Finch looking in the window

I’ve heard of a raven, but what does it mean when I have a finch that’s gently rapping at my patio door?

My boys 6 and 3 have invented a creature called a “holesound” which has 2 arms and 4 legs, needles “like a hedgehog” that can poke you and it blows air out of its mouth to “make the instrument work but if it gets to your nose it catches on fire.” Sounds terrifying honestly

I quit checking in over there a while ago- it’s either ai generated garbage or “Grok is this true? Grok, Grok, Grok?”

That’s that non determinism sneaking in

My son 6: “Well I don’t want to keep living in a crack house” my wife: “what?” My son: “Well there are cracks in the wall over there”

Add this to the plan- “ok implementing the plan”

Weird thing I found out this AM- AI agents are not very good at TDD. They tend to flop around trying to fix errors. They are far better at writing tests that match what they've already created which doesn't give me a huge amount of confidence honestly

a man with long curly hair and a beard is eating a piece of cheesecake Alt: a man with long curly hair and a beard is eating a piece of cheesecake and making the chef’s kiss sign

Just saw Project Hail Mary and it was marvelous

The biggest lesson I’ve learned as a father is sometimes you just need to go on a tickle spree

Just spin ‘em

Hey Git worktree is actually pretty nice for running multiple agents at once- now I can pretend to successfully run 2 in parallel

Add in interview with the CEO

A picture of a chat window with chatgpt where it says, “So your assumption was mostly right, just slightly simplified” I responded, “Well your response was largely the same but over complicated so we’ll call it even” “Fair enough…” was the response

I think I might be going insane

They just keep increasing the size of mine

I made the mistake of asking Claude to help me plan a project and it proceeded to try to download the entire internet

Basically, while I felt a little unsure last week about whether AI will take all dev jobs I'm now certain it won't. I think this context problem is AI's achilles heel. It will be very fast at bootstrapping new things but as projects get bigger it will struggle more and more with context

One thing I noticed though is that on large existing projects it tends to make incorrect assumptions that could lead to some very subtle and very catastrophic bugs. I think the issue is that it can't pull the whole codebase into its context. If I wasn't vigilant it would have made some nasty bugs

It writes some pretty good code. It follows my style guide pretty well. I'm impressed.

I've used Claude code now for about a week so clearly I'm an expert /s

I wanted to give my first impressions 🧵

I feel like it’s worse than what was mentioned in the article by Vincent. I doubt they took the effort to load it into their llm, it probably just scooped it up at one point and they said “make an article” and didn’t really read it closely

Just fire all the developers

How do we know this wasn’t prompted?

We like to pretend that our fancy pattern matching code will gain consciousness any time now and solve all our problems

I just manually installed the LEMP stack on a server like someone out of the early to mid 2010s. It felt amazing

Now we’re talking

The Moltbook Experiment Failed YouTube video by The PrimeTime

This is too funny
youtu.be/6OXE65fjjsU?...

I got a streak checking my yahoo throw away email the other day 😅

It snowed

My response: "you made that up didn't you?" AI response: "Ha! You caught me. Yes, I did. I apologize for that - I shouldn't have made up the..."

This right here sums up my frustrations with AI coding perfectly

GitHub - Flowframe/laravel-trend: Generate trends for your models. Easily generate charts or reports. Generate trends for your models. Easily generate charts or reports. - Flowframe/laravel-trend

Just found out about github.com/Flowframe/la...

Man I could have used this before now... 😅