igergov.com

Writing

Essays on software, infrastructure and smart homes, written for colleagues and students. Some have examples you can play with.

RSS
14 posts

2026

  1. Docker build secret cache: why is the step still CACHED? Why changing a secret does not request a fresh build step. Docker build secret cache, CACHEBUST, --no-cache-filter and verification limits. DevOpsDockerSecurity 3 min · BG DE EN ES
  2. Can you retry a POST request after a timeout? Can you retry a POST after a timeout? Unknown outcomes, idempotency and the limits of Stripe API v1 keys, with retry timing kept separate. APIHTTPDevOps 3 min · BG DE EN ES
  3. Learning you can see Why I teach abstract classes with paladins and necromancers, what research says about pictures, attention and games in learning, and which popular "facts" are simply not true. LearningGamificationLumen Academy 8 min · BG DE EN ES
  4. A broken test is good news A red pipeline means the bug was caught early, at a point where fixing it is cheap. DevOpsCI interactive 5 min · BG DE EN ES
  5. One app, ten problems How a small notes app goes through the whole of DevOps when each step solves the problem the previous one created. DevOpsTeaching interactive 7 min · BG DE EN ES
  6. Smart home: the truth we would rather not hear My talk at VarnaConf 2026. Why the smart home is not smart yet, what it is missing and when it should ask the person. Smart homeConferences interactive 7 min · BG DE EN ES
  7. A traffic light that sees the traffic Ten years of algorithms and real deployments of smart traffic lights: what works in a real city and what still only exists in simulation. PublicationSmart cityReinforcement learning 3 min · BG DE EN ES
  8. Smart home: the decisions you should not underestimate My talk at BurgasConf 2026. How a single temperature sensor gradually becomes a system, and which decisions along the way carry the most weight. Smart homeConferences interactive 6 min · BG DE EN ES
  9. Kubernetes without the weight We started with a Kubernetes emulator for the DevOps course. It turned out to be easier to build our own small version. KubernetesRoyak interactive 7 min · BG DE EN ES
  10. The agent writes the test. You own it. Agents check the same thing dozens of times and burn tokens, or they write throwaway tests nobody keeps. How we got to Proba, step by step. QAProba interactive 7 min · BG DE EN ES
  11. Can a language model help bring the power back after an attack How large language models can help restore critical infrastructure after a cyberattack, and how far it is safe to trust them today. PublicationCritical infrastructureLanguage models 3 min · BG DE EN ES
  12. Who wins the auction depends on what you measure The idea came from a cloud prototype where you offer the capacity you don't use and buy what you need on top at the best price. We tested its market with agents that learn on their own. ResearchReinforcement learning interactive 8 min · BG DE EN ES
  13. What really helps when the city stops A systematic review of how artificial intelligence helps restore energy, water, transport and hospitals, and where the evidence still falls short. PublicationCritical infrastructureArtificial intelligence 3 min · BG DE EN ES

2025

  1. How to tell whether a model has learned from your data A hidden watermark in the fine-tuning data of a language model, and an honest measure of how far you can rely on it. PublicationLanguage modelsSecurity 3 min · BG DE EN ES