E. Lewis

Got called meta in a big source of truth discussion today. I think it was a compliment?

"Truth is what happens to the User."

When I tell my manager I'm competitive, I wonder if he realizes I mean I'm "checking git analytics for my entire team" levels of competitive. Anyway, thinking about what it'd take to get into the top 5 for the whole department...

The giant document I wrote in 2023 has saved another engineer whose team didn't preserve institutional knowledge. The fact that the document itself is lost as part of the institutional knowledge every time though... well, culture.

a desk with keyboard, mug, 3d printed coaster, and a hint of a monitor with three heavy tech books as it's stand.

Making Recurse my home away from home again as my NYC apartment decides it's a Screaming Radiator kind of day. A silent focused office is hard to beat.

Rust syntax and error checkers gently tapping me on the nose with a new bit of syntax every time I try to skip ahead in the tutorial.

I usually say something like "Conflict is a sign that people care about things. If nobody cared, there would be no conflict. Caring is good, so it is better to have conflict, then to have apathy." Unsure if this is the exact same pitch or you mean something else.

Anyone know if there are any studies on how consistent code standards across a codebase affect use of LLM auto completion? looking @ Google folks in particular

* Better test accounts (Read 'working' for 'better')
* Buggy data relationship removed
* Constants updated
* Goat Sacrificed on the Altar of the Sales Driven Roadmap
* Tickets for week old bugs
* Updated on-call hand off document

What I really want today is a shorter TODO list and unfortunately, that means doing things. Specifically, it means doing all the little tasks taking up an inefficient amount of mental space. It should be *possible* to have a metric for focus, but would it be valued if we measured it?

I have (and love) the top left bedspread, got mine at three80atlantic in Brooklyn, where they stock 11.11 and other brands with an interest in indigo & texture like INDI+ASH.

Did you look at hackathons and creative coding sessions with time limits as possible material? Did a Creative Coding at Recurse recently, where my goal was to write a bunch of test cases understanding Things Programmers Get Wrong About Time. It was educational in terms of speed VS understanding.

...and when my team meets my lower expectations instead, well, that's why we continue to have these contrasting views.

Someone who only talked about how we should push for the ideal situation would lose my confidence quickly. Someone meeting with the other stakeholders the same day to tell them they need to go further than I asked? That's a confidence boost and a lesson for me and a benefit to the whole team.

As a cynic, I'm grateful to report to an ideal-chaser with (this is the important part) a lot of skill in coming up with practical action items.

When I'm in the moment, trying for some small improvement, I can count on them to push for the larger improvement with all the force of the skip-level.

Some images from the Twitter years in a thread (since they are no longer accessible) 1/n

Thinking about writing a blog post about my estimation checklist, but first I would need to make a proper blog. Hmmm, domain names...

The other person doing formal methods at Recurse found it without me pointing it out, and used it as far as I know.

Just got my first liveness constraint working in a custom TLA+ spec 🎉🎉

A diagram matching Hillel's translation of the code to the actual code, via rainbows and too many arrows.

Very beautifully translated. Have a diagram in honor of my failed attempts. Tomorrow, I'm finishing the Core section of the book and then I'm going to try to Do Something.

Sorry, it's the future lock, not the past lock. And also something something conveying the brackets. Not simple!

TLA+ redefining my mental definition of 'fluency' in programming languages over here as I take three+ layers of mental translation to get anywhere.

Glaring at `[][lock # NULL=> lock' = NULL]_lock` like, "Always Lock Not Equal To Null At Least As True As Past Lock Equal To Null Or Unchanged Lock" ...ok, try again... "A non-null lock... is always at least as true as... the past value of lock either being NULL or unchanged?" Ok, again...

First day of pure flow state in ... months, probably. Thank you, Recurse Center.

Been setting up Linux Mint on an old laptop and enjoying the symmetry of having one Mac, one Windows PC, and the minty fresh new (old) kid all in one place.

If you're digging the hole, you get to choose your own shovel. (And now, the fighting over who gets to dig the hole.)

I do actually like this though, especially from a small team perspective.

Today I learned print has an argument limit length. Which makes sense, but also, was not something I ever expected to have to think about. Accidentally stress-testing other people's CI scripts, here we go.

Finally got around to making a non-work bsky account. If this is relevant to you, you're probably already on follow over there.

The best side effect of writing the pytests presentation was taking a few minutes to figure out coverage for our legacy tests VS the pytests and realizing there's so much overlap that we should be able to cut 75% of the legacy tests, saving ~$1k a month of CI/CD budget not to mention developer time.

My pytests presentation went well for a "multiple staff engineers saying they learned a lot and want another discussion period for questions" level of well. Hoping that gets me a lot of trickle-down effect in code reviews across the company. Meanwhile, there's removing all the legacy tests...

oh, yes please.