I do a good portion of my work on my MacBook Air M1. It's not super flashy and doesn't have impressive specs. It is the base model, without upgrades. It's the first model to feature the Apple Silicon chips, and it's almost 6 years old. There are stickers on the back to remind me of conferences and projects I've been a part of. I enjoy using it, and it's comfortable.
In the past couple of years, an issue has kept recurring that has only gotten more persistent and irritating. The storage is full. As I said, it is a base model, so the storage isn't anything fancy, just 256GB of space.
When I checked the storage, it said that my system data was taking up a whopping 120GB of space. Add in my documents, photos, applications, and MacOS, and you can see pretty fast how I kept approaching that 256GB limit.
If you are like me, the phrase "system data" doesn't immediately tell me much about where to look for a culprit. After a bit of poking around, I found that it's a grab bag of things that don't fit neatly into categories like documents, photos, applications, etc. Basically, it's your home folder, application caches, and if you are a developer like me, the projects you are working on.
I checked around my home folder and found some things to clean out and remove. It helped a little, but it wasn't anything close to the missing 120GB of space. It quickly filled back up
I grabbed the application GrandPerspective, which displays the content of your drive in a nice graphical format. It showed me some additional caches, local LLM models I downloaded with Ollama, and a few other things I purged. But I didn't find anything like 120GB. All evidence pointed to a large cache of files in my home folder, but I wasn't finding it.
I consulted Claude and Gemini to find out what was going on. They explained that it was something in my home folder or user space that was eating up a lot of resources. They gave suggestions and led me down quite a few rabbit holes, but there was no lasting relief from the oppressive space constraints.
Finally, I walked back through my drive, which I'd become quite familiar with by this time. I had Gemini help me walk through everything, and it asked me to list the contents of my /Users folder. Now, I thought this was silly because I'm the only person who has used this computer, but I did it anyway. Sure enough, there was another user's folder consuming 80GB+ of space right there in plain view.
Worse yet, that second user was me. I had decided to change the name of my user for some reason a couple of years back and forgot about it. I'd left the account files there, silently lurking even though the account itself was gone.
It was a cache of random files sitting in the user space. Just like I was looking for from the beginning. It just wasn't my current user. It was the ghost of my prior user.
I didn't need fancy tools or LLMs. One ls /Users would have done it.
Sometimes the answer is right there.
You just need to stop and look at it.
This week on the blog
Please take a minute to check out what is new on my blog this week:
CTEs: Making your SQL readable - The SQL for Python Developers series moves forward, showing you how to improve the readability of your subqueries with Common Table Expressions (CTEs).
Road to Agentic Notes - How to get started documenting your personal context so that you can use it with LLMs and reference it in the future yourself.
What I'm learning
This week, I listened to a Beyond the Noise Podcast episode with guest Kelsey Hightower. I've been interested in the discussion around LLMs, the impact they are having, and what their capabilities really are on our work, our lives, and our productivity.
It was a great conversation, and one thing I took away was a quote from Kelsey.
And so when people talk about LLMs and say, my God, look how amazing this is. We're writing more code than ever. ... But then I asked, okay, is your backlog gone? Because ideally, if you're getting a 10x productivity boost, then your backlog should be disappearing at a rate we've never seen before.
I've been doing a lot of work with LLMs over the past year or so. I've done some vibe-coding. I've done some strategy work, some brainstorming, and some editing. You name it. I've been very productive.
My personal backlog is larger than ever.
I haven't decided what that means, but it's now on my radar.
Links worth your time
Next week, I'll be moving this section out of my newsletter. If you have been enjoying these, you can continue to find them on my website
Google announces TurboQuant - Google announced TurboQuant, a technology that reduces the resource consumption of models while speeding up their performance. Seems like a significant win and I look forward to seeing what this does in the local LLM space.
Meta TRIBE v2 - Meta is introducing the TRIBE v2 foundation model, which simulates the human brain. They claim that it can predict how the brains of people it has not interacted with before respond to stimuli.
Google discloses quantum vulnerabilities in cryptocurrency - Google researchers highlight the potential for future quantum computers to break current cryptocurrency security (ECDLP-256). They propose a "Responsible Disclosure" model using zero-knowledge proofs to verify threats without aiding attackers, and urge a transition to post-quantum cryptography (PQC) sooner than later. Elliptic Curve Cryptography is easier to break with quantum computing than previously thought.
So many interesting technological developments are happening in the world right now.
I hope that you have some time to explore them.
I hope that you have had some time this week to spend with family and friends.
I hope that your backlog is right where you want it to be.
Jamal
That's it for this week. If this was useful, the best thing you can do is share it with someone who would get something out of it.
Not subscribed yet? Fix that here: jamalhansen.beehiiv.com