Link Dump #255

Let's talk about... or let's read something to talk about it later.

1. Software Architecture
   1. Infrastructure efficiency: How a classic math problem helped reduced our capacity buffers
      Learn how statistical modeling and software engineering can be combined to manage the efficiency-reliability balance inside a private PaaS environment.
   2. Residuality Theory: A Rebellious Take on Building Systems That Actually Survive #PickOfTheWeek
      Barry O'Reilly developed the residuality theory, a rebellious approach to software architecture that focuses on designing systems that withstand unexpected stresses and changes. This isn’t just philosophy. It is based on actual complexity science. And this is definitelly approach worth knowing.
      
2. Software Development
   1. Measuring AI’s Impact on Delivery Speed
      If your organization calculates AI ROI based purely on survey impressions or pull request volume, your metrics are misleading. Reflecting on software delivery engineering, James Shore demonstrates why comparing teams against their own historical estimation curves is the only way to isolate real speed gains. In this summary, the analysis highlights how randomized task assignment provides clear data on whether AI coding tools actually save time
   2. The quantum countdown
      Post-Quantum Cryptography is not a future IT initiative. It is an active risk management, governance and compliance requirement. The structural vulnerabilities introduced by Harvest Now, Decrypt Later operations mean that organizations delaying their PQC transition accumulate unmanaged liability today.
   3. Context Engineering: The Missing Piece in Agentic Systems #PickOfTheWeek
      Context engineering is becoming essential as AI agents take on more software development work. An agent can plan, code, test, investigate incidents, trigger CI, and help deploy software. But none of that matters if it is operating without the right information.
   4. Measuring AI’s Unintended Consequences #PickOfTheWeek
      AI is great for making teams faster. But is it sustainable? What about maintainability, and burnout, and lock-in? In this part, we’re looking at how to measure and model the unintended consequences of using AI.
   5. Why shift left is dead?
      Traditional shift-left security assumes code already exists, but AI introduces risk earlier through prompts, agents, packages, permissions and access to sensitive information. Organizations need to evolve from the conventional SDLC toward an Agentic Development Lifecycle that governs AI tools, models, data access and agent activity throughout development.
   6. Defend your downtime to improve agentic output #PickOfTheWeek
      In the article, the author challenges the urge to constantly fill the waiting time created by AI coding agents. As agentic workflows handle execution tasks, developers often feel pressured to parallelize work by context-switching into next tasks. The author argues that actively defending mental downtime is essential for consolidating complex problem context and generating creative architectural breakthroughs.
      
3. Testing
   1. Stop Chasing 100% Test Coverage
      Striving for complete test coverage forces teams to spend hours asserting trivial logic while missing critical system failures. In this article, the author explains how focusing automated testing on high-impact areas maximizes deployment confidence without slowing down release cadence.
      
4. Agile
   1. AI makes Agile more alive #PickOfTheWeek
      Speed alone doesn’t solve problems such as unclear priorities, slow decision-making, weak feedback loops or poor collaboration. AI can amplify both agility and dysfunction. That makes agility more valuable, but it also changes what teams need to optimize for: not simply how much software they can produce, but how quickly they can turn an idea into an experiment, learn from it and decide what to do next.
      
5. Leadership
   1. What They Don’t See
      Frontline employees operate in short timespans where the immediate result of skipping an inspection seems harmless. In this article, the author explains how a time lag of three months between a defective part's assembly and a customer's complaint obscures the cause-and-effect relationship for workers, leaving them seeing only isolated reprimands rather than the full operational chain.
      
6. Fun
   1. Deep work