AI Toolbox Turnout

AI Toolbox Turnout: Alex Ezell

2026-07-16T00:00:00+00:00

Meet Alex. Survivor of 30 years building web apps. Runs agent vs agent fight club. Cannot abide dead links.

On a normal working day, what’s actually open on your machine? Walk me through your real AI stack: which models, which agents, which tools?

I like the native apps primarily so it’s Claude Code and Codex (which is ChatGPT now I guess?). I like to have them spar against each other all day, writing and reviewing each other’s code.

I also have Zed windows open to those same repos among others but I don’t use AI tooling in Zed. It’s sort of just a porthole for me to peer into a project.

When I’m using Sprites, I use Herdr as my tmux replacement so I can easily keep track of what’s happening in all of the Sprites.

Where’s your line between letting an agent run wild and grabbing the wheel yourself? What have you learned to never hand off, and what do you now delegate without a second thought? Has that line moved in the last six months, and which way?

I have a high risk tolerance in general so I let the things run free. There’s a bit of intuition around doing some early planning and investigation. But, a lot of my work is “do more of what’s already there” so pointing it at existing implementations gets me a long way toward the goal.

I find that’s where the AIs have gotten much better in the last six months. They do more work to build context with surrounding code and the system as a whole so adding a LiveView or processing some new data into an existing LiveView is very easy.

The thing I still never hand off at all, what I don’t even grant agents access to, is databases. I just don’t want my AI making database queries on its own. Even if the access is read-only, I’ve been in enough Postgres incidents to draw that line. Sure, I’ll let it write code that does queries but that’s the first thing I review when it’s done. I guess that’s where my risk threshold lives.

When did you last spin up a Sprite, and what for? What did you do before Sprites?

I spun up a Sprite to do a bunch of link checking. More specifically, I asked Claude to write a script to do that from a Sprite.

I play around in the indieweb/smallweb world and lots of folks use gif buttons to link to other sites. If you remember the webring days, that’s what I’m talking about. I wanted to easily check a site’s button links to see which are dead. That work eventually turned into deadbuttons.online which anyone can use.

Before Sprites, I would have run all of this locally from my computer. But, I don’t really have checkpoints on my computer so when I made a big refactor that went completely wrong, as I did on the Sprite, I couldn’t just go back to the previous checkpoint. In this case, the Sprite checkpoint saved me.

What’s accreted on your setup that you’d be gutted to lose? The CLAUDE.md, the skills, the hooks, the conventions, the muscle memory. What took you a while to get right?

A few of my Sprites have agents on them with very specific memories of how to diagnose issues in our environments using some specific tools they have through Connectors. That knowledge and tooling took a while to all come together for the Sprite to easily know what to reach for and when.

I’ve been looking into ways to sort of “multihome” that memory so it’s not trapped.

What are you building on the side with this stuff, the thing that has nothing to do with Fly? What does it do and how did you build it?

I mentioned Deadbuttons above. I started Precider on Sprites. Precider helps folks find the right preworkout. The site itself is a straightforward Elixir/Phoenix web app.

The real beauty of AI and Sprites was having an agent do all the research for me. I started with asking the agent to list all the pre-workout companies it could find. Then, I took that list of URLs and had the agent find all the pre-workout products with their names and prices. The agent created a small database of these which I knocked into a quick little CLI script so I could review them quickly, rejecting those which were wrong.

Once that was done, I did try to have the agent gather all the data about the ingredients and their amounts. Filtering and sorting all these products by the amount of a specific ingredient is what Precider does that didn’t really exist before. That had mixed results and in the end took a lot of manual work from me but I still used the agent to feed me the product and give me a quick way to enter the data.

Once all of that data was collected, I asked the agent to create a SQLite file of it. I stood up the first version of the site on the Sprite for testing and iteration. Once it was in a good place, I used fly launch to stand it up on a Fly Machine.

What’s the part of your workflow a colleague thinks is unhinged, or overkill, or a waste of time, but you’d defend to the death?

I think I’m fairly bog standard and probably behind the times with most of this stuff.

I would argue that the native apps have something special about them that makes the results better in some way I can’t quite put my finger on. I have no data for that. Just vibes.

What did you think these tools were useless at a while ago that you now trust them with? And the reverse: what did you overtrust and have to walk back?

Letting an agent build a GenServer a while ago was a nightmare. It just never got the lifecycle of the messages right. That has changed in the last few months. I don’t know if it’s just more exposure to that pattern or better training but in the rare cases I’m writing a GenServer, it typically starts from scratch with the agent.

On the other hand, I often let agents go a bit astray with how they manage CSS classes and componentizing the HTML for a LiveView. It always looks fine but structurally it can be quite messy and repetitive where it shouldn’t be.

What can you flat-out do now that you just couldn’t before? Not the same thing faster, a genuinely new capability.

Write Go code for CLIs. As a person who has spent almost 30 years building web apps, the idea of writing a CLI or even a native OS app was pretty far outside of my wheelhouse (things like Electron notwithstanding). But, I do it a lot now and it’s because the agent takes care of the boilerplate and the structural pieces that would have taken me too long to learn to make the ROI make sense.

AI Toolbox Turnout: Lubien

2026-07-10T00:00:00+00:00

Meet Lubien. Maker and fixer of all things. Sprites power user. Serial cat collector.

On a normal working day, what’s actually open on your machine? Walk me through your real AI stack.

I’ve been using Zed for quite a while now and it has my Claude API key, which makes it stupidly easy for me to get started using LLMs anywhere. If the repo is on my own machine I just open Zed and choose the repo from “Recent projects”.

When I’m in a Sprite, I use zed ssh://sprite-name.sprite-org. I have a neat setup script that prepares Sprites for me to just SSH into them. Run the script once and I’m done.

I do have a single exception: since I want to be able to easily edit my blog, I have a Sprite connected to my Tailscale tailnet, so I don’t use the script above, just ssh machine-name.tailnet-subdomain.

Where’s your line between letting an agent run wild and grabbing the wheel yourself? Has that line moved in the last six months?

For quick prototyping I let the agent run wild first, then I start yelling at it to do better. I’ll often say “write me this LiveView that has X and Y, make sure to add tests” or “we need a new controller route for /path/to, make sure to add tests”, and the model most often uses mock data. The magic here lies in “make sure to add tests”, because I can start a TDD red-green cycle without any trouble at all.

I also ask it to componentize wherever possible. I pasted this on the Elixir Brasil Telegram group when asked about it:

Use components from core components or create new ones. The LiveView render function should be componentized as far as possible. If I see a lot of Tailwind classes it means you can componentize more and you should. Use slots to make it easy to position parts of the code. Instead of writing several Tailwind classes repeatedly you can also add collocated CSS with custom classes with @apply on Tailwind.

In the last 6 months the major change is that I need to yell less at it. Agents are increasingly better at understanding existing patterns before doing stuff. Before, I’d have to explicitly add “do it just like path/to/file.ex does”, but nowadays I just ask for what I want and assume it will do the grunt work and give me reasonable enough results.

When did you last spin up a Sprite, and what for? What did you do before Sprites?

This past Sunday. I was recording a series of videos for my YouTube channel about supply chain attacks and how to protect yourself from them, so I spun up a Sprite to PWN myself (lubien/not-a-scam-trust-me) and also show how to use some tools. Tune in to youtube.com/@lubiendev for more.

Before that I would have blindly trusted that code I cloned was safe enough to run. But this year, daily supply chain attacks have taught me otherwise.

I don’t code from my phone. I’m already doing it more than I should in my home office, so my phone is for talking to friends, watching brainrot and getting my data harvested by Big Ads.

What’s accreted on your setup that you’d be gutted to lose? The CLAUDE.md, the skills, the hooks, the muscle memory.

I don’t use skills other than whatever is already in a Sprite. For my Fly.io work it’s AGENTS.md from the Phoenix Framework, plus a few edits I made to teach agents about our architecture. I prefer yelling at the agent over spending time writing skills. I know there are some great skills out there that can automate stuff like checking Sentry, preparing plans, creating PRs, etc. I’m just OK with writing that down. Sorta feels like rubber ducking, but with my keyboard instead of with my cats (all 6 of them).

What are you building on the side with this stuff, the thing that has nothing to do with Fly?

comunidade.devsnorte.com. It started on phoenix.new (I promise this is not an ad), then later I ejected it so I could do things on my own with my friends. We use it to list events and other projects from Brazil’s North region (we kinda divide ourselves into 5 geopolitical regions, you know … continent-sized country stuff). I’m always coming up with new ideas to improve it and we get community feedback too, so our backlog is growing.

Where has AI crept into the parts of your life that aren’t code?

I hate AI support. It fails me 99% of the time. Surprisingly, I had one positive AI support case yesterday (after several failed ones) and it was from a driving school WhatsApp contact. Speaking of WhatsApp: Meta now has this feature that summarizes unread messages. It would be useful, since I have hundreds of them (all related to programming and communities), if I wasn’t chronically online.

What’s the part of your workflow a colleague thinks is unhinged, but you’d defend to the death?

I’m that colleague. I keep AI usage low, so I think it’s unhinged when I see folks automating their lives down to wiring up integrations for task boards, time sheets, replying on Slack, etc. Seems odd to me.

What did you think these tools were useless at that you now trust them with? And the reverse: what did you overtrust and have to walk back?

Back when ChatGPT first appeared, I tried to use it to port flyctl to Elixir, but it failed miserably from the beginning. I’m sure now Claude and friends could handle it given enough tokens and testing. Not that I want to do all that anymore. I don’t think I ever over-trusted AI. On the contrary, I downplayed it for a while.

What can you flat-out do now that you just couldn’t before? Not the same thing faster, a genuinely new capability.

Pretty shell scripts. I can now make em all. All my test shell scripts sucked, but now I use AI to write them for me. I never bothered with (nor wanted to) learning shell scripting, to be honest. I just googled things as needed.