[new blog post]
TLA+ as a Design Accelerator: Lessons from the Industry
muratbuffalo.blogspot.com/2026/03/tla-...
10.03.2026 21:33
👍 11
🔁 1
💬 0
📌 0
[new blog post]
TLA+ as a Design Accelerator: Lessons from the Industry
muratbuffalo.blogspot.com/2026/03/tla-...
TODAY. Come see John and Peter talk about C bounds checking and real time locks!
DC Systems tonight! Come for the tech talks, or come for the beer, or just come see 30-40 of your closest friends!
Monster Scale Summit (free and virtual) starts tomorrow. Pop in to learn from an incredible lineup (@skamille.themanagerswrath.com, @antirez.bsky.social, @martin.kleppmann.com, @muratdemirbas.bsky.social, @dominiktornow.bsky.social, @teivah.dev Pat Helland, Joran Greef....) www.scylladb.com/summit
I've spent the past couple months hand crafting a new B-tree map implementation for Go. It's fast. Faster than Rust's BTreeMap. Faster than the fastest C++ (frozenca/btree). And about 2x faster than my current Go btree (tidwall/btree).
https://scour.ing/ has gotten pretty good at surfacing what new stuff I actually want to read on the internet, better than following subreddits. You can see my feed of mostly database things at scour.ing/@linearizable. It surfaces small personal blogs particularly well.
High-level and low-level projects are allowing AI-assisted contributions. Yes Zig, NetBSD, GIMP, and qemu ban AI. But Linux, curl, liburing, MariaDB, and LLVM all have AI-assisted commits already.
Yeah possibly, the Linux kernel for example allows ai assisted code changes but not docs! Which is kind of interesting. But if it's all one repo then I doubt the policy would differ but maybe.
And oops thanks for the pointer.
In all cases, and this is something I'm learning, I am most looking for submissions that talk about existing projects, not focused on things you (the author) built. If you wrote code during the course of the article that's fine, but the focus should not be on your own work.
If you're a CS grad student in the US, I'll pay you a small fee to write for The Consensus. I'm especially looking for articles that compare the state of research to the state of what devs do in practice, because there are often interesting discrepancies.
theconsensus.dev/contribute.h...
Very happy to have the first contributed article published in The Consensus. And very happy to turn Alperen loose on you all to talk about type systems and gradual typing.
It was shown but I agree it could be clearer. Perhaps display it as a table would be better.
Thank you for the feedback and kind words!
It has not been easy to decide what goes into each tier. But for the moment the Student tier gets you access to paywalled articles, Standard tier gets access to a full content RSS feed and the monthly newsletter.
If you have signed up for the Standard tier after the newsletter was sent, you can find the newsletter archive here to read it online.
theconsensus.dev/n/2026/02/st...
The Consensus launched in February, payments were only in place two weeks ago; I was prepared to send out this edition to only myself as practice. But in the last two weeks nearly 80 individuals subscribed at the Standard tier. The support and interest has been astounding. 🙇
The very first of The Consensus Standard just went out. This is a monthly newsletter for folks subscribing at the Standard Subscription tier.
A collection of 1) articles written for The Consensus, 2) interesting jobs, 3) funding announcements, and 4) external articles I enjoyed.
New batch of jobs dropped. These job postings are not sponsored (if they ever are, they will be labeled). They're simply interesting-looking opportunities related to software infrastructure, pulled from around the internet.
theconsensus.dev/jobs.html
AI writing is like store-bought cake. It might be perfectly fine, maybe even as good as something you could make yourself, but it’s weird to give it to someone and say it’s homemade
Just spent an hour today in our agentic AI coding class showing all the way Claude Code produces *subtly* bad programs for a pretty trivial application. Wrong in some ways, but just bad in other ways. Reading its output is always a sobering experience.
The performance team I lead at Automattic is hiring - automattic.com/work-with-us... #webperf #php #jobs #wordpress #woocommerce
Thanks for clarifying that!
While yes those projects exist, no they did not account for any of these major projects I surveyed. :)
Some of the AI-assisted commits I found made me chuckle: a three-line CSS change by Claude in Mattermost, or a change by Claude "reviewed by NOBODY (OOPS)" in an Apple engineer's commit to WebKit (granted it's a comment-only change).
That is very generous, thank you.
We - like many other technical teams - are hungry for the market for trusted, neutral, technical advice, but it's hard to find. The big analyst firms aren't sufficiently forward-thinking, and their publications are rarely detailed enough, geared towards managers rather than practitioners.
No I am not publishing it as an indictment. The 40 or so projects that didn't use AI explicitly I suspect many of them are using AI without mentioning it.
I surveyed 112 major source-available projects to understand their AI contribution policy and whether or not they have actually accepted explicitly-labeled AI contributions.
Only 4 projects banned AI completely: Zig, NetBSD, GIMP, and qemu. 70 already have AI-assisted commits.
