Don't Pick a Side

Steelmanning the loudest debates in AI engineering — then landing on a rule, not a tribe.

Latest

Parallelism vs coherence

Oct 20, 2026

The fifteenth debate I keep having with myself: when do more agents shorten the work, and when do they only multiply the effort required to trust the result?

Protected vs left behind

Oct 15, 2026

A special issue on Europe's rules, the new personal agents, and what we lose while both sides argue.

Skills vs memory

Oct 13, 2026

The fourteenth debate I keep having with myself: should an agent get better by remembering me, or by loading a procedure for the work?

Clean evals vs dirty reality

Sep 15, 2026

A clean benchmark measures capability. A dirty one measures whether you can sleep.

Context vs contracts

Sep 6, 2026

The twelfth debate I keep having with myself: when an agent gets something wrong, do I give it more background or write a rule it cannot politely forget?