cat mission.txt

Stay on the right side of the line.

Generative AI, in the hands of an operator who never built the underlying mental models, does not make that operator smarter. It makes them dependent. This publication exists to keep you on the other side of that line.

the-problem

An operator who already understands a system can use AI to move faster. The AI drafts; the operator checks the draft against a model they carry in their head. An operator who does not understand the system has nothing to check against — they cannot tell a good answer from a confident wrong one. The output ships because it looks right, not because anyone knows it is right.

The gap does not stay small. Every shortcut you take today is a mental model you do not build, and every model you skip makes the next answer harder to judge. The dependency compounds — quietly, across a career — until the day a system breaks in a way the AI has never seen and you are the one holding the pager.

An operator who never built the mental models is not in partnership with the AI — they are dependent on it, and dependency feels exactly like competence right up until the moment it fails. The result is work you cannot trace, cannot debug, and cannot defend. When someone asks why the change is safe, the honest answer is "the model said so." That is not an answer.

where-i-stand

The climb is the point. The understanding you fight for is the understanding you keep, and the understanding you keep is the only thing that holds when the tool goes quiet. Difficulty is not a bug in learning — it is the mechanism. Skip the friction and you skip the model the friction was building. 

$ set friction-budget 20%  # of your week, no AI
  rule[0] read the source
  rule[1] estimate before you prompt
  rule[2] rebuild a thing by hand
$ remaining 80% OK  # use every tool — you can judge the output

CTRL ALT PRESS is anti-hype on purpose. It will not tell you the tool is magic, and it will not tell you the tool is useless. It tells you where the tool earns its place and where it quietly hollows you out — and it asks you to do the harder thing on purpose.

trust --why

I have spent twenty-plus years operating enterprise IT under real production load — the kind where the outage is yours and the explanation is yours too. This is one operator's publication, not a brand.

45%

of twelfth-graders score below NAEP Basic in math — the lowest since 2005.

source · NAEP 2024 Mathematics Assessment, twelfth-grade

I have seen this movie before. I grew up in a school system that walked away from rigor — show-your-work dropped, mental math dropped, "the calculator will handle the steps in between." That was a lie, the same kind the experts tell about AI today. The mechanism repeats with AI, only worse: a calculator only touched math, where AI touches everything.

ls readership/

The deep dives are written for five operator archetypes, sorted by altitude.

  • Supporter L1–L2 The instance — one ticket, one device, right now
  • Builder L3 The system — root cause, and the playbook that ends it
  • Architect Principal The topology — failure prevention by design
  • Orchestrator EM / Director The team — throughput and the humans
  • Strategist CISO / CTO / VP The capital — ground-truth ROI

Show your work. Read the source. The friction matters.

— Chris Bounds