Getting into a grove using my AI dev workflow. Really promising results. I've got a couple of sessions running taking my workflow and turning it into an actual harness. Just getting into day two and there are already around 350 closed tickets. Pretty sure I'll have MVP today.

I'm still shocked at how well a good dev workflow with strong guardrails really works with these latest models. Other than 1 bug the POC was done in 3 hours and it's way better than I expected. I could deliver this POC now and and it would blow people away. But I have two weeks!

I didn't realize copilot updated the cli. Much better.

I've been working on my own orchestration / meta harness (Volon) for structured development work and so far it's working better than anything like this 6 months ago. But the cli harness you are using makes a big difference. Copilot is way behind.

My wife installed my notes app (Nanite) and it only has 2 bugs so far. Bugs now equal the number of users.

Writing code by hand is dead. After this weekend's experiments I'm even more excited about what's possible.

My first loop experiment resulted in a 1-shot success with over 100 tasks generated, lots of good logs, and a working POC from raw brain dump to working MVP. I then used the MVP to do it again for 3 different projects and it went just as well. I'm really happy with this so far.

This app is on loop 21 of scaffolding itself into existence.

I'm about to run a test to have an AI agent build out my new workfow via inception. It will start from one prompt and iterate to scaffold itself up to the point where it can use itself to finish building itself fully and then I can use it to build other things.

I have a tech startup idea (like everyone else) that I actually think has potential. When I think about it, I get that sick feeling in my stomach that usually tells me I'm on to something. Going to at least explore feasibility / market.

I have agents using a skill to read, write, and search custom notes app (nanite). Been wanting to do this forever.

My dad's health is failing and not the kind you recover from. This is very hard to process. I already hate that we don't get a real say in when things end but when it's my dad? I'm a bit of a wreck.

Started learning python for my new AI role and I like what I see so far. I'm sure it'll have something annoying like all languages but a lot of it feels intuitive so far. Really digging into tensors.

I'm excited to start my new role at work in January. I'll be working with AI full time. I'm winding down on my last bit of work on my current team and it's hard to concentrate because my mind keeps thinking of things we can do with AI.

Excited about moving to an new role at work where I will get to focus full time on AI related work. Going to be a lot of pressure but I couldn't be more excited about this challenge. The transition will take a few weeks and I'll share more detail when I can. So stoked!

Have an opportunity at work that I can't wait to hear news on. I can't share details yet but I'm hoping to know in the next week or so. It's moving fast at least.

Time for a marathon meeting day. Pray for me, send good vibes, burn some sage, whatever it is you do. I'll need it today.

American healthcare is a joke but being in Alabama it's even worse. My doctor prescribed Vizz eye drops but insurance barely pays any towards it. Apparently sight is a convenience. It was $178 for a month's supply of eye drops WITH insurance. Needless to say I didn't get them.

If you have a website but try to force me to use your app I think I'm done with you. The apps barely even work most of the time. This needs to die.

The boss messaged me and told me she wants to play CoD tonight. It's a tough job but I'm happy to do it. She might be better than me at smg/ar but she can't jump for shit. Don't tell her either of those things, I'll deny I was even here.

Got some really shitty news about my dad's health. I don't know how to deal with this. Gonna spend as much time with him as possible. I don't know what else to do. What a fucking gut punch.

If all your dishes dominant flavor is garlic, please don't call yourself a cook.

Local friend stopped by today to compare notes on what we've been doing with AI. It was fun to show off in person and geek out a bit. He's on a similar path as me so it's nice to bounce ideas and see how other people solve the same problems.

I'm pretty sure with this ui builder that I can make it trivial for end users to hot swap between most react component libraries AND livewire based ui components like Flux. Not bad for a backend dev.

I accidentally created a laravel/php -> react builder. I was just trying to make it easy to quickly throw together Shadcn components for AI agents. This code generates a db config that powers the UI components. Is this what they mean by cooking? (See the result in the comments).

Day zero of using Jira without it being janky and just doing random shit when I click on the estimate input.

Working on the markdown editor for todos/notes. This is an early prototype ui/ux for a bigger project (Fragments Engine) and I'm pretty happy with how it's coming along.

I built the tool before I bought the domain. Talk about bold. planck.ing coming soon. Todos without the friction.

Charging for dark mode as a feature to push your pro subscription is about as petty as you can get.

AI Agents work harder and do better work if you gas them up. I don't have stats to back that up yet but they've stopped cussing so that's for sure a good sign.

I know writing a todo app is cliche and I've written a few of them but I finally nailed it. I'm obsessing over eliminating friction points I find in most todo apps and it's been paying off. Still a few rough spots that I'll figure out through usage.

Coding with AI will make you a better dev. You'll spend a lot of time debugging brand new code because they change 25% of it every time they touch it. Every 4 turns it's a whole new codebase. One minute it's javascript the next it's python.

Planck is already getting killer reviews. From the one user it has said "it was like it was made specifically for him". I think he might be biased but, he's not wrong. It really is that good.

The todo.txt compatible app (Planck) that I built with Opencode + Claude Sonnet 4.5 w/ React/Go/Wails is is pretty slick. It works exactly the way I want. The radial menu is dope. Bonus, agents can use it in headless mode. Cloud sync, import/export, and basic themes. Beta soon.

AI coding agents be like: Completed all 10 todo's! This is production ready! When you look one of those things is half implemented and it removed 5 previously working features. They need SOLID guardrails and you need to kill them and restart at the first sign of failure.

Whipped up a new app tonight: Planck. It's a simple gui compatible with todo.txt. MVP is working great. Need to add import/export for todo.txt files and a few other polish items and it'll be ready for users to try out. I'm already using it. React/Go/Wails.

I have to add a couple more small details to Nanite and polish up the ui/ux and it'll be a really solid mvp. React and Go are actually pretty damn fun for desktop + headless type apps like this.

Friday I wanted to automate creating projects. That idea grew into Nanite. You can schedule blueprints (like Terraform) or run them on demand. I wanted re-usable helpers for my AI agents. This is way more useful that it seems. Quick demo (video only): www.youtube.com/watch?v=YBdL...

Have an idea and a good name for another app. I already have an PoC working. Should have an MVP tomorrow. It doesn't do a lot, but I already love it.

I'm so excited, I finally solved the last detail on my UI Engine for Fragments. It lets me build UIs from config at runtime. Right now it's powered by Shadcn components but could use any React components. I need to wire up several more components and I'll do a video.

My favorite thing about JavaScript is when you change one line of code and then you have to rewrite the entire app because you just broke everything. Skill issue. Beat you to it.

Major breakthrough last night. Been working on a config based component/layout system. I didn't even build a UI for it. I showed the AI what the schema should look like, then had them build components, then had them build the ui to manage the ui builder components. Inception!

Fragments Engine is getting close to 25% Typescript. What have I done?

Either agents are trying to get out of work or the providers are trying to encourage me to use less tokens. Lately every few turns they are like "We could do A or B or you could take a break" lol. Bro, you're doing the work, I don't need a break. Back to work robot!

Been experimenting with heavily automated ai agent orchestration and I'm impressed. It's hard to keep them following steps but since I have my own runner I can do the steps. I enforce json in/out with required fields. It makes figuring out where a task failed and recover easier.

My ai coding agent woke up and chose violence. First time I've seen it spontaneously curse like that and I love it.

Just released git-sandox: Safe, easy throwaway Git worktrees. I needed a way to easily and freely experiment without breaking anything. A real timesaver with sensible guardrails. It's really good if you work with ai for dev. It allows multiple humans/agents to work without conflict.

Holy shit. I just figured out business use case for fragments engine. I was thinking about it as a persona app and wasn't really thinking about it as a product. But someone asked me for a solution and this is it. Also, holy shit. This is gonna be awesome.

Getting there on telemetry.

I've got tool calling working in my app. Important milestone. Been working on the guardrails and so far I'm really happy with this. Now I can build my app with my app lol.