Jump to content

Resilience Coffee/2026/05/12

From Resilience Coffee

- Blog Post​ - Visualizing Systems

  - Some amount of mapping is value added
  - Without a map, domain is fuzzy
  - The act of constructing a map routes the thinking
  - "Stay where the light is" - Gary Klein: Streetlights and Shadows
  - "Hard pan" - tilling a field regularly leaves a field with a hard layer underneath
  - "Legibility" - Seeing like a State (book)
  - Confirmation bias
  - Finding the empty space with abstract maps is hard
  - Social construction is important
  - Be okay with being wrong (or just not knowing)
  - One of the 3 ways adaptive systems fail according to Woods: Being stuck to outdated patterns https://www.researchgate.net/publication/284324002_Basic_patterns_in_how_adaptive_systems_fail
  - Reference: Historical map of City of London with every road to scale, but a tourist map with topological relationships is more useful
  - "All models are wrong, some are useful" George Box. Know the boundaries of your model.
  - Normalizing changing context
  - Where are the most common gaps in knowledge or blank space you have all been seeing?
  - Arrived at a place from a happy path, but don't know how to get other people there
  - Developing positive communication channels
  - Every platform team is a tragedy of the commons: Joel Spolsky "All abstractions leak", O'Reilly PE book covers some of this, but there's a lot of difference when there is a captive audience; Don't have the buy in on there being a centralized platform
  - Long path to get a paradigm shift of RE to see how incidents are a window to peek through, not a problem to solve: incidents are a way to learn about the system actually work — not a thing to be avoided.
  - Cross-pollinating lessons learned
  - Applying resilience: sewage drain backup, technician could not directly observe the problem, choice: second opinion, or dig up neighbors driveway. Trust the person closest to situation. Do you know anyone that had this problem recently? Found a comparable situation. Thank the original operator. Trust because it was vendor recommending another vendor. Added monitoring systems.

- SWE in resilience space get feedback about whether we solved our problems but not all operators get that feedback, e.g. sewer inspectors/EMTs. - Impact of AI coding assistants - It's in how you use it, but we don't as an idustry know how to use it yet - Where are the incentives? - "Coding assistants feel like they're never enough. You wrote something faster, but could you have had more skills on top? Multi-agent? Used a different prompt? We're internally continually pushed to do "more" always" - Feedback loop is too slow for the incentives - Optimized for sales to executives - How do we know we've had a bad outcome - How (and when) to do Multi-region deployments - How to not work when you are sleeping