🌸 Sun, Mar 1, 2026 🌸
gemini-cli with Gemini 3 feels faster and smarter. Might be placebo, though.
Only God knows C++ to A. Bjarne Stroustrup knows C++ to B. And if you know C++ to C, pat yourself on the shoulder.
I am getting into the rhythm of AI coding. I write the code by myself, and ask AI to fix bugs. It is pretty good at spotting typos, and sometimes it can add some good ideas and polishing touches.
I thought today was Sunday; what a disappointment.
YouTube, can you fix the shuffle feature? I have 500+ in the playlist, and it only plays a handful from that in the loop.
I am done with that game. Next…
https://mika314.itch.io/endless-drive
“if at first you don’t succeed try try again” is that definition of insanity?
Is that addiction? I keep telling myself: “I should stop using AI, it is not good for my mental health”. And I keep using it?
Debug renders sometimes look beautiful. Here I was trying to find out how the normalized vector from the right headlight of the car to the surface looks like.
🌸 Mon, Feb 9, 2026 🌸
Darn, having phantom tobacco smells since yesterday, after 4 months of no long COVID symptoms, I am getting it back.
🌸 Sat, Feb 7, 2026 🌸
What’s the deal with AI’s obsession with always running a clean build? Every iteration is 5-10 minutes wasted time.
🌸 Fri, Feb 6, 2026 🌸
Kill switch philosophy in game development
First, call it Feature.Enabled; don’t call it a kill switch. Feature.Enabled is easier to understand and removes cognitive load when parsing configuration files or looking at a dashboard panel.
Second, Feature.Enabled should be on by default. Disabling a feature should require explicit human input—this is how the kill switch is engaged.
🌸 Fri, Feb 6, 2026 🌸
There is a new way of self-gaslighting. It used to be that you’d look at old code and think, “Who is the idiot that wrote this?” - only to run git blame and realize it was you. But now, I run git blame, see my own name, and think, “There is no way I wrote that stupid code; I must have vibecoded it.” 😄
A line of code is worth 1000 words of documentation.
I am learning about screen space shadow techniques, and I do not understand one part: why we need to ray march in the loop for that short distance, why we just cannot project the depth buffer to the shadow map from the point of view of the light. In my mind, it is N-raymarching steps faster without losing any quality. I understand the limitation of screen space shadow techniques: only visible geometry will cast a shadow, and that argument applies to the N-raymarching steps. So don’t bring me that argument. Maybe constructing the highmap mesh is expensive?
Generic - good. Branded - bad. Generic is cheap and compatible. Branded is expensive and has its own useless marketing twist.
I feel Apple has an edge in the AI race because of the Apple M-series chips. No one has figured out a better chip in terms of performance-to-power ratio. Apple could build data centers using these chips and make inference up to 60% cheaper than competitors.
TIL: Ctrl+C / Ctrl+V does work in Blender. 😲 Am I that old? Or was it recently added?
The last iOS update finally fixed one annoying minor inconvenience. I had one unread iMessage, and before the update my iPhone was refusing to show that unread message. Finally, after the update, it actually showed the message. And that annoying red bubble with the lonely 1 is gone.
Here is a simple iPhone app called Focus Friend. The intent of the app is to reduce smartphone screen time. I’m not an extensive iPhone user, but I thought I just wanted to make a lot of socks and decorate the room. But I realized that app has some positive aspects; it reminds me to take a break, go to the bathroom, drink water, and stretch. TLDR; it became for me an UnFocus Friend.
Apparently, youtube-dl is no longer maintained, so to download YouTube videos now we should use yt-dlp.
🌸 Fri, Jan 9, 2026 🌸
It is quite unsettling that you can publicly state clearly untruthful things despite the video evidence. And a huge percentage of people would still blindly support you.
🌸 Thu, Jan 8, 2026 🌸
I am wondering if building AI data centers based on ARM architecture can reduce energy consumption and cooling requirements. Since ARM processors are highly efficient for running inference, would they offer a significant advantage?
🌸 Thu, Jan 8, 2026 🌸
I just got the new update on my Meta Ray-Ban Display glasses—handwriting and teleprompter features. Quite interesting! In true Meta fashion, though, it’s not without annoying bugs. I have been stuck at the pairing stage with WhatsApp because the QR reader is bugged in various ways.
Have you heard about the small JS engine for embedded devices?
https://github.com/bellard/mquickjs
1 - 2 - 3 - 4 - 5 - 6 - 7 - 8 - 9 - 10 - 11 - 12 - 13 - 14 - 15 - 16 - 17 - 18 - 19 - 20 - 21 - 22 - 23 - 24 - 25 - 26 - 27 - 28 - 29 - 30 - 31 - 32 - 33 - 34 - 35 - 36 - 37 - 38 - 39 - 40 - 41 - 42 - 43 - 44 - 45 - 46 - 47 - 48 - 49 - 50 - 51 - 52 - 53 - 54 - 55 - 56 - Next