Tight loops and feedback
One of the interesting, and frustrating things about the course I'm doing in Adaptive Systems is its generality; being quite cybernetic, it can be applied or observed in many contexts (biology, robotics, environmental, organisational), and it's down to us to develop our own idea of what it means.
What this means is that I've been reading a few papers on robotics, and in particular developed a taste for the work of Rodney Brooks. At a time when the GOFAI crowd were obsessed with symbolic approaches with AI, he was rolling up his sleeves and encouraging his group at MIT to build stuff in the real world.
Battling Reality is one of my favourite papers so far, for its frankness ("Many of the preconceived notions entertained before we started building our robots turned out to be misguided"), and its clear message that the act of construction was in itself educational. In particular,
"Unless you design, build, experiment and test in the real world in a tight loop, you can spend a lot of time on the wrong problems"
"Understanding the environment and truly discovering the constraints on cognition are more readily done by building one robot than by thinking grand thoughts for a long time."
…which set a few agile/lean bells ringing for me (including the one marked "confirmation bias", of course).
There's also an air of Warwick-style showmanship to someone who authors a paper titled Fast, cheap and out of control: a robot invasion of the solar system, too…