the server agrees, it responds with a confirmation, and the connection is elevated. This creates a persistent, two-way communication channel over a single TCP connection - This channel stays open. Unlike the classic request-response model of HTTP, the server can now push data to the client at any time - This eliminates the need for constant polling. Data is sent only when it's available, which drastically reduces unnecessary network traffic and server load - The result is a low-latency, real-time connection, perfect for applications where speed is critical. 🤔 over to you, where else we can use WebSockets? Let me know in the comments! 👇 --- You've got System Design skills but showcasing them in 60 minutes interview is a completely different challenge for e.g. where to start? How much time to spend on HLD vs LLD etc Let me share a FREE 7-day email course that provides a step-by-step framework for acing your 60-minute System Design interview 📣 Join 1300+ subscribers and sign up for the course at systemdesign.engineer, to learn time allocation to demonstrate full range of design skills

67

17 Comments