Jeff Miller (orange hatband)Mar 6
#aonw Opening Circle. April Jefferson with a reminder and an introduction to
Open Spaces, 5 Principles, 1 Law; butterflies, honeybees, ...and how to pull sticky notes without curling.
Show threadJeff Miller (orange hatband)Mar 6
Elevator pitches for Open Spaces topics at AONW. 4 corners of the gym, 2 spaces in each of 2 classrooms, the library, and "wherever".
Show threadJeff Miller (orange hatband)Mar 6

perpetual retrospective space on the large landscape decorated board - ok to put stickies.

colored balloons marking gathering spaces in larger rooms (2x or 4x)

#aonw

Show threadJeff Miller (orange hatband)
For session notes, see LinkedIn under the hashtag #aonw2026 . (AONW org: https://www.linkedin.com/company/agile-open-northwest )
AGILE OPEN NORTHWEST | LinkedIn

AGILE OPEN NORTHWEST | 397 followers on LinkedIn. AONW is an annual conference about agile practices and techniques. | AONW is an annual conference about agile practices and techniques. Using Open Space, the participants themselves make the conference they want to attend. Come prepared to share your latest ideas, challenges, hopes, experiences and experiments!

Mar 6 at 7:01pmWeb
Show threadJeff Miller (orange hatband)Mar 6
Topics for the sessions are moderately heavy on LLMs and agentic techniques.
Show threadJeff Miller (orange hatband)Mar 6
"tasks that were previously not worth it are now done" (e.g. docs and warning linting; part of a session on agent augmented programming and behavior shaping)
Show threadJeff Miller (orange hatband)Mar 6
topic wanted? "Harm Reduction in LLM augmented software development"
Show threadJeff Miller (orange hatband)Mar 6
A link on "The Green Path" of behavior change before context and advocacy (in software coaching context) - Marian Heather Hartman https://changethatsticks.substack.com/p/when-we-stop-teaching-and-start-designing
When We Stop Teaching and Start Designing

Why Behavior Comes Before Explanation

Change that Sticks
Show threadJeff Miller (orange hatband)Mar 6
A terrible metaphor from Blue Eyed Samurai about the value of old vs. new techniques re pre LLMs and new LLM augmented techniques.
Show threadJeff Miller (orange hatband)Mar 6
A happy spirit haunting this session: Arlo Belshee. #aonw2026
Show threadJeff Miller (orange hatband)Mar 6

Listening to a discussion on evolving and growing skill and ability in agile coaching and I think about Alistair Cockburn and the Crystal family of just enough method for the scale.

#aonw2026

Show threadeighteyesMar 6
i'm curious what agile means in a world of solo teaming. i'm doing product and engineering at my job, which would have been 6-8 people in the before times. the bottleneck isn't pre-ship planning, it's post-ship review. my sprint cycle is daily. there's no sense in the time overhead of ceremony.
Show threadJeff Miller (orange hatband)Mar 6
@eighteyes "Theory of Constraints" is showing up for me as a place to refer processes. It specifically looks at constraints and how they suggest what to do when they change or are different from what were present when a process was described or designed.
Show threadJeff Miller (orange hatband)Mar 6
@eighteyes The other reference that surprised me (okay I am a little slow) as increasingly relevant was automated code analysis. When coding isn't the bottleneck in development, other parts take more stress, and may need to change.
Show threadeighteyesMar 6
@jmeowmeow my experience with automated analysis is it is prone to miss things if you're applying it broadly. everyone working with llms should know about context rot and the need to combine mechanical with inference based approaches. for example. i just tackled a list of 50 bugs in my software project. giving the whole list to the ai in one bunch, only solved about 10 of them, the easy ones. by forcing it to iterate, fix then validate ( using tooling ), to think about each one as an isolated unit, was much more successful.
Show threadJeff Miller (orange hatband)Mar 6

@eighteyes Small tight iterative slices, each one verified, appear to be a useful pattern for agentic augmented coding.

Lada Kesseler had a post about patterns and antipatterns in augmented coding which I appreciated.

Show threadeighteyesMar 6
@jmeowmeow well said. i agree. how do you find that delta? the map is not the territory.
Show threadJeff Miller (orange hatband)Mar 6
@eighteyes I am not enough practiced in these perspectives to give examples, but I just heard the next chair over in the agile coaching session say that Gerald Weinberg's approach was to find the most troublesome point (constraint or blockage in a process) and investigate that.
Show threadJeff Miller (orange hatband)Mar 6
@eighteyes "people getting stuck in processes" is another topic from the coaching circle
Show threadJeff Miller (orange hatband)Mar 6
@eighteyes What people are saying in the coaches circle is that ways out of being stuck include either referring to the "why" is important (principles), or leading by example (using techniques that they are confident will improve the situation)
Show threadeighteyesMar 6
@jmeowmeow agreed, i've found that especially with LLMs, giving it the 'why', and providing tight product-focused context, gives it the fuel it needs when it has to guess. you can't solve everything with specs, inevitably it will guess, and you want it to guess right. setting context and vision is the role of leadership.
Show threadeighteyesMar 6
@jmeowmeow one of the great challenges of using AI on teams, is that everyone has a different mental model of software and how it's built. by systemizing a given set of mental augmentation tools, you're nerfing folks who don't think in that way. i don't have a solution for this, except to build 10x tools which makes all prior knowledge obsolete. :D
Show threadJeff Miller (orange hatband)Mar 6
@eighteyes There's also the fear of a rug-pull from service suppliers, not unjustified to those who have been through a few boom and crashes.
Show threadeighteyesMar 6
@jmeowmeow do you mean the llm providers? yea, we are in the uber-growth, vc subsidized period, but we're already at the point where you can get adequate results for 80% of software dev needs from a $3k local hardware investment. that is the floor the frontier models will have to beat. and it's always rising. Uber had fixed costs in drivers / cars. Inference costs for value will go down over time. I'm not too worried about it, nothing I can do about it anyway. I'm done with coding by hand. Someone will offer heirloom-quality, hand-crafted software, but I've got too many things to build. lol.
Show threadJeff Miller (orange hatband)Mar 6
@eighteyes I looked at your site and you look very busy. Even a piano tutor. It's been a while since I've dusted off my keyboard.
Show threadeighteyesMar 6
@jmeowmeow that was the last non-ai project i made. :D i'm only getting started.