https://www.jasonbelnick.com/writing/ Practical notes on AI adoption and local models. en-us Tue, 07 Jul 2026 00:00:00 +0000 https://www.jasonbelnick.com/writing/the-mac-mini-that-never-swaps/ https://www.jasonbelnick.com/writing/the-mac-mini-that-never-swaps/ Tue, 07 Jul 2026 00:00:00 +0000 Why the unattended jobs in my local-AI setup run on a dedicated always-on Mac mini instead of my 64GB Mac Studio, serving one small model that never gets evicted. https://www.jasonbelnick.com/writing/i-built-a-loop-that-learns-from-my-editing-mistakes/ https://www.jasonbelnick.com/writing/i-built-a-loop-that-learns-from-my-editing-mistakes/ Sat, 04 Jul 2026 00:00:00 +0000 Part 2 of the local writer story: a self-improvement loop whose first act was rejecting its own idea, and why that rejection is the feature https://www.jasonbelnick.com/writing/i-trained-a-local-model-to-write-like-claude/ https://www.jasonbelnick.com/writing/i-trained-a-local-model-to-write-like-claude/ Fri, 03 Jul 2026 00:00:00 +0000 I trained a 9B to write like Claude, benched it blind, and the ending flipped: the bigger model with the distilled skill took the seat https://www.jasonbelnick.com/writing/the-judge-that-runs-on-my-desk/ https://www.jasonbelnick.com/writing/the-judge-that-runs-on-my-desk/ Thu, 02 Jul 2026 00:00:00 +0000 I benched local LLMs as judges, watched a bare prompt invert the winner, and swapped a cloud judge for a local one once the rubric was armed https://www.jasonbelnick.com/writing/the-4-30-am-shift/ https://www.jasonbelnick.com/writing/the-4-30-am-shift/ Thu, 02 Jul 2026 00:00:00 +0000 A stack of cron jobs on local LLMs runs my knowledge base overnight. Two failures I watched in the logs changed how I build every local pipeline. https://www.jasonbelnick.com/writing/a-wiki-my-models-maintain/ https://www.jasonbelnick.com/writing/a-wiki-my-models-maintain/ Thu, 02 Jul 2026 00:00:00 +0000 My local LLMs maintain a knowledge base: they capture sources, summarize them, and commit the output daily. The RAG engine I built on top was the wrong idea https://www.jasonbelnick.com/writing/one-big-model-at-a-time/ https://www.jasonbelnick.com/writing/one-big-model-at-a-time/ Thu, 02 Jul 2026 00:00:00 +0000 How I schedule large local models on a 64GB Mac Studio with a router and a swap tool instead of juggling memory by hand every day https://www.jasonbelnick.com/writing/we-delete-more-models-than-we-keep/ https://www.jasonbelnick.com/writing/we-delete-more-models-than-we-keep/ Thu, 02 Jul 2026 00:00:00 +0000 How a nightly coding benchmark and a strict parity rule turned a season of hyped local models into deleted weights and a stack I trust https://www.jasonbelnick.com/writing/ai-adoption-missing-90-percent/ https://www.jasonbelnick.com/writing/ai-adoption-missing-90-percent/ Wed, 01 Jul 2026 00:00:00 +0000 73% of surveyed enterprises use AI but only 10% run on it. Missing ownership, sign-off paths, and measured workflows create the gap; the models are fine https://www.jasonbelnick.com/writing/the-open-weight-frontier-mid-2026/ https://www.jasonbelnick.com/writing/the-open-weight-frontier-mid-2026/ Wed, 01 Jul 2026 00:00:00 +0000 In one June window, open weights reached the coding frontier. What I'd build on, what I'd admire from a distance, and which numbers to trust https://www.jasonbelnick.com/writing/local-llms-on-a-64gb-mac-studio/ https://www.jasonbelnick.com/writing/local-llms-on-a-64gb-mac-studio/ Wed, 01 Jul 2026 00:00:00 +0000 The awkward middle tier of local AI in mid-2026, the capacity and bandwidth math that governs it, and a 75 percent verdict on what a 64GB Mac Studio actually delivers https://www.jasonbelnick.com/writing/the-week-the-cloud-blinked/ https://www.jasonbelnick.com/writing/the-week-the-cloud-blinked/ Wed, 01 Jul 2026 00:00:00 +0000 Fable 5 vanished for 18 days under an export-control order. Frontier model access is now a dependency risk you have to engineer against https://www.jasonbelnick.com/writing/routing-work-between-local-and-cloud/ https://www.jasonbelnick.com/writing/routing-work-between-local-and-cloud/ Wed, 01 Jul 2026 00:00:00 +0000 Model rankings churn monthly; a routing policy survives. Classify LLM workloads by risk, privacy, latency, and volume, and keep the decision log. https://www.jasonbelnick.com/writing/self-refinement-that-knows-when-to-stop/ https://www.jasonbelnick.com/writing/self-refinement-that-knows-when-to-stop/ Wed, 01 Jul 2026 00:00:00 +0000 Nous Research's Autoreason finds naive self-refinement makes output worse, then fixes it with a blind tournament where doing nothing can win