Jack Danger

@jackdanger.com

Engineering @ LaunchDarkly, burritos, acapella

Every warning-against-AI written by a frontier labs employee makes me wonder if they even know that someone has to actually *pay* for the computers to execute software

Serious question: How are eng leaders bringing their teams safely through this AI shit? There is so much to learn, so much to grieve, so much fun to be had, and everybody is handling it differently.

Two AI truths through which I see everything: * commoditization is inherent - everything is a race to the bottom because there is zero stickiness anywhere * what you can do with AI is limited by your own knowledge

I’m not sure I’d buy the idea of AI addiction (people destroy themselves and their families to feed an addiction), but it obviously generates intense dependency (like phones do)

I’ve been working on a longform blog post, but I realized the whole takeaway is this: computer science degree matter again. Source code is free forever, but actually running software has not become any easier.

I see folks interpreting the SpaceX IPO as a call to be more like Elon. Richness is giving to others more than you take, creating value that wouldn’t exist without you, and leaving every person better than you found them. Invest in your own internal moral compass and generate joy for yourself.

If you want: * The public to think AI is scary * To pump up mentions of your name before IPO * To stop losing $400/query I imagine you might shut down your giant uses-8-racks-to-answer-questions model when any fed gives you a chance

I thought Python would be the default language of the future because the folks who built LLMs used it the most. These days I'm doing everything in Rust because only correctness and performance matters if I'm not seeing the code anyway

Anthropic, as a company, has many strengths. But their leadership exists in a realm where everything is free. So nothing about their products or tooling is informed by customer cost-sensitivity

When inference prices go up (likely this summer IMO) software engineering will look even more like engineering: methodical and intuitive cost/value trade-offs over a complex system

I'm using Opus 4.8 for some extremely difficult software problems and it's failing in consistent ways. It's like mentoring a promising junior engineer (that moves at 10,000x human speed). Feels like it's asymptotically approaching autonomy, but every step toward it is exponentially more expensive

writing code with AI relies on two huge drivers: * model inference * developer tooling (e.g. Claude code) There’s no lasting competitive advantage in the models (every part gets cheaper every year). And the dev tooling is fun and easy to make with good models.

I voted for Steyer but I’m happy with Becerra in part because all the wonkiest people I know in CA say he’s legit, if not colorful. After seeing the outcomes of a dramatic government, I bet I’d be pretty happy with a merely effective one

there are only a couple things about AI that I’m pretty certain of: * We should store every one of our input and output tokens. These are ours. * In a few years your laptop will run all the AI you need for 99% of cases. 100% for most people

Start planning now to attend a No Kings protest on Oct. 18. We need this one to be even bigger than the last one. Invite your family and friends, and ask them to invite their family and friends. People power can defeat fascism.

Bild

I don’t think anyone is prepared for what they just did w/ ICE. This is not a simple budget increase. It is an explosion - making ICE bigger than the FBI, US Bureau of Prisons, DEA,& others combined. It is setting up to make what’s happening now look like child’s play. And people are disappearing.