Activity for engineer.sitereliability.best

Ok should I learn kuberay?

Zach is so great for this site and i hope mobs never run him off.

Zach Weinersmith@zachweinersmith.bsky.social

The evidence that they're lying can't consist of their position itself, because now you've logically ruled out that anyone can ever sincerely disagree with you. Moreover, shouldn't you also assume that people who agree with you are also likely to be paid agents?

One might say the human ability to reframe, adapt, and gracefully extend the systems we are a part of is the primary source of robustness in the world.

Mity Cheese@mitycheese.bsky.social

This points up a key problem with how a lot of companies attack robotics and computing. They're can often succeed at a task, but unlike biological entities, their failure modes are not very robust.

I worked for record labels during this period. Figured I would do it for the rest of my life. But by 2010 things were feeling hollow and precarious. We stopped signing the whole band and only signed the frontman. 360 deals took growing chunks of all revenue streams. Terrestrial radio had collapsed.

Zach Weinersmith@zachweinersmith.bsky.social

This also explains the experience some of us had where roughly from 2005-2015 ish, plus or minus a few years on each side, it was easier to make a living. The cost of distribution had cratered, but we were still acting like artisans, rather than making commodities for the aggregators.

i made a karaoke app that has a secret "weird al mode" where you type in a one-word prompt and it sends the word-aligned/diarized lyrics to an llm to rewrite w/ a themed parody version. it rules.

ellie lockhart (she/they)@eleanor.lockhart.contact

Thing that eventually will happen: Singing real karaoke to AI generated songs

If I had a boss right now I would be sharing this with them. I like the exec-friendly framing for why sre-in-a-box products will always miss the point and the practical steps you can take to build your own w/o hiring an ai team.

The practice of using coding agents to write comprehensive test suites to build confidence in code changes and the geometric speedups we're seeing in coding model inference are on a crash course. If I ran my harness in fast mode it would spend 75% of the time waiting on tests to finish.

I'm realizing how different Star Wars must hit if you think the droids are all mindless drones. A bunch of themes and narrative arcs just turn to vapor.

Alex Wise@engineer.sitereliability.best

Gotta say I prefer Zach's approach over Hank's. IMO if he hadn't tried to use C3PO as a relatable proxy for "autonomous machine with human-equivalent moral concern and qualia" there would've been less hubbub. It seems a lot of folks don't consider threeps a moral patient.

Anybody who can say with conviction that students are making a mistake when they discount the value of traditional education signaling in favor of more social means of access and status has too much conviction. We can't promise "learn xyz and you'll have a good life" and they know it.

Colin M. Henry@henryhenryhenry.com

recommend reading the paper in entire: papers.ssrn.com/sol3/papers....

Not to take a joke too seriously but a big factor is the problem space for IaC is different from regular software delivery, and we want solutions that feel like traditional software but address a different set of concerns and tradeoffs.

amos@fasterthanli.me

infrastructure as code? *looks at code* n..no thanks

evergreen

i am saying this as a professional@turbocri.me

infrastructure-ass code smh

The "local AI" scene seems to be quickly becoming a mire of FUD-peddlers selling terrible products that have worse privacy/security/capability/cost envelopes than hosted alternatives.

At my last job I shaved a little of the cloud bill by shutting down some old infra. Devs were on a $20 OAI plan at the time and many ran into quota limits. I asked to run an experiment where we gave one dev a Vertex API key w/ a months worth of the savings to see if impacts delivery.

Jake Gold@jacob.gold

AI token budget consensus I'm hearing from friends at >150 developer companies is ~$3000-4000/mo per developer. Which is equivalent to buying every developer a new Macbook Pro every month. My guess is we spend more and more per developer going forward, but hopefully on raw compute not model APIs.

Chris had a script for Odyssey and that did pretty well. They copied the idea for Brand New Day and that turned out great. Maybe try having a script for this one too!

IGN@ign.com

First look at Frozen 3 at #D23!

Me daring my kid to write in cursive, 2026.

The Verge@theverge.com

For comparison www.theverge.com/tech/979583/...

A lot of the harnesses today are jumping to concurrency for speedups because LLM responses are relatively slow. As faster inference providers come online, I wonder if we see single-threaded harnesses on fast inference outperform the swarmy systems on some tasks.

“Of all the marvellous achievements of modern science, the Electric Telegraph is transcendently the greatest and most serviceable to mankind. It is a perpetual miracle, which no familiarity can render commonplace.” Hold my beer.

apenwarr@apenwarr.ca

Ok this is very neat

o no, the janky toy harness got complex enough i prob want a real workflow engine. this way lies madness.

One of the important things an eng interview should try to figure out is "does this person understand that outcomes at scale get very weird and unintuitive?"

dax@thdxr.com

deepseek is insanely good at inference - they are hitting 96.56% cache ratio on our heavy traffic second best provider for us is doing 91.60% this doesn't seem like a lot but it means they're using ~2x less GPU time

Love digging into "why did this agent go off the rails?" and finding the answer was a subtle word choice in a single prompt 4 conversations ago. A large part of working w/ coding agents is developing a spidey sense for when you're about to monkey's paw yourself into a big mess.