Software Engineering

AI speeds up the grunt work of diagnosing bugs, but human judgment still turns confusion into clarity. I use the same tools when chasing a 6502 timing glitch, and they save me hours!

Debugging eats time because you spend most of it re-creating context, not typing fixes. I've been leaning on AI tools for a while now, and what they do best is collapse the search space so I spend my brain cycles on reasoning, not rummaging; they speed up the boring, repetitive parts without taking ownership of the investigation. I use five classes of helpers that actually save hours. Log and trace summarizers turn thousands of lines into a timeline with anomalies and “this changed before it broke” moments. A repo-aware Q&A surface where state is mutated and who owns which boundary,...

Good recap of the Clawdbot saga and what changed along the way.

I've migrated dozens of Mule projects over the years, so trust me when I say every DataWeave file needs an actual rewrite, not a find-and-replace. The header syntax changed , percent prefixes are gone except for %dw, fun uses =, and output/var/namespace lines are different, so those little header quirks will bite you first. The thing that breaks the most code is when/otherwise turning into if/else, because the condition and value swap places; that alone will make tests explode. Also, using(...) blocks become do { var ... --- body }, which is more verbose but way more flexible once you...