Fucking, during the design process, Claude had the UI/UX design choice, to "make sure the phone data entry has BIG buttons for GLOVED hands in the field. I mentioned nothing about this. Only context as a dairy farmer, and it designed an effective UI/UX expirence.....
And make Google data analytics views of his feed costs, and do forecasting, ect. All shit he never was able to have time for, and I built it in 16 hours. By hand it would have taken me 2 months? Hard to say, and it wouldn't have had the dynamic data views, and intelligent data processing
Ya, in 16 hours last Saturday I made my dairy farmer brother an app powered by 10 years of unstructured Google sheets data, that not only lets him now track things in a structured way, with a nice UI and phone AND desktop layout, but I hooked up an MCP server and now Claude can analyze his data
This - AI is real folks. Stop trying to fucking say it doesn't work or won't replace you, cause it can and will. Orginize now, figure out what we are gonna do, cause if you keep putting your fingers in your ears about what's coming it will roll you over.
There is a world where people learn to completely control it themselves, who knows.
Anthropic provides more control of these knobs then most, even in their consumer facing knobs. And Claude code lets you see the number of tokens in their prompt, plus since these things run locally, you can look at the code for some of it. That said, with local models getting close behind
Ya, people need to really get past the fucking "LLMs are just stocasric parrots that don't work!". This hasn't been true for like a year, and especially it completely shifted with the recent wave of models since November or so of 2025
I’m saying you have no idea the kind of cocaine that’s going to be available by 2030
the things we can synthesize only increase in their fidelity. ChatGPT voice mode really sets my teeth on edge, and at the same time, it’s extraordinary science fiction technology. TTS didn’t sound like this EVER.
It’s so much more than “oh here is fake homework”
or “here is a fake influencer, saying propaganda”
that’s the little league of this shit
What’s going to happen when people have their own personal, on-demand artificiality? there will be versions worse than app sports betting, in impact
the “it doesn’t work” folks did us all a disservice, even themselves
I don’t think we’re ready for the social consequences of all the artificiality that’s about to wash up
Living through a boom of network abundance transformed the world in ways we’re still making sense of
this will be… rougher
Only possible upside of this whole thing is convincing more countries and people to go all in on electric. Though it does mean way more profits for the oil companies in the mean time so 🤷♂️
I didn't think when I wrote
"Additionally, because oil deposits inevitably run dry, we often rely on extraction efforts happening in other countries which may or may not be friendly.... and [its] cost can change quite dramatically based on supply and demand"
it would be so prescient but, well...
It's even better when you are able to charge at home cost wise (like for me in Colorado about 1/5 the price) but the current crisis really shows why electric is awesome! Have some solar panels and you don't have to rely on a constant supply of shit to burn
I charged my EV last night (from 20 percent to 100 percent) and it cost me about $17.
Beyond the no gas, the car just rides so beautifully. Got it in July and cannot imagine ever going back to gas car.
1) Single source of truth, data driven extensibility - basically types / enema are defined once and everything flows from there.
2) fail fast, validate input boundaries and validate and when broken do the error
3) domain logic separation from infrastructure
4) generated contracts (API specs)
This allows natural common, good practices to bubble up a bit, and keeps the code base cleaner. I just added a principles skill that define principles they should follow during design, implemention and refactoring.
Which for agents I think have the biggest universal weight are
The pattern I have adopted when I am more hands off (depends on the side project) - is I have a refactor plan, implement refactor plan then a pattern extractor skill that finds common design patterns 3+, creates/updates a patterns skill that is referenced in my design and implement workflow skills.
Absolutely, still requires over site and stearing
github.com/nklisch/skil...
Asked Claude and he was like, ehhhh, this is fast moving, maybe its worth it if you want to freeze the contract. Shrug, I'll just do my next major project in rust
First major open source project for the agentic age....
I think it's the right abstraction for skill distribution.
Handles looking for updates for skills easily sharing, a little security, ect
Decentralized contract anyone can adopt.
skilltap.dev
github.com/nklisch/skil...
Maybe I'll have it rewrite my first open source tool this agentic age rewritten in rust 🤣🤣
I have been wondering if I should start having it do rust here. I have mostly been doing strict typescript cause I know it well enough and my rust is weaker, but I should give it a shot in my next project. Done some go at work cause everything is go there soooo
My current take is something a friend said:
AI multiples everyone by at least 10x, but the base scale supports negative numbers
I mean it's why China is pushing so hard for electrification.
I do morn for the craft, I morn for the time I was thinking at the lower interface level at all times, but I do not see how us seniors "hold the line" without holding people back....
It has let me build documentation that is better and more comprehensive then has ever existed for the services I maintain, and inherited. Those docs have helped reduce friction for other engineers when integrating into these systems.
Multi-repo systems, point to the exact traces by file name and line of code to allow me to quickly find my dependencies, and root out bugs. It has let me prototype refactor and improvements put on the back burner for years in like a few days.
What I could, when given the chance on new projects try to follow good patterns and practices, build for sustainability and scalability when it makes sense, but those opportunities are few and far between for most. The AI tooling has let me slog through legacy code base across million line
Been in understanding legacy systems, half rewrites, attrition, random new library adoptions, some random over abstraction so guy did cause he thought it was cool.
The vast majority of software in my career I have touched has been half baked, duct tapped together already.
I have tried to clean up
