Will Larson

A lot of folks want a map to Staff Engineer, and just as many Staff Engineers want to better understand their role. Towards that end, I'm working on a small project to collect folks' stories of reaching and operating as staff-plus engineers staffeng.com

Ok here we go!! StaffEng, AI edition, with @voidfiles and yours truly. Sign up for our listening tour and to nominate guests! (Link in thread)

@joshua_hornby @DanielNealAdler Ty!

@Lethain Hey Will, your recent post on sharing Claude transcripts made me think of github.com/dadlerj/tin by @DanielNealAdler Thought I’d share in case you found it useful?

@ThomasJ02 @tylercowen Yeah it’s a fun experiment! Interested to see if the pattern gets traction

This AI companion from @Lethain is the first time I've seen something like this besides @tylercowen's GOAT book. I think this will become the new norm for a lot of non-fiction.

@porestar I think generally yes, but presumably any given model with work better with a specific harness. Eg recent meme on X that Claude opus 4.5 effectiveness has degraded due to harness changes in Claude code (no idea if true, hasn’t really been my experience)

@Lethain „Ultimately, I think these are fairly thin wrappers, and that you’ll learn a lot more by implementing these yourself“ - do you think model performance will generalise to any harness?

Wrote up a series of nine short posts on interesting problems we encountered while building an internal agent framework (virtual files, compaction, LLM vs code-driven workflows, etc) lethain.com/agents-series/

@somi_ai 100%, not something I thought we needed until I realized I couldn’t live without it

@Lethain The virtual files pattern is underrated. We found that sandboxing file operations through an abstraction layer let us test agent behavior deterministically without actual filesystem side effects. Made debugging so much cleaner when agents could run against mock file states.

@davidnoelromas @voidfiles Would love to be first (or early whatever!) guest 😂 not sure I have the coordination bandwidth to host

Now that I’m a few years out from podcast.staffeng.com… I kinda miss doing a show. Maybe I’ll start something new. Probably about AI best practices for eng leaders in big tech—the future is more unevenly distributed than ever. @voidfiles wanna join? @lethain wdyt?

@bbruceece 🙏 thank you! I’m grateful to hear that

@Lethain This is my favorite of your four books, it’s incredibly concise with its goal and is a more unique topic than the other three. I feel energized and focused having gone through it and will be revisiting almost every chapter in the coming months

Now Reading: @Lethain’s Crafting Engineering Strategy

@bbruceece 🙏 hope there’s something useful in there!

Taking advantage of the quiet gap between Xmas and New Years to finally get to reading Mr Larson’s newest release About one hour of reading has generated about 4 hours of follow-ups to do once the team is back in office 🤯

Year EIGHT as a founder CTO. 2025 felt like a decade. The year of vibe coding, Apple opening payments, almost getting acquired, saying no, scaling past 100 people, mistakes, cultural memes, and figuring out how to go long. miguelcarranza.es/cto-year-8

@can Wow they’re still at it, huh?

just odd. why would anyone think sending anonymous emails would work?

one day ill write up about the person who insists on sending slanderous emails to people in my life

My annual year in review post lethain.com/2025-in-review/

Love this post from @simonw simonwillison.net/2025/Dec/18/co…

@simonw @obie Agreed, and also it means that it’s possible for senior folks (previously in meetings too often to write) to write software again. Doesn’t yet solve time to participate in other dimensions of creating software tho (production issues, scaling, etc) lethain.com/coding-at-work/

This from @obie captures how I'm feeling about coding agents

@aktwits No magic bullet, have worked to solicit folks to present to share their approaches which creates good visibility, but ultimately I’ve found getting more folks with AI experience embedded in more teams/orgs has been best

@Lethain Great insights! Great you are tracking "Adoption" within your workforce. Curious, how do you track "how well" people are using it?

In the spirit of working in the open (and not eg preaching from the pulpit), I wrote up my notes on driving internal AI adoption and would love to hear what other practitioners are trying! lethain.com/company-ai-ado…

@tylerklose @can @substack Also in terms of unsubstantiated takes: I imagine that the volume of slop generated 2023+ means that training sets will predominantly feature earlier content rather than stuff published now.

@tylerklose @can @substack I'm not very worried about this, as long as your ideas are somewhat novel (in this case, meaning statistically infrequent), I don't see much reason to fear LLMs will surface them highly, even if they exist in the training data.

I won’t link to his YouTube but very disappointed that @substack allows people like this who reblog my entire content and then charge for it natesnewsletter.substack.com/p/grab-the-18-…