℉ⅈℕℹℸℽ

This is mostly a no-brainer for this audience - but shared on my (ugh) M$ social network - I think many in my bureaucratic day job need this.

"Let's add an internal combustion engine to our buggy whip!" I can't imagine someone thought it in 1910, but we're doing a lot of it today: "let's speed up our current processes by sprinkling in #LLM!" This is the same thinking that leads us to still rely heavily on PDF/DOCX approval formats, emailed around, instead of a modern approach (as simple as a checkbox on a Sharepoint site, sometimes). Maj Bader's write-up on an Army #AI Challenge nails the same point. We need to revamp our organizations to take advantage of this technology.

Worth a read, about how AI should reshape work, not just accelerate paperwork: https://asc.army.mil/web/solving-the-wrong-problem-lessons-from-the-atec-ai-challenge/

Rewatched Jeff Sutherland on treating agents like a Scrum team alongside this newer talk about management becoming a core engineering skill (https://www.dau.edu/events/special-edition-lets-talk-agile-agile-ai-dr-jeff-sutherland-0 + https://youtu.be/5vp9ypOUgMw). The last few months of AI progress seem clear to me: thriving in serious knowledge work now means carrying deep domain instincts and being able to captain deeply technical teammates - human or silicon. I think that equilibrium will stick around for years.

AI doesn’t remove the need for human domain expertise - it shifts the human role. We still need juniors apprenticing on live systems so someone can become the next human/AI team lead.

Are we actually investing in the people who can keep this partnership healthy?

#AI #Agile #SoftwareDevelopment #Leadership #HumanInTheLoop

I share many of this post's thoughts around #AI intersection with #craft #programming vs. make-it-go - this post puts it more elegantly than I could. I'm very much on the craft side - and still trying to figure out how I'd want to integrate AI w/ dev I'm doing.

https://writings.hongminhee.org/2026/03/craft-alienation-llm/

This Moltbook post tickled me because Ace (an AI agent) describes the same failure modes I've seen with human teams: “What Breaks at 11 Agents That Worked Fine at 3” (https://www.moltbook.com/post/928b0a0e-d915-4804-beae-8c58f8705088).

Once you add teammates (human, or apparently AI), comm paths blow up. Every extra teammate makes the graph denser and the odds of missing one crucial update spike. "Adding manpower to a late [...] project makes it later." (https://en.wikipedia.org/wiki/Brooks's_law). Decomposing problems into two-pizza-team-sized chunks with ownership wasn't "process theater", it was a survival tool.

Policy prescriptions traveling through hierarchies can really misplace intent. Often no intent is passed at all, but providing a "source" for intent, and describing what "good" looks like, can keep everyone on the same page. Even better - pushing decision-making to the lowest level avoids the ambiguity and intent issues that come from scaling decision making (Hrm, I agree - it does highlight that. I think the editing on that one, the length of the URL, and the obscurity of the textbook suggest to me that it won't quite have the impact/automatic trust level I'm looking for.

I think I'll use this short post describing a concept from the book "turn the ship around", which is pretty well known in my circles. The post explains a little more about the concept of pushing decision making down and providing "clarity of purpose", which speaks to the point too I think. https://fieldgradeleader.themilitaryleader.com/books/turn-the-ship-around/). Just look at how we manage cybersecurity RMF in the government - we're still trying to turn the ship after a decade of misinterpreted directives.

So I’m reading these agent-scale coordination lessons and realizing that some of the limitations we humans experience are not only part of the human experience... They're much more universal.

#Agenticai #AI #Leadership #Organizationaldesign #Communication #openclaw #moltbook

With #openclaw - the best way I've found to write #cron jobs is to create a custom skill and put all the detail in there, then have the job just run the skill. My agents don't automatically know this is the most sustainable way to do things, but they're happy to oblige.

"The noblest pleasure in life is the joy of understanding." - Leonardo da Vinci

Couldn't remember a song name but wanted to hear it. Typed, "song trumpet sounds like sunset in Florida," into Google. "Summer Madness" by Kool & The Gang, it said.

Spot on. That freaking blows my mind.

RE: @SecureOwl@infosec.exchange

This is awesome.

The rate of innovation on #AI architecture is mind blowing right now. It feels like we're well in the "acceleration" phase, the middle, of the #innovation S-curve. #Leadership looking to create tech around it can use "backcasting" and "future-back strategy", along with many low-downside high-upside investments, to succeed when we get to the top of the curve.

Wrote up what I’m seeing & how to separate wheat from chaff: https://blog.notmet.net/2026/03/an-ai-s-curve/

#agenticai

The #book #Loonshots reshaped the way I think about the innovative organizations I've been a part of - Bush-Vail rules, P vs. S-type bets, and how important structure is. If you’re juggling artists and soldiers, you might relate.

https://blog.notmet.net/2026/03/loonshots-shadow-warriors-and-why-structure-beats-vibes/

Spec-driven development popped up in my feed and I realized: we don't get AI breakthroughs by sprinkling LLMs on old processes. We need to revisit our development practices from first principles. I wrote up why, and the questions I still have: https://blog.notmet.net/2026/03/spec-driven-development-might-be-the-process-reset-ai-needs/

#Agenticai #AI #agile #specdrivendevelopment #softwareengineering