Hello 👋, I'm Patrik
Latest articles and social posts
Latest Articles
I built pbp to solve a problem I was having with my growing collection of side projects. You probably have the same problem.
Changing how I consume music in 2025 to rely less on automated recommendations and get a deeper appreciation of it
In this follow up to my post about how I set goals that align with my values we dig into how I use Obsidian to actually do all the things I talked about in the previous post.
All the way from my core values down to what I decide to do day by day. We go through weekly and daily planning based on values and how to figure out how to bucket time by mapping out an ideal week.
Will generative AI take over what we used to believe to be fundamentally human - creativity?
This is how you get off the couch and achieve everything in life. A step by step guide.
As AI becomes better at writing and creating, is there intrinsic value in human created content?
How do we separate knowledge? How do we know that anything is correct?
An informative guide to managing time effectively using the methodology of 'forwards thinking - backwards planning', especially for parents.
Exploring the misconception of multitasking, its cognitive costs, overconfidence in abilities, and negative health implications.
An exploration into the harmful effects of overworking and constant productivity, offering insights on how to maintain a healthy work-life balance.
Latest Posts
I wonder if we will see a rise of RUP (Rational Unified Process) and similar heavy-weight docs focused methodologies to better instruct AI agents and when turnaround time will be minimal
Being able to put a chain of AI-agents to work on creating PoCs in like 30 minutes with full test coverage and suprisingly good code quality is wild. But - it's like creating a conference talk. 1 minute of talk = 1 hour of prep. So 30 minutes of AI work = 30 hours of human finish. Not
feasible (yet) to think of the 30 minutes as the full work. This is where tech AI hypers are wrong atm and why CEOs will be in for a rude awakening.
Idk what it is but I own two Hans Zimmer vinyl records. One is the Interstellar soundtrack and the other is the Dune Part 2 soundtrack. Love Zimmer but both records came with factory defects. Not ok since I had to sell one of my children to pay for the records.
The irony that Microsoft blocks access to AI agents fetching webpage data when pushing agents/CoPilot as the future of everything
One thing I've noticed when working with AI agents is how extremely easy it is to get sidetracked with scopecreep when you can just fire off another agent to go do something. The age old "focus on one thing at a time" is still relevant as agents never actually 100% finish the tasks for me.
It's a lot of building more WIP.
Immich is such a promising project. After running it for a few versions I have concluded that it's not yet read for primetime production use (for me at least).
Omarchy 3, now with a red dot.
How have I not owned a 3d printer before now? My cable management game is about to hit a new level.
Somehow gen ai lead to it being acceptable to say "Yeah I'm not sure the content in the document is totally correct". Not a good thing. You deliver it you own it. Doesn't matter how you created it.
More and more "normal people" I talk to keep saying that "I'm starting to doubt this whole AI hype. We are not really seeing any tangible results".
Having a proxmox host going 0 byte disk sure was interesting. So much strange things happening when disk runs out - ended up getting to bored and just rebooted after putting disk back.
It's gonna be a long time before letting LLMs run normal people browsers will be a non idiotic thing to do. "Our testing has identified scenarios where Claude could be manipulated to: Extract and share sensitive information with bad actors. Delete important files. Perform unintended
actions on websites that could result in harm to you or others.". I mean this is from the actual providers own support article... https://support.anthropic.com/en/articles/12012173-getting-started-with-claude-for-chrome
Tip of the day: Remember to set the right power profile when working on a laptop. "Balanced" is turtleemoji throttling
Just found out about floating windows in Zellij. So nice for a (or more!) bash scratchpads that can be toggled on/off with a keybind. Great stuff.
A hosts-file with all random AI domains would be helpful. Should start blocking at DNS level..
I wonder if we will see a rise of RUP (Rational Unified Process) and similar heavy-weight docs focused methodologies to better instruct AI agents and when turnaround time will be minimal
Being able to put a chain of AI-agents to work on creating PoCs in like 30 minutes with full test coverage and suprisingly good code quality is wild. But - it's like creating a conference talk. 1 minute of talk = 1 hour of prep. So 30 minutes of AI work = 30 hours of human finish. Not
feasible (yet) to think of the 30 minutes as the full work. This is where tech AI hypers are wrong atm and why CEOs will be in for a rude awakening.
Idk what it is but I own two Hans Zimmer vinyl records. One is the Interstellar soundtrack and the other is the Dune Part 2 soundtrack. Love Zimmer but both records came with factory defects. Not ok since I had to sell one of my children to pay for the records.
The irony that Microsoft blocks access to AI agents fetching webpage data when pushing agents/CoPilot as the future of everything
One thing I've noticed when working with AI agents is how extremely easy it is to get sidetracked with scopecreep when you can just fire off another agent to go do something. The age old "focus on one thing at a time" is still relevant as agents never actually 100% finish the tasks for me.
It's a lot of building more WIP.
Immich is such a promising project. After running it for a few versions I have concluded that it's not yet read for primetime production use (for me at least).
Omarchy 3, now with a red dot.
How have I not owned a 3d printer before now? My cable management game is about to hit a new level.
Somehow gen ai lead to it being acceptable to say "Yeah I'm not sure the content in the document is totally correct". Not a good thing. You deliver it you own it. Doesn't matter how you created it.
More and more "normal people" I talk to keep saying that "I'm starting to doubt this whole AI hype. We are not really seeing any tangible results".
Having a proxmox host going 0 byte disk sure was interesting. So much strange things happening when disk runs out - ended up getting to bored and just rebooted after putting disk back.
It's gonna be a long time before letting LLMs run normal people browsers will be a non idiotic thing to do. "Our testing has identified scenarios where Claude could be manipulated to: Extract and share sensitive information with bad actors. Delete important files. Perform unintended
actions on websites that could result in harm to you or others.". I mean this is from the actual providers own support article... https://support.anthropic.com/en/articles/12012173-getting-started-with-claude-for-chrome
Tip of the day: Remember to set the right power profile when working on a laptop. "Balanced" is turtleemoji throttling
Just found out about floating windows in Zellij. So nice for a (or more!) bash scratchpads that can be toggled on/off with a keybind. Great stuff.
A hosts-file with all random AI domains would be helpful. Should start blocking at DNS level..
