This video may just become my personality
06.03.2026 05:45
👍 1
🔁 0
💬 0
📌 0
This video may just become my personality
like, it goes from feeling pretty good, to feeling perfect
that tilt, tho
As a kid I went from an Apple II single button mouse to a TrackMan Marble - this post awakened something within me. The [now] home one arrived a couple days ago, the work one will be here tomorrow.
I vibe-coded a stt windows service a few weeks ago based on whisper, and it's worked great. Today it sort of heard a YT video, and hmmm
"(distant chatter) - It's desperate. (distant chatter) - I need my eyes. (distant chatter) - Do that guys. (distant chatter) - Oh, it's strong, it's strong." 😄
Post a pic you took, no context, to bring some zen to the feed.
You absolute legends
“Shithead runs his mouth” is not actually a genre of story you need to keep returning to.
We’ve found @tolkoto.com’s arch-nemesis.
Uh, hey. I have a launch-era (April 2022) 256gb SD with a 1tb sd card in it, and a dock. I got an Ally X a while ago, so it's been sitting in a drawer for months. Shoot me a DM - I got you.
My aunt was a rep until 2020 (Sue Ford, representing Grafton’s 3rd district), and man, some of the stories she told about the folks she worked with.
Bravo. Y’all knocked it out park.
And they only get paid $100/yr, so it attracts lunatic true-believers.
This really is a great picture:
Nintendo of America still undefeated as the shittiest law firm in gaming.
I didn't @ him because I figure he gets enough attention, but ed3d dot net on here has a good head on his shoulders for AI stuff, and his tools are clearly built with enterprise development in mind (lots of Agile concepts - the planning stage all but asks for the Jira ticket number, lol)
I guess what I'm trying to argue is, there are a TON of people releasing shitty vibe-coded projects out into the ether, but that doesn't mean that's all the tools are good for. You can run a full SDLC, design review, all that stuff, AND YOU SHOULD, while also using LLMs.
Yeah, this is kind of the key to how the tool chain I've been using works. Just constant reference back to the project plan and detailed acceptance criteria. Opus sends Sonnet agents out to do work, they come back with results that are reviewed against the requirements, rinse repeat.
Now, is it deterministic? Maybe not, but that's not _really_ how I think about it. Does the code meet the AC and pass all tests? Does it pass UAT, which the skill generates a test plan for? Does it pass the security scanning architecture I've built?
Here's what I've been using. github.com/ed3dai/ed3d-.... Getting from plan to working code takes easily 3-4x longer than with ootb Claude, but I've been impressed with the output. The readme does a better job explaining how it works than I can here.
It's the suit version of the "butt scrunch" yoga pants
I think this is right (although decreasing) for straight LLMs.
After using Claude Code Pro for a month, with some very well built Skills from Ed3d, I’ve really changed my mind on the utility of these tools. I just talk to it.
But is it WILDLY easier for anyone to deploy insecure software now? 💯
Straight out of the box, Claude Code will absolutely do this. But like any software if you figure out how to use it well…
(FWIW: I don’t rely on Claude for testing - it creates the AC {with my approval} and TCs, but tests run through Jenkins)
I've spent the last 2 hrs being the most miserable scrum master imaginable to Claude Code, and it's much more fun than it should be.
(Been using @ed3d.net's tools for a bit, which integrate ideas like Definition of Done and Acceptance Criteria. Tonight I decided we're going to have proper retros)
Get every one of their asses. Shit like this is why I subscribe to the Verge.
a. Hell yeah
b. I really like that bowl
100% - my brother in law also has a degree from MIT and worked at Draper, but on satellites not subs. I guarantee you want people like him working for you. This is so self-defeating.
I hope it works though - High On Life x Sunset Overdrive (which is what this is) could be fun as hell
