Will Larson's Blog, page 6

Modeling driving onboarding.

The How should you adopt LLMs? strategy explores how Theoretical Ride Sharingmight adopt LLMs. It builds on several models, the first is about LLMs impact on Developer Experience.The second model, documented here, looks at whether LLMs might improve a core product and business problem: maximizingactive drivers on their ridesharing platform.

In this chapter, we’ll cover:

Where the model of ridesharing drivers identifies opportunities for LLMsHow the model was sketched and developed using let...

Modeling impact of LLMs on Developer Experience.

In How should you adopt Large Language Models? (LLMs), we considered howLLMs might impact a company’s developer experience. To support that exploration, I’ve developed a system model ofthe developing software at the company.

In this chapter, we’ll work through:

Summary results from this modelHow the model was developed, both sketching and building the model in a spreadsheet.(As discussed in the overview of systems modeling,I generally would recommend against using spreadsheets to develop m...

Testing strategy: avoid the waterfall strategy trap with iterative refinement.

If I could only popularize one idea about technical strategy, it would be thatprematurely applying pressure to a strategy’s rollout prevents evaluating whether the strategy is effective.Pressure changes behavior in profound ways, and many of those changes are intended to make you believe yourstrategy is working while minimizing change to the status quo (if you’re an executive)or get your strategy repealed (if you’re not an executive). Neither is particular helpful.

While some strategies are ...

Should we decompose our monolith?

From their first introduction in 2005, the debate between adoptinga microservices architecture, a monolithic service architecture, or a hybrid between the two, has become one of theleast-reversible decisions that most engineering organizations make.Even migrating to a different database technology is generally a less expensive change than moving from monolithto microservices or from microservices to monolith.

The industry has in many ways gone full circle on that debate, from most hyperscale...

Executive translation.

One of my most unexpectedly controversial postsis Extract the Kernel, which arguesthat executives are generally directionally correct but specifically wrong,and it’s your job to understand the overarching direction without gettingdistracted by the narrow errors in their idea.

Some executives are skeptical of this idea because they don’t like the implicationthat they’re usually wrong, but they weren’t the audience that was offended.But the folks who got particularly upset were non-executive...

Video of Developing Eng Leadership Styles.

The last chapter I wrote for Eng Executive’s Primer was this one about developing engineering leadership styles.It’s an interesting chapter to me peronally, precisely because it’s not something I would have agreed with or written five years ago.

This past Friday I gave a conference talk on this topic at LeadingEng New York, 2024.If you’re interested, you can watch a recording of an earlier practice session from a few days before the talk,and can review the slides.I think that the practice se...

Numbers go up.

There’s a genre of computer games called incremental games, whose entire design philosophy can be summarized as, “numbers go up.” These games focus on the fundamental gaming loop rather than plot, characterization or anything beyond the foundational satisfaction of numbers increasing. The initial idea here was social commentary exposing the addictive core of many games, but like all good commentary it’s also inadvertently spawned the genre of gacha games that focus on extracting revenue from add...

When to write strategy, and how much?

Even if you believe that strategy is generally useful,it is difficult to decide that today’s the day to start writing engineering strategy.When you do start writing strategy, it’s easy write so much strategy thatyour organization is overwhelmed and ignores your strategy rather thaninvesting time into understanding it.

Fortunately, these are universal problems, and there are a handful ofuseful mental models to avoid both extremes.This chapter covers:

when to write strategy, in particular t...

Hey folks, I’m Will Larson!If you’re new to my writing, t...

Hey folks, I’m Will Larson!If you’re new to my writing, take a look at some of the popular stuff I’ve written,or my books: Staff Engineer, An Elegant Puzzle,andThe Engineering Executive’s Primer.You can hear from me a bit more frequently on my my weekly newsletter or via the RSS feed.You can also find me as @lethainon Twitter or Mastodon.

Developing domain expertise: get your hands dirty.

Recently, I’ve been thinking about developing domain expertise, and wanted to collect my thoughts here. Although I covered some parts of this in Your first 90 days as CTO (understanding product analytics, shadowing customer support, talking to customers, and talking with your internal experts), I missed the most important dimension of effective learning: getting your hands dirty.

At Carta, I’m increasingly spending time focused on our fund financials business, which requires a deep understanding...