Data & AI consultant, agentic entrepreneur, and founder building practical AI products.https://achrafez.com/https://achrafez.com/posts/2026/one-prompt-thirty-two-calls-or-seven/https://achrafez.com/posts/2026/one-prompt-thirty-two-calls-or-seven/Batching an LLM checklist is a blast-radius decision before it is a performance one: a seeded simulation of three call shapes, and why a call count is arithmetic while a speed-up is a measurement.Sat, 19 Sep 2026 00:00:00 GMThttps://achrafez.com/posts/2026/perfect-precision-zero-recall/https://achrafez.com/posts/2026/perfect-precision-zero-recall/A document check cleared every genuine file and every forgery in a labelled set; a seeded synthetic run shows why specificity on genuine inputs is free and recall is the first number to ask for.Sat, 19 Sep 2026 00:00:00 GMThttps://achrafez.com/posts/2026/the-constant-nobody-could-re-measure/https://achrafez.com/posts/2026/the-constant-nobody-could-re-measure/A magic number survives because there is nothing to re-measure it with. The calibration data is often already in a log you throw away, and the same data sets the noise floor that makes the result readable.Sat, 19 Sep 2026 00:00:00 GMThttps://achrafez.com/posts/2026/the-forecast-metric-that-hides-the-failure/https://achrafez.com/posts/2026/the-forecast-metric-that-hides-the-failure/A volume weighted error of 7.6 percent and a net bias of 0.2 percent can describe a forecast that loses to a seasonal naive on 77 of 116 slow moving series. A seeded panel shows what to measure instead.Sat, 19 Sep 2026 00:00:00 GMThttps://achrafez.com/posts/2026/from-brief-to-rendered-animation/https://achrafez.com/posts/2026/from-brief-to-rendered-animation/A failed scene edit shows why reliable AI video systems need approved scripts, timed artifacts, deterministic fallbacks, and media read-back.Sat, 08 Aug 2026 18:00:00 GMThttps://achrafez.com/posts/2026/calendar-features-for-tabpfn-time-series/https://achrafez.com/posts/2026/calendar-features-for-tabpfn-time-series/TabPFN-TS turns forecasting into a table-plus-features problem. A 24-hour edge case shows why geometry, checkpoint compatibility, and release evidence are separate gates.Tue, 09 Jun 2026 00:00:00 GMThttps://achrafez.com/posts/2026/how-far-should-you-tune-lightgbm/https://achrafez.com/posts/2026/how-far-should-you-tune-lightgbm/LightGBM tuning can win, but only a deployment-faithful baseline and nested validation can show whether the last fraction of a point pays for its search.Sun, 17 May 2026 00:00:00 GMThttps://achrafez.com/posts/2026/tabular-foundation-models-worth-the-hype/https://achrafez.com/posts/2026/tabular-foundation-models-worth-the-hype/Tabular foundation models are a real shift, but the durable bet is about information sets, task breadth, and synthetic priors - not easy benchmark wins.Wed, 29 Apr 2026 00:00:00 GMThttps://achrafez.com/posts/2026/insurance-ai-needs-evidence-graphs/https://achrafez.com/posts/2026/insurance-ai-needs-evidence-graphs/A synthetic insurance claim shows how hashes, typed facts, contradictions, abstention, and human review make AI recommendations auditable.Sat, 11 Apr 2026 00:00:00 GMThttps://achrafez.com/posts/2026/from-pdfs-to-citable-verdicts/https://achrafez.com/posts/2026/from-pdfs-to-citable-verdicts/A practical design for separating relevance from proof, preserving document lineage, abstaining safely, and evaluating a citable Qualiopi pre-audit.Fri, 27 Feb 2026 00:00:00 GMThttps://achrafez.com/posts/2026/from-human-like-text-to-human-voice/https://achrafez.com/posts/2026/from-human-like-text-to-human-voice/Short AI text can already pass bounded source tests. The harder question is whether a model can sustain one person's situated voice across contexts and time.Tue, 03 Feb 2026 00:00:00 GMThttps://achrafez.com/posts/2026/i-stopped-prompting-claude-code-and-started-engineering-the-loop/https://achrafez.com/posts/2026/i-stopped-prompting-claude-code-and-started-engineering-the-loop/A green test suite missed an integration seam; a frozen mutation experiment shows how repository rules, executable checks, scoped review, and replay state make agentic work more inspectable.Wed, 14 Jan 2026 00:00:00 GMThttps://achrafez.com/posts/2026/the-keyword-that-broke-unequal-length-periodograms/https://achrafez.com/posts/2026/the-keyword-that-broke-unequal-length-periodograms/How Aeon's PeriodogramTransformer pads unequal-length collections, why its list and 3-D branches need parity, and what one case study teaches.Fri, 21 Nov 2025 00:00:00 GMThttps://achrafez.com/posts/2026/the-fourier-smoother-that-halved-the-signal-it-kept/https://achrafez.com/posts/2026/the-fourier-smoother-that-halved-the-signal-it-kept/A bounded Aeon DiscreteFourierApproximation case study shows how conjugate symmetry, passband gain, and cutoff semantics expose concrete Fourier failure modes.Thu, 06 Nov 2025 00:00:00 GMThttps://achrafez.com/posts/2026/your-acf-can-be-right-while-its-confidence-interval-is-wrong/https://achrafez.com/posts/2026/your-acf-can-be-right-while-its-confidence-interval-is-wrong/A missing-data bug in statsmodels left ACF estimates unchanged while confidence intervals and Ljung-Box statistics used the wrong sample size.Tue, 14 Oct 2025 00:00:00 GMT