Link Dump #244

To read or not to read - what a silly question:

1. Software Architecture
   1. Ghost Ownership: You Own the Capability, Not the Decision #PickOfTheWeek
      Your engineering teams might own the microservices, the deployment pipelines, and the API gateways for your latest AI features, but do they actually own the business decisions happening inside them? In this sharp critique, the author introduces the concept of "Ghost Ownership" - a hidden architectural trap where teams manage the operational capability while an upstream, non-deterministic LLM quietly rewrites their domain's core logic.
   2. Domain Types: The architecture your subdomain deserves #PickOfTheWeek
      Applying the same architecture everywhere often leads to overengineering, underengineering, or both. Discover how domain types can help you choose the right level of architectural complexity.
      
2. Software Development
   1. When code costs nothing to produce, how do you review it all?
      This sharp take exposes the dangerous illusion of relying blindly on automated test loops that can only verify what someone thought to encode. Find out why attempting to accelerate software velocity by abandoning code reading forces human engineers to step up a level of abstraction - shifting their daily roles from manually inspecting line-by-line changes to actively designing, tuning, and auditing the very systems that inspect the changes.
   2. What Mechanical Engineers Know About Friction, Failure And AI #PickOfTheWeek
      Lessons from lubricated systems for designing and operating AI-enabled work without losing expertise, feedback and control.
   3. double, BigDecimal, or Fixed-Point?
      This analysis explores the architectural sweet spots between double, BigDecimal, and fixed-point alternatives. Learn how to audit your computational context based on precision needs, rounding boundaries, and performance budgets, and find out exactly when an optimized primitive long or double is structurally superior to a heavy object wrapper.
   4. The Intent Debt #PickOfTheWeek 
      An AI agent can write your tests, rewrite your legacy modules, and help you understand an unfamiliar framework, but it can never tell you what you originally meant to build. Find out why neglecting your project’s written constraints transforms your automated workspace into an engine for system drift, and why externalizing your architectural decisions is the only remaining way to stay in control of your software.
   5. Navigating today’s AI token crisis
      This analysis explores why corporate heavyweights like Uber and Microsoft are radically scaling back cloud AI access after burning through entire yearly budgets in a matter of weeks. Discover why agentic workflows break traditional cloud infrastructure monitoring by appearing perfectly healthy while quietly chewing through thousands of metered API dollars, and learn how to implement strict token-based circuit breakers before a runaway loop tanks your delivery margins. 
      
3. Leadership
   1. Your Company Doesn’t Just Need a Defensible Strategy – It Needs One that Can Adapt
      Many executive teams design their corporate strategies like fortresses, building high walls around their current product lines and calling it defensible. Discover why defining your organization by your current physical processes rather than the underlying problems you solve sets you up for sudden disruption, and learn how to shift your executive perspective so your business model can evolve fluidly when the marketplace changes beneath your feet. 
   2. Dilbert Principle #PickOfTheWeek
      In many large software organizations, a bizarre promotional pattern emerges: the developers who struggle most with code are suddenly elevated to management positions. Reflecting on the systemic ironies of software engineering hierarchies, this sharp take exposes the danger of using promotions to hide employee performance issues.