seeing a branch of software engineers not caring or learning about model development lifecycle and this is quite concerning

grok @bot hooked up to google slides is a superpower

atticus is pretty cool! and ofc powered by @grok voice!

Does it run a watchOS fork?

hardware guy going to deliver so hard for the fanboys

it makes me sad how sloppy iOS messages app interactions are.

To make a model useful, you need to give it the right environment and the set of tools and let it cook Same applies to people

more things i've stopped doing by hand in the last couple weeks: rolling out feature flags, watching rollout metrics, daily health & reliability checks. good @bot

best feeling in the world: telling someone to vibe code the feature they asked you to add because you set up the system slop resilient

post genomic

I see 2 schools of thought when using agents these days. - Agents are useful tools, you as the engineer is responsible for all it does at the end of the day, - Agents are the future, your job as the engineer is to manage and enable as many of them to unleash their true power. 1 is safer than 2, 2 is faster than 1, 2 is definitely more fun than 1. Either way, need to know your audience and pick the right approach accordingly

at which point a task becomes worthy of automation

Have been using Bazel recently in a large monorepo and reminded of the days at G where it was a pain in the butt to edit BUILD files by hand Love it when the agent takes over the annoying parts of my job

Kind of annoying that @Xcode window and the headless mcp are using separate caches when building (long waits for an already built product), and if you try to build one while other is running it fails poorly. I guess its directionally good

Books are great ways to learn about the world but there's also lots of knowledge on the streets, wonder how we'll teach that to the computer

Old La Honda (bridge to mailboxes)

who’s making husbandbench for bros to hillclimb

me and who