Automation makes sense, especially in a world where data is ever more abundant. However automation hides away the process, and if done in a non-deterministically (for example by using machine-learning) may be the source of huge headaches. Automate responsibly, and use deterministic workflows whenever you can.

Knowledge was already democratized (books and public education), and now even technology is becoming unaffordable because of some vague promise of super-intelligence (whatever the f**k that means)

In the software world "open source" has a well defined meaning: if a program is actually open source, you should be able to rebuild it youself from (guess what) the source code. Open weight models are not open source! They might share some of its characteristics, such as (for certain licenses) the ability to be used for commercial purposes, and have derived work generated from them However, without access to the actual training data they will never be open source. Not even close.

Given all the praise I heard about Cursor AI, after a number of months trying to use local LLMs on my laptop with limited success, I decided to spend 20 USD of my hard earned salary to witness this amazing new tool in all of its glory. The results have been a far cry from all the "10x", and "100x" improvements in productivity that have been claimed left and right.

Engineering shall never die.

Hot-take: vibe coding is a nothing burger.

On (re)using old-ish small form factor PCs VS single board computers for clustering.