Adron Hall

Slava Ukraini. ✊🏻

I've gotta say, in some ways I'm ok with Starbucks and in other ways Starbucks just pisses me off to no end. One of my biggest complaints is how they do such a shit job training their baristas and don't teach them anything sane about coffee or espresso. They just teach them this made up bullshit Starbucks nonsense... which leaves the "baristas" being "Starbucks" employees NOT baristas. They aren't taught what a cappuccino actually is. They have those dumb ass sizings: Tall, Grande, and Venti. Latte is sort of a latte at Starbucks. A macchiato is just a nonsensical made up misappropriation of the word. No macchiato at Starbucks is anything like an actual macchiato that an actual trained barista would make. It's basically like Starbucks is just this giant revisionist bullshit factory. Starbucks really should just recede to being the "Sugar Flavored Coffee Factory" store or something. It's never really been about coffee except to lie about it. They're really just a fluff *luxury* brand for folks that don't have taste buds beyond "SUGAR". It's utterly nonsensical. Anyway, I'm done. The "partners" should get a bump in pay and the demands they've made. It shouldn't be that big of a deal. Starbucks has effectively closed all their "low performing" or "crime heavy" (re: they went union busting by closing the stores, same as Wal-mart does) stores so they have really no excuse to offer the union what they're requesting. It's really not that much.

I'm in the process of building out a new dev workflow via LLM and I've decided it's not even going to include Github. Github has become a brick wall of problems and an overall bottleneck. Something has to change so I'm going to start creating that change. Wish me luck!

A column type is a promise about what goes in it. Pick the wrong one and every row after fights you. Here's how to pick right. https://interlinedlist.com/blog/picking-the-right-column

The fastest working list you'll ever build starts with someone else's homework. Pick a template, tweak two things, done. https://interlinedlist.com/blog/the-60-second-list

Maximize Git Efficiency with Parallel Agents and Worktrees Discover how to overcome Git's one-tree problem by leveraging worktrees for parallel AI execution, enhancing your agents' efficiency. https://compositecode.blog/2026/08/19/worktrees/

Maximize Git Efficiency with Parallel Agents and Worktrees

Discover how to overcome Git's one-tree problem by leveraging worktrees for parallel AI execution, enhancing your agents' efficiency.

compositecode.blog

The Importance of Planning Before Coding: A Workflow Solution Learn how to prevent coding mishaps by implementing a structured planning process before AI agents make changes to your codebase. https://compositecode.blog/2026/08/20/plan-then-build/

The Importance of Planning Before Coding: A Workflow Solution

Learn how to prevent coding mishaps by implementing a structured planning process before AI agents make changes to your codebase.

compositecode.blog

Considering where things are, know anybody working on a Github/heavy source management replacement? Ya know, something that is super lean? I'm switching up my architecture now for some projects but it seems like - as some have already pointed out - Github is just not really that reliable for even basic projects at this point.

A blur of train travel.

Even wooden frames and more! 😃

Drool. 🤤

When multiple agents on a machine are running, do you use worktrees (like this https://compositecode.blog/2026/08/19/worktrees/) or do you use another method to split out your work? I've been pondering some new ideas, but wanted to ask the general community how you all are getting greater work separation to ensure reduced conflicts. So what ya got? Anybody? Anything?

Maximize Git Efficiency with Parallel Agents and Worktrees

Discover how to overcome Git's one-tree problem by leveraging worktrees for parallel AI execution, enhancing your agents' efficiency.

compositecode.blog

Imagine you're starting work on an existing repo. It's been around for a while and has had a lot of people working in that repo over time. It takes a long - LONG - time to get up to speed on that repo. Let's say you point Claude or Cursor at it instead to get up to speed. But - and I'm paraphrasing - Claude or Cursor effectively come back and shrug, saying the repo is a mess and it's difficult to tell what exactly is going on and lot's clean up work is needed. What would you call that kind of repo? That a seasoned engineer AND introspecting with an LLM both tell you it's a wreck? I was thinking something like "Toxic Repository" or maybe "Poison Pattern Repo"? There has got to be something more concise though. Got anything?

Giving Every Agent Its Own Branch: Git Worktrees for Parallel AI Work To note, I've written this post and at this time... it's kind of a *halfway* solution and really only fixes part of the problems of multiple agents running on a singular machine workspace. Give it a read and I'll have some suggested solutions coming your way real soon. https://compositecode.blog/2026/08/19/worktrees/

Maximize Git Efficiency with Parallel Agents and Worktrees

Discover how to overcome Git's one-tree problem by leveraging worktrees for parallel AI execution, enhancing your agents' efficiency.

compositecode.blog

Did a reaction vid to this years ago too... https://youtu.be/9T1y-DNgYY0?si=IBf6oOxW5_4Q7BD0

Reaction for Fervence's "Hezekiah"

Thrashing Code

youtu.be

A Research Agent on a Budget Leash, and How I’d Sequence the Whole Build https://compositecode.blog/2026/08/18/a-research-agent-on-a-budget-leash-and-how-id-sequence-the-whole-build/

Mastering Durable AI: Building an Efficient Research Agent

Discover how to build a durable AI research agent with effective workflows, cost management, and essential guardrails in this insightful guide.

compositecode.blog

Cost-Effective AI Strategies for InterlinedList Development https://compositecode.blog/2026/08/17/making-llm-features-cheap-enough-to-actually-ship/

Cost-Effective AI Strategies for InterlinedList Development

Explore cost-efficient AI strategies in Durable AI for InterlinedList. Learn how to optimize token usage and prevent unnecessary expenses.

compositecode.blog

A specifically interesting one! Trees on a tree stump in case you like trees on tree stumps!