LLMs Explained for Swift Developers in 5 Minutes Coming from a Swift and C background, I found LLMs surprisingly tough to learn—especially since most tutorials jump right into neural network internals and data science jargon. But understanding these models is more relevant than

It's now invoking tools after I submitted continue, but my input isn't shown on screen. Maybe it was doing something after all?

1/ Cannabis multi-state operators are flourishing and are the backbone of what the industry can become. $VRNOF $TCNNF $LEEEF $CRLBF $GTBIF $MSOS Let's take a look...

Hmmm... what am I supposed to do now? Is Droid still working or did it exit? No idea. I'll just submit 'continue' and see what happens

Well, this is interesting. Droid decided to just delete its implementation and replace it with placeholders. Did it lose track of where it was in the planning? I think it's safe to say that Spec Mode and planning are the weak points of Droid. Also interesting: Droid used the old

"Operation timed out..."

Removed the files. Now OpenAI doesn't want to play along. There's always something ;) "Error: An error occurred while processing your request. You can retry your request, or contact us through our help center at https://t.co/7mzXhz83dX if the error persists."

Great. Droid timed out. That's a good time to remove the existing files and start from scratch :)

THE KING IS BACK 👑

Ok. I reviewed the rest of the code this morning and it wasn't great. The main issue I believe is that Droid's spec mode is quite limited. ChatGPT's deep research is more thorough and easier to interact with. Droid went down the wrong path for the tokenizer. So we're going to

Energy abundance? We don’t nood energy abundance (apparently…) 🤦‍♂️

Americans should stop larping European antique culture. Have some self respect and put a massive Walmart parking lot there.

The German government is trying to cancel the EU combustion engine ban in 2035, while German car makers already have EVs with gas car ranges at gas car prices in 2025. Make it make sense

I noticed Droid did something naughty. I used the old ObservedObject. I guess you have explicitly mention only to use modern Swift/SwiftUI syntax in agents. md. Fixed in the latest commit. But... the architecture is still to complicated. Let's fix that too

Day 2. I'm still unable to add droids. We'll do that later.

First iteration based on a TextView pushed to Github. Not exactly what I wanted, but a start. I also tried to add a sub-agent (a beta feature) that benchmarks the parser at every iteration and saves the benchmark in a file. But alas, sub-agents don't seem to work reliably yet

Droid came up with a plan to test wrapping a few existing C, ObjC and Swift libraries and benchmarking them. I think that's a great idea. LFG! (and yes, it understood that)

Actually, let's plan what we need first. With shift-tab we can enter spec mode where Droid can research and plan. Since we already have our agents . md file, it knows what we plan to do.

Playing with a new feature: custom droids. These agents run in the background and can do specialized tasks. Not sure if it makes sense, but I'm going to add agents that continuously test for Markdown and KaTeX compatibility. Looks promising but maybe I should plan first

Sorry for the delay. I needed to fix an app in the App Store that broke for macOS 26 users (why Apple? Why??) and I somehow screwed up the Github repo (not sure how I managed that). I just added the first iteration of the https://t.co/3SWa64d5Ki file that explains the goals, dos

Join me live here: https://t.co/4jofaEQpno

If you want to code along, here's my referral code for $40 worth of credits and a Eurovision song-along GIF for no particular reason https://t.co/XJekDdfIP4

Repo created. https://t.co/6wF8ao5Iiu