https://efe-genc-portfolio.vercel.app/writing/ Essays by Efe Genc on building production systems — proactive AI, autonomous pipelines, ownership boundaries, and the things worth not shipping. en-gb Sun, 02 Aug 2026 12:00:00 GMT https://efe-genc-portfolio.vercel.app/writing/ninety-five-percent-less-water/ Vertical farming's sustainability promises, 95% less water, zero pesticide, eight times the yield, are controlled rather than grown. What a TÜBİTAK-funded project's committed targets taught me about where the climate value of agtech lives: in the software layer holding the set points. Sun, 02 Aug 2026 12:00:00 GMT https://efe-genc-portfolio.vercel.app/writing/ninety-five-percent-less-water/ https://efe-genc-portfolio.vercel.app/writing/the-controller-you-can-read/ Inside a sealed vertical farm with no logged history, I chose fuzzy logic over a learned model, because a control decision you cannot explain is a crop you cannot diagnose. When rules beat models, why traceability was the deciding property, and the condition under which the choice expires. Sun, 02 Aug 2026 12:00:00 GMT https://efe-genc-portfolio.vercel.app/writing/the-controller-you-can-read/ https://efe-genc-portfolio.vercel.app/writing/a-token-budget-is-a-product-decision/ Every model call has a fixed budget and there is always more relevant material than fits. What a system leaves out is a product choice, so the five-layer allocator I built treats it as a declared policy per use case instead of the tail of a slice. Thu, 30 Jul 2026 12:00:00 GMT https://efe-genc-portfolio.vercel.app/writing/a-token-budget-is-a-product-decision/ https://efe-genc-portfolio.vercel.app/writing/an-interview-inside-the-browser/ Real-time video, an embedded code editor and proctoring, running together in one tab while someone is being assessed for a job. Three subsystems that each assume they own the main thread, and the loading decision I had backwards to begin with. Thu, 30 Jul 2026 12:00:00 GMT https://efe-genc-portfolio.vercel.app/writing/an-interview-inside-the-browser/ https://efe-genc-portfolio.vercel.app/writing/default-deny-rights-documentary-pipeline/ A pipeline that publishes without a human in the loop has to make rights a condition of entry, not a review step. The failure mode you choose is the design decision. The check you write is only the implementation. Thu, 30 Jul 2026 12:00:00 GMT https://efe-genc-portfolio.vercel.app/writing/default-deny-rights-documentary-pipeline/ https://efe-genc-portfolio.vercel.app/writing/knowing-when-not-to-speak/ A proactive assistant pays far more for a bad interruption than it earns from a good one, so suppression deserves to be architecture rather than a threshold. What I built instead: one delivery gate, twelve ordered checks, and every rejection logged with its reason. Thu, 30 Jul 2026 12:00:00 GMT https://efe-genc-portfolio.vercel.app/writing/knowing-when-not-to-speak/ https://efe-genc-portfolio.vercel.app/writing/one-outbox-record-per-notification/ Two independent producers for push and in-app notifications will eventually disagree, and the disagreement is a category of bug rather than an instance of one. Making them the same record made divergence structurally impossible, and it cost me something I think was worth paying. Thu, 30 Jul 2026 12:00:00 GMT https://efe-genc-portfolio.vercel.app/writing/one-outbox-record-per-notification/ https://efe-genc-portfolio.vercel.app/writing/showing-ownership-private-repositories/ Every repository I have shipped production code into belongs to somebody else. This is the method I use to make ownership claims that can be judged rather than believed, and the claims it forces me to make smaller. Thu, 30 Jul 2026 12:00:00 GMT https://efe-genc-portfolio.vercel.app/writing/showing-ownership-private-repositories/ https://efe-genc-portfolio.vercel.app/writing/surviving-your-ai-providers/ A retry loop around a dead provider multiplies delay instead of providing a fallback. One gateway, a capability-to-tier router, a cross-provider chain and a breaker per provider, plus the parts of this design I still count as a cost rather than a solved problem. Thu, 30 Jul 2026 12:00:00 GMT https://efe-genc-portfolio.vercel.app/writing/surviving-your-ai-providers/ https://efe-genc-portfolio.vercel.app/writing/the-feature-i-chose-not-to-ship/ An over-the-air update splash screen that would have covered a narrow gap during app boot, and degraded every background-to-foreground switch instead. Reproducing it on a real device, reading the dependency, and reverting. Thu, 30 Jul 2026 12:00:00 GMT https://efe-genc-portfolio.vercel.app/writing/the-feature-i-chose-not-to-ship/ https://efe-genc-portfolio.vercel.app/writing/the-spec-is-the-fast-path/ The received wisdom is that writing the design down is what you sacrifice for speed. Over one twelve-month tenure the documents and the release cadence grew together: 255 specifications, 207 plans, 17 runbooks and 377 production releases. This is my account of why those two things were never in tension. Thu, 30 Jul 2026 12:00:00 GMT https://efe-genc-portfolio.vercel.app/writing/the-spec-is-the-fast-path/