https://theonlyengineer.com Notes from building generative-AI systems that survive real users: agents in production, evaluating LLM output, and the architecture and cost underneath. en nizar@theonlyengineer.com (Nizar) Thu, 27 Aug 2026 00:00:00 GMT https://theonlyengineer.com/blog/context-is-the-hard-part https://theonlyengineer.com/blog/context-is-the-hard-part Thu, 27 Aug 2026 00:00:00 GMT Why the same AI feature works at one company and fails at another: canonical definitions, current signals, and quality enforcement — the layers nobody budgets for. nizar@theonlyengineer.com (Nizar) llm-architecture-and-cost context architecture production https://theonlyengineer.com/blog/the-most-expensive-call-is-the-one-that-fails https://theonlyengineer.com/blog/the-most-expensive-call-is-the-one-that-fails Thu, 27 Aug 2026 00:00:00 GMT In a step-billed system your costliest interactions are the ones the model works hardest on and still can't finish. The effort paradox, and the four controls that bound it. nizar@theonlyengineer.com (Nizar) llm-architecture-and-cost cost agents production https://theonlyengineer.com/blog/what-an-llm-feature-costs-to-run https://theonlyengineer.com/blog/what-an-llm-feature-costs-to-run Thu, 27 Aug 2026 00:00:00 GMT Cost is steps times tokens times retries, the failures dominate, and the pilot will not show you. The four buckets to model and the levers that actually move the bill. nizar@theonlyengineer.com (Nizar) llm-architecture-and-cost cost architecture production https://theonlyengineer.com/blog/where-ai-belongs-in-your-architecture https://theonlyengineer.com/blog/where-ai-belongs-in-your-architecture Thu, 27 Aug 2026 00:00:00 GMT Models at the edges, deterministic code in the core. The patterns that don't need a model, the ones where it earns its keep, and the question to ask before adding it. nizar@theonlyengineer.com (Nizar) llm-architecture-and-cost architecture llm cost https://theonlyengineer.com/blog/observability-for-llm-features https://theonlyengineer.com/blog/observability-for-llm-features Thu, 27 Aug 2026 00:00:00 GMT What to record on every model call, why the rendered prompt is the field that matters most, and why instrumentation has to exist before the first incident. nizar@theonlyengineer.com (Nizar) trustworthy-llm-output observability production llm https://theonlyengineer.com/blog/test-one-prompt-a-hundred-times https://theonlyengineer.com/blog/test-one-prompt-a-hundred-times Thu, 27 Aug 2026 00:00:00 GMT Single-run prompt testing hides variance, tail latency, token spread and intermittent misbehaviour. How to build the batch harness that surfaces them. nizar@theonlyengineer.com (Nizar) trustworthy-llm-output evals testing llm https://theonlyengineer.com/blog/constrain-the-model-dont-review-it https://theonlyengineer.com/blog/constrain-the-model-dont-review-it Thu, 27 Aug 2026 00:00:00 GMT Inconsistent generated output is a constraint problem, not a capability problem. Schemas and closed vocabularies beat review queues, and here is why. nizar@theonlyengineer.com (Nizar) trustworthy-llm-output llm architecture testing https://theonlyengineer.com/blog/how-to-evaluate-an-llm-feature https://theonlyengineer.com/blog/how-to-evaluate-an-llm-feature Thu, 27 Aug 2026 00:00:00 GMT Why normal testing doesn't work on a probability distribution, what goes in an adversarial case set, and how to score it without fooling yourself. nizar@theonlyengineer.com (Nizar) trustworthy-llm-output evals llm testing https://theonlyengineer.com/blog/an-agent-is-a-state-machine https://theonlyengineer.com/blog/an-agent-is-a-state-machine Thu, 27 Aug 2026 00:00:00 GMT Orchestration is not reasoning. Which decisions to encode in code, which to leave to the model, and what an explicit generation graph buys you. nizar@theonlyengineer.com (Nizar) agents-in-production agents architecture https://theonlyengineer.com/blog/long-running-agents-belong-in-a-queue https://theonlyengineer.com/blog/long-running-agents-belong-in-a-queue Thu, 27 Aug 2026 00:00:00 GMT Why multi-step generation cannot live in an HTTP request, what the queue-plus-stream architecture looks like, and the worker settings that decide whether it degrades gracefully. nizar@theonlyengineer.com (Nizar) agents-in-production agents architecture production https://theonlyengineer.com/blog/narrow-agents-beat-smart-ones https://theonlyengineer.com/blog/narrow-agents-beat-smart-ones Thu, 27 Aug 2026 00:00:00 GMT A boring agent that produces a useful artifact every time beats a clever one that mostly does. How to scope a task narrowly enough to be validated, tested and trusted. nizar@theonlyengineer.com (Nizar) agents-in-production agents production https://theonlyengineer.com/blog/what-it-takes-to-ship-an-llm-agent https://theonlyengineer.com/blog/what-it-takes-to-ship-an-llm-agent Thu, 27 Aug 2026 00:00:00 GMT The checklist that separates a working demo from a system you can leave running: scope, control flow, queues, evaluation, constraints, cost and visibility. nizar@theonlyengineer.com (Nizar) agents-in-production agents production llm https://theonlyengineer.com/blog/modernizing-on-prem-infrastructure https://theonlyengineer.com/blog/modernizing-on-prem-infrastructure Thu, 13 Aug 2026 00:00:00 GMT Bringing cloud-native tools into your on-prem infrastructure for automation and resiliency. nizar@theonlyengineer.com (Nizar) hpc cloudstack virtualization https://theonlyengineer.com/blog/automating-hpc-infrastructure https://theonlyengineer.com/blog/automating-hpc-infrastructure Thu, 13 Aug 2026 00:00:00 GMT Using Apache CloudStack to manage cloud resources in your on-prem infrastructure. nizar@theonlyengineer.com (Nizar) hpc cloudstack