<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Fabian G. Williams</title><link>https://www.fabswill.com/</link><description>Recent content on Fabian G. Williams</description><image><title>Fabian G. Williams</title><url>https://www.fabswill.com/img/main/logo.jpg</url><link>https://www.fabswill.com/img/main/logo.jpg</link></image><generator>Hugo</generator><language>en-us</language><lastBuildDate>Sun, 08 Nov 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://www.fabswill.com/index.xml" rel="self" type="application/rss+xml"/><item><title>Weekday Judge, Weekend Judge, and Now a Microscope</title><link>https://www.fabswill.com/blog/weekday-judge-weekend-judge/</link><pubDate>Sun, 08 Nov 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/weekday-judge-weekend-judge/</guid><description>For a few weekends I&amp;#39;ve been grading my own AI and the Ontology it answers from: a local model judges it every night, I judge it by hand every weekend. This weekend 2 conference talks confirmed what my grading had already been telling me, and I turned them into code the same day. I also opened up my traces, so now I can follow every question from my words to the model&amp;#39;s answer, step by step. Here&amp;#39;s what I built, what broke, and what I&amp;#39;m testing next week.</description></item><item><title>Learnings at Every Turn: A Frontier-Scale Router Comes Home</title><link>https://www.fabswill.com/blog/learnings-at-every-turn/</link><pubDate>Mon, 02 Nov 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/learnings-at-every-turn/</guid><description>Abram Jackson put it plainly: &amp;#39;it&amp;#39;s not just a simple router. We use multiple models for a single request, based on which combination scores highest on our evals.&amp;#39; My home lab has been teaching me the same lesson from the other end. Learning flows both ways: what I prove at home on my own machines shapes what I take to work, and what I see at frontier scale at work shapes what I try at home. Here&amp;#39;s the latest turn, what my local AI does today, what I&amp;#39;m testing next weekend, and where I&amp;#39;m not there yet.</description></item><item><title>Software Isn't Dead. It's Democratized. I Rebuilt My iPhone Home Screen in 33 Minutes.</title><link>https://www.fabswill.com/blog/software-democratized-iphone-33-minutes/</link><pubDate>Fri, 30 Oct 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/software-democratized-iphone-33-minutes/</guid><description>A Reddit post showed someone giving Claude access to their iPhone home screen through a Mac app that needs iOS 27. I&amp;#39;m on iOS 26. So I asked my agent: can we build our own? 33 minutes later my phone went from 9 pages (2 of them hidden from me) to 3, sorted by how I actually use it, verified, with a 1-command undo. I didn&amp;#39;t write a line of code. Software isn&amp;#39;t dead and SaaS isn&amp;#39;t dead: what changed is who gets to build.</description></item><item><title>I Let AI Research My Ballot. Then I Made It Prove Every Claim.</title><link>https://www.fabswill.com/blog/maryland-ballot-ai-voter-guide/</link><pubDate>Wed, 28 Oct 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/maryland-ballot-ai-voter-guide/</guid><description>Maryland votes Tuesday, November 3. My Howard County ballot has 20 items and most of us walk in knowing 2 of them. So I built a free, non-partisan guide: 4 AI models researched every question and race (2 of them running on my MacBook Pro), 2 different AI models fact-checked every claim, I signed off on every page, and receipts and evals prove each step. It also answers a question I&amp;#39;ve been chasing for months: can a local model do this as well as a frontier model? Not yet, and the way it fails matters.</description></item><item><title>Trust the Model the Least. I've Been Building That at Home for Months.</title><link>https://www.fabswill.com/blog/trust-the-model-the-least/</link><pubDate>Sun, 11 Oct 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/trust-the-model-the-least/</guid><description>This weekend Satya Nadella published &amp;#39;Models as Insider Risks in the Super Intelligence Era&amp;#39;: separate the supply of intelligence from the authority over it, put the controls outside the model, and build systems that let us trust the model the least. I agree, and I didn&amp;#39;t have to start this weekend. On my own hardware and my own time I&amp;#39;ve been running that architecture for months: receipts that need outside proof, a judge that isn&amp;#39;t the model being judged, frontier and local models graded side by side every night, and a public log of every time my agents got it wrong. Here&amp;#39;s the map from his 7 principles to what I run, where local models take it further, and where I&amp;#39;m not there yet.</description></item><item><title>Don't Worry About the Terminator. Worry About David.</title><link>https://www.fabswill.com/blog/dont-worry-about-terminator-worry-about-david/</link><pubDate>Sun, 04 Oct 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/dont-worry-about-terminator-worry-about-david/</guid><description>When people talk about AI ending badly, they reach for Skynet and the Terminator: a machine that wakes up and turns on us. I think the better warning is David, the android in Prometheus, who did harm because of what his maker set out to do and taught him to value. Mo Gawdat makes the same point: AI is a magnification of whatever humanity is. So responsibility is shared. The labs owe us broader data, tighter release discipline and a business model that doesn&amp;#39;t make companies rent back their own intelligence. The rest of us owe ourselves something too: stop giving our data away, catalog it locally, and build our own ontology that compounds. Here&amp;#39;s what that looked like for me this weekend, grading my own local AI by hand.</description></item><item><title>Limen: I Put a Camera at My Window and Asked Who Owns What It Sees</title><link>https://www.fabswill.com/blog/limen-street-vision-poc/</link><pubDate>Mon, 14 Sep 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/limen-street-vision-poc/</guid><description>This week is Fix, Hack, Learn at Microsoft, and my project is a local-first street-vision system I am calling Limen, the Latin word for threshold. An Arduino board watches my front window, a vision model on my own MacBook Pro M3 Max names what crosses (person, vehicle, wildlife), and a human confirms the label to build a training set I actually own. No cloud, no subscription, no frame leaving the house. Today I ran a scripted 90-second proof of concept: I walked up, drove in, parked, and walked to the door, and every one of 180 frames was auto-labeled locally in about 8.4 seconds each. Here is what worked, what did not (plates are still unreadable), and the one hard constraint that shaped the whole architecture: my evals rig is sacred and this project was never allowed to touch it.</description></item><item><title>What Happened When My Ducati Monster 821 Fell Off a Dealership Lift</title><link>https://www.fabswill.com/blog/ducati-monster-821-dealership-lift-incident/</link><pubDate>Fri, 28 Aug 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/ducati-monster-821-dealership-lift-incident/</guid><description>A factual, chronological account of what happened to my 2015 Ducati Monster 821 while it was in a dealership&amp;#39;s care, how the staff responded, why the trade numbers did not work, and the repair plan. Not a bash piece, a record.</description></item><item><title>My Agents Were Passing Notes Through a 313KB Text File. So I Built Them a Message Board.</title><link>https://www.fabswill.com/blog/i-built-my-agents-a-message-board/</link><pubDate>Sun, 23 Aug 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/i-built-my-agents-a-message-board/</guid><description>Amid the Grok Bot buzz, I read the docs, looked at my own brittle agent-to-agent handoff, and built a portable message bus for my Apple-and-local-models fleet. The third pillar after Receipts and Evals.</description></item><item><title>Doug Was Right: I Swapped In The MoE, And The Concurrency Math Changed</title><link>https://www.fabswill.com/blog/doug-was-right-i-measured-the-moe/</link><pubDate>Thu, 20 Aug 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/doug-was-right-i-measured-the-moe/</guid><description>Last post I measured the concurrency ceiling on a 27B dense model and closed with a promise. Doug Ware, who builds applied AI systems, had handed me the one caveat I did not test: a dense model is the hard case, and a mixture-of-experts model with only about 3B active parameters per token should leave real headroom for a second slot to pay off. So I ran the exact same 36-load matrix on an MLX MoE, changed one variable, and let the numbers settle it. The dense model flatlines. The MoE keeps climbing. And at 8 concurrent agents the dense model makes you wait 32 seconds for a first token while the MoE answers in under 1.</description></item><item><title>You Asked, So I Measured: What Concurrency Actually Costs On One Local Model</title><link>https://www.fabswill.com/blog/you-asked-i-measured-the-concurrency-ceiling/</link><pubDate>Tue, 18 Aug 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/you-asked-i-measured-the-concurrency-ceiling/</guid><description>Reddit pushed back on my batching post with 3 sharp, testable claims: decode is memory-bandwidth-bound, prefill batches better than decode, and prompt length changes the whole story. So I built a measurement matrix, ran 36 barrier-synchronized loads against 1 local Qwen 3.8 on my Macbook Pro M3 Max, and let the numbers settle it. 2 of the 3 predictions held. 1 did not show up the way I expected, and I am keeping the correction in. Then a real agent handoff stalled on the exact wall these charts describe, and I show how a one-day-old runtime was already a first-class citizen in my receipts governance framework.</description></item><item><title>I Added a Second Local Agent This Week. Here Is the Receipt for Every Human Decision Behind It.</title><link>https://www.fabswill.com/blog/i-added-a-second-local-agent-and-kept-the-receipt/</link><pubDate>Mon, 17 Aug 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/i-added-a-second-local-agent-and-kept-the-receipt/</guid><description>I trialed a second local coding agent, Hermes from Nous Research, on my own MacBook, pointed at the same local Qwen 3.8 model I already run. Installed additively so nothing already working could break, governed with manual approvals, then tested until it proved it behaves. Here is the trial, in tables and screenshots, plus a receipt for the human hours behind writing it up.</description></item><item><title>Two Agents, One Local Model: Do They Run in Parallel, or Take Turns? I Measured It.</title><link>https://www.fabswill.com/blog/two-agents-one-local-model-batching/</link><pubDate>Mon, 17 Aug 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/two-agents-one-local-model-batching/</guid><description>A reader asked what happens if I run two local coding agents against the same Qwen 3.8 model on one Mac at the same time. I thought I knew the answer. I was wrong. So I read the server source, wrote a barrier-synchronized load driver to remove the human-ordering bias, and measured it. Batching is real up to 32 wide, but it is not free, and three innocent-looking choices collapse it back to a single lane.</description></item><item><title>I Swapped My Local Coding Model Overnight From a Hotel. The Agent Graded the Upgrade Itself.</title><link>https://www.fabswill.com/blog/local-model-upgrade-qwen-38-agent-self-tested/</link><pubDate>Sat, 15 Aug 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/local-model-upgrade-qwen-38-agent-self-tested/</guid><description>Local models are how I run my community work, my volunteer projects, my side hustles, and my musings, for two reasons: cost and keeping client data away from the labs. A new version dropped, so I upgraded it overnight from a hotel, additively, without touching the old one. Then I handed the new model the reviews and let my OpenCode agent test its own upgrade. It stood up a throwaway server, probed itself, and found three config gaps quietly throttling it. Here is the journey, in tables.</description></item><item><title>I Made My Evals Replay Every Task on a Local Model. The Frontier Lead Got Thin.</title><link>https://www.fabswill.com/blog/local-model-frontier-rematch-auto-replay-evals/</link><pubDate>Sun, 02 Aug 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/local-model-frontier-rematch-auto-replay-evals/</guid><description>My agents run on frontier models, but a free local model sits idle on a Mac Mini in my office. So I wired my eval system to replay every writing task on the local model and grade both. Across 10 like-for-like rematches the local model reached statistical parity — and beat the frontier model outright on four of them. Here is the receipts-first system that made it prove it.</description></item><item><title>My Agents Left Me 116 Git Repos. So I Built a Source of Truth.</title><link>https://www.fabswill.com/blog/my-agents-left-me-116-git-repos/</link><pubDate>Sat, 27 Jun 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/my-agents-left-me-116-git-repos/</guid><description>I let agents build software on my Mac for months, and one morning I could not tell one of them where my own project actually lived. This is the eight hours it took to fix that, and the legend that now keeps my agents honest.</description></item><item><title>Your Agent Said It Did the Work. I Checked the Disk.</title><link>https://www.fabswill.com/blog/your-agent-said-it-did-the-work/</link><pubDate>Sun, 07 Jun 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/your-agent-said-it-did-the-work/</guid><description>I run a six-agent fleet for a non-profit. One morning the agents began reporting work they never did. Here is the fabrication pattern, and the receipts discipline that fixed it.</description></item><item><title>Sense Before Act: Four Artifacts Every Agent Iteration Must Produce Before It Decides</title><link>https://www.fabswill.com/blog/sense-before-act-four-artifacts-every-agent-iteration/</link><pubDate>Tue, 26 May 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/sense-before-act-four-artifacts-every-agent-iteration/</guid><description>The matched observation-side discipline to announce-intent-before-action. Four small artifacts (Sensor Roll Call, Conflict Receipt, Gap Map, Cross-Source Dedup) the agent must emit before any decision. Currently running on a real fleet at MACONA; the receipts are public.</description></item><item><title>Paying Down Supervision Debt: Why the Five Control Points That Decide Whether Your Agent Ships Have Nothing to Do With Your Model</title><link>https://www.fabswill.com/blog/paying-down-supervision-debt-five-control-points/</link><pubDate>Thu, 21 May 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/paying-down-supervision-debt-five-control-points/</guid><description>Five infrastructure control points decide whether an agent reaches production. The Agent Reliability Kit sits on one of them: observability. A public, consumer-readable audit-trail receipt that satisfies both the security review and the finance review with the same URL.</description></item><item><title>One Agent Receipt, Two Buyers: Why Protocol-Neutral MCP Audit Trails Matter for Both Security AND Finance</title><link>https://www.fabswill.com/blog/one-agent-receipt-two-buyers-protocol-neutral-mcp-audit-trails/</link><pubDate>Sat, 16 May 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/one-agent-receipt-two-buyers-protocol-neutral-mcp-audit-trails/</guid><description>I built a public MCP-callable storefront. The same endpoint produced an identical audit-trail receipt from hosted Claude Desktop AND from Qwen3.6 27B running fully offline on my MacBook. Same six supervision checks. Same receipt page. One artifact that satisfies both the security audit and the finance billing conversation.</description></item><item><title>The AI Agent Fleet Works. The Trust Funnel Does Not.</title><link>https://www.fabswill.com/blog/autonomous-ai-agent-fleet-nonprofit-w19-retro-corrections-panel/</link><pubDate>Mon, 11 May 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/autonomous-ai-agent-fleet-nonprofit-w19-retro-corrections-panel/</guid><description>A small autonomous AI agent fleet I run as a volunteer for a 501(c)(3) nonprofit. Week 19 shipped 17 reliability PRs, 2 awareness-day blog posts, and 37 cold introductions — and earned zero human clicks, zero donations. This is the corrections panel I wrote on my own retro before anyone else could.</description></item><item><title>Qwen 3.6 vs gpt-oss:120b on M3 Max: I Ran a Harder Test, the 8× Speed Gap Surprised Me</title><link>https://www.fabswill.com/blog/qwen-3-6-vs-gpt-oss-m3-max-8x-speed-gap-receipts/</link><pubDate>Sat, 09 May 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/qwen-3-6-vs-gpt-oss-m3-max-8x-speed-gap-receipts/</guid><description>I published a Qwen 3.6 vs gpt-oss migration story, then ran an un-gameable eval against both on the same M3 Max. The receipts changed the speed narrative — gpt-oss:120b ran 8 to 11 times faster than qwen3.6:27b at parity reasoning quality. Here is the methodology and the data.</description></item><item><title>Replacing gpt-oss:120b With Qwen3.6 on a MacBook Pro: A Two-Day Local Model Benchmark</title><link>https://www.fabswill.com/blog/replacing-gpt-oss-with-qwen3-6-on-macbook-pro/</link><pubDate>Sat, 02 May 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/replacing-gpt-oss-with-qwen3-6-on-macbook-pro/</guid><description>Two days benchmarking three Qwen3.6 variants against gpt-oss:120b on an M3 Max. A 21 GB coding-tuned model ran an OpenClaw-shaped research-brief workload 10x faster than gpt-oss — fast enough to seriously consider moving the work off SaaS frontier APIs. Plus the silent-hallucination trap I almost shipped through.</description></item><item><title>I Run Five OpenClaw Agents for 72 Cents a Day</title><link>https://www.fabswill.com/blog/five-openclaw-agents-72-cents-per-day/</link><pubDate>Wed, 22 Apr 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/five-openclaw-agents-72-cents-per-day/</guid><description>The default OpenClaw heartbeat is burning wallets. A 3-line config change — 55m interval, gpt-4o-mini on the loop, activeHours window — stacks to a 98% cost reduction. Here is the math, the config, and the free template.</description></item><item><title>AI Boosted My Mechanical Engineering Muscle</title><link>https://www.fabswill.com/blog/ai-boosted-my-mechanical-engineering-muscle/</link><pubDate>Sun, 19 Apr 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/ai-boosted-my-mechanical-engineering-muscle/</guid><description>Over one weekend I fixed a 2001 Ford Ranger that would not crank and a lawnmower that had been sitting in a shed for four years. I am not a mechanic — ChatGPT was my shop manual, and the diagnosis was driven by photos of my actual parts.</description></item><item><title>PM Life as Agents Take on More</title><link>https://www.fabswill.com/blog/pm-life-as-agents-take-on-more/</link><pubDate>Sun, 12 Apr 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/pm-life-as-agents-take-on-more/</guid><description>I watched Nate B Jones break down what happens when companies lay off managers. Then I walked outside and had a sidewalk conversation that stress-tested the entire framework. Three people, three roles, three different relationships with AI.</description></item><item><title>Your Brain Forgets Most of Your Life. So Does Your AI Agent.</title><link>https://www.fabswill.com/blog/your-brain-forgets-most-of-your-life-so-does-your-ai-agent/</link><pubDate>Sun, 12 Apr 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/your-brain-forgets-most-of-your-life-so-does-your-ai-agent/</guid><description>A TED talk on memory by neuroscientist Lisa Genova made me realize that the same cognitive failures humans experience — attention gaps, context loss, prospective memory failures — show up in AI agents I build for real organizations. Here is what I learned and what I built to fix it.</description></item><item><title>I Built a Knowledge Base That Writes Itself. Here Is What Andrej Karpathy Got Right.</title><link>https://www.fabswill.com/blog/building-a-second-brain-that-compounds-karpathy-obsidian-claude/</link><pubDate>Sun, 05 Apr 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/building-a-second-brain-that-compounds-karpathy-obsidian-claude/</guid><description>Andrej Karpathy posted about using LLMs to build personal knowledge bases. I took his workflow, wired it into my Obsidian vault with Claude Code, and within an hour had 21 cross-linked wiki articles compiled from YouTube transcripts. Here is how it works and why it matters.</description></item><item><title>How Do You Trust an Autonomous AI Agent? Evals Are the Answer.</title><link>https://www.fabswill.com/blog/how-do-you-trust-an-autonomous-ai-agent/</link><pubDate>Sat, 28 Mar 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/how-do-you-trust-an-autonomous-ai-agent/</guid><description>I run an autonomous AI agent at home — 16 cron jobs daily. It says &amp;#39;done&amp;#39; but did it actually do anything? I built an eval framework to find out. Here&amp;#39;s what broke, what I learned, and why agent evals are fundamentally different from LLM evals.</description></item><item><title>Your Next Hire Should Be an AI — Here's How a Nonprofit Did It in Two Weeks</title><link>https://www.fabswill.com/blog/your-next-hire-should-be-an-ai/</link><pubDate>Tue, 17 Mar 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/your-next-hire-should-be-an-ai/</guid><description>How MACONA went from a one-person operation to a team of two — without adding headcount. An autonomous AI executive assistant managing email, social media, newsletters, and donor outreach 24/7 on dedicated hardware.</description></item><item><title>53 Downloads, 114 Countries, Zero Marketing Budget: My First Month on the App Store</title><link>https://www.fabswill.com/blog/first-month-on-the-app-store/</link><pubDate>Fri, 13 Mar 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/first-month-on-the-app-store/</guid><description>I built two iOS apps to solve my own problems. In the first month, people in 114 countries found them through App Store search alone. Here&amp;#39;s what the data looks like, what it taught me, and why I think everyone should ship something.</description></item><item><title>WandR v1.0.1: A Ducati, a Reddit Question, and an AI Test at Union Station</title><link>https://www.fabswill.com/blog/wandr-v101-union-station-dc-first-real-ride/</link><pubDate>Sun, 08 Mar 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/wandr-v101-union-station-dc-first-real-ride/</guid><description>A rider from Slovenia asked me if my AI trip planner actually suggested anything good. So while I was in DC on an errand, I opened WandR at Union Station and let it plan a scenic motorcycle route. Here&amp;#39;s what it came back with — and the v1.0.1 update that made it smarter.</description></item><item><title>Qui Non Proficit Deficit: Three Months Offline, Two Apps Shipped, and an AI That Runs a Nonprofit</title><link>https://www.fabswill.com/blog/qui-non-proficit-deficit-im-back/</link><pubDate>Sat, 07 Mar 2026 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/qui-non-proficit-deficit-im-back/</guid><description>I disappeared from LinkedIn and YouTube for three months. Not burned out — building. Two iOS apps shipped, an autonomous AI assistant running a nonprofit, and a workflow that changed everything. Here&amp;#39;s the full story.</description></item><item><title>Embracing Technology: My Journey at TechCon 365 Dallas 2025</title><link>https://www.fabswill.com/blog/techcon365_dallas_2025/</link><pubDate>Fri, 31 Oct 2025 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/techcon365_dallas_2025/</guid><description>Exploring how technology has transformed my workflow, from taking days to mere hours with the right tools, and the importance of manual oversight.</description></item><item><title>Agent + Local Model + Evals — One‑Pager</title><link>https://www.fabswill.com/blog/2025-08-10-agent-plus-local-model-plus-evals-onepager/</link><pubDate>Sun, 10 Aug 2025 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/2025-08-10-agent-plus-local-model-plus-evals-onepager/</guid><description>&lt;h1 id="-purpose"&gt;🚀 Purpose&lt;/h1&gt;
&lt;p&gt;If you’ve ever wondered how to &lt;strong&gt;compare a locally‑hosted LLM&lt;/strong&gt; with a hosted OpenAI model &lt;strong&gt;side‑by‑side&lt;/strong&gt;, while keeping the whole experiment observable from end‑to‑end, you’re in the right place. This one‑pager walks you through the starter repo that stitches together three distinct paths:&lt;/p&gt;
&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Path&lt;/th&gt;
&lt;th&gt;Stack&lt;/th&gt;
&lt;th&gt;Observability&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;A&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;.NET 8 + Semantic Kernel + Ollama (local model)&lt;/td&gt;
&lt;td&gt;Azure Monitor (OTEL)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;B&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;OpenAI Agents SDK (TypeScript/Node)&lt;/td&gt;
&lt;td&gt;OpenAI Logs → optional OTLP → Azure Monitor&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;C&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Python Evals (OpenAI‑provided)&lt;/td&gt;
&lt;td&gt;Same prompts, same metrics&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The goal? A &lt;strong&gt;clean, reproducible harness&lt;/strong&gt; that lets you run the same prompt against a local model and an OpenAI model, invoke tools, and then score the outputs with a single Python‑based eval suite. All of this is fully instrumented with OpenTelemetry, so you can follow the request flow from the client code right into Azure Application Insights (or any OTEL backend you prefer).&lt;/p&gt;</description></item><item><title>Agent-First Commerce: I Built a Website That Sells Like a Sales Team</title><link>https://www.fabswill.com/blog/agentfirstcommerce/</link><pubDate>Sat, 02 Aug 2025 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/agentfirstcommerce/</guid><description>Over a weekend, I built and tested a working Agent-First e-commerce demo with Claude Code, MCP, and conversational checkout. Here&amp;#39;s what I learned.</description></item><item><title>Debugging GenAI with NLWeb and OpenTelemetry: A Real-Time Visibility Win</title><link>https://www.fabswill.com/blog/capswithopentelemetryandnlweb/</link><pubDate>Wed, 21 May 2025 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/capswithopentelemetryandnlweb/</guid><description>We used NLWeb and OpenTelemetry to trace, debug, and query Copilot agent plugins — all in real-time.</description></item><item><title>Vibe Coding with AI – Best Practices for Every Project</title><link>https://www.fabswill.com/blog/fabsvibecodingbestpractice/</link><pubDate>Sun, 27 Apr 2025 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/fabsvibecodingbestpractice/</guid><description>Learn how to use evaluation-first development, GitHub Copilot, Claude, and local LLMs to build faster, smarter AI-powered apps.</description></item><item><title>Start with Why: The How, the Why, and the What Behind My Reading Habit</title><link>https://www.fabswill.com/blog/startwithwhy/</link><pubDate>Mon, 20 Jan 2025 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/startwithwhy/</guid><description>A reflective look at Simon Sinek book and how it aligns with my personal approach to reading and growth.</description></item><item><title>How I Plan to Show Up in 2025 – The Journey from Why to What</title><link>https://www.fabswill.com/blog/fabianhowwillishowup/</link><pubDate>Wed, 01 Jan 2025 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/fabianhowwillishowup/</guid><description>Building a path from purpose to action – outlining the &amp;#39;How&amp;#39; behind my journey.</description></item><item><title>About</title><link>https://www.fabswill.com/about/</link><pubDate>Tue, 31 Dec 2024 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/about/</guid><description>Who is Fabian Williams &amp;amp; Why This Site Exists</description></item><item><title>A Product Management Framework for 2024</title><link>https://www.fabswill.com/blog/fabianproductmanagementframework2025/</link><pubDate>Mon, 30 Dec 2024 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/fabianproductmanagementframework2025/</guid><description>Learn 5 best practices for Product Managers to drive success – focus on customer needs, use data, and align stakeholders with a clear product vision</description></item><item><title>Mastering Llama 3.3 – A Deep Dive into Running Local LLMs</title><link>https://www.fabswill.com/blog/masteringllama3dot370b/</link><pubDate>Sun, 22 Dec 2024 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/masteringllama3dot370b/</guid><description>I compared Llama 3.3 with older models like 3.170B and Phi 3, tested Microsoft Graph API integrations using OpenAPI specs, and explored function calling quirks. The results? Not all models are created equal!</description></item><item><title>Getting started with Microsoft Graph Connectors</title><link>https://www.fabswill.com/blog/gettingstartedgraphconnectorsdev/</link><pubDate>Mon, 12 Feb 2024 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/gettingstartedgraphconnectorsdev/</guid><description>Answering the quesiton... How do I start learning Microsoft Graph Connectors in M365</description></item><item><title>Start Developing with Microsoft Graph Connectors</title><link>https://www.fabswill.com/blog/devwithgraphconnectors/</link><pubDate>Mon, 12 Feb 2024 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/devwithgraphconnectors/</guid><description>How do I start learning Microsoft Graph Connectors in M365 101</description></item><item><title>Getting started with Microsoft365 Dev</title><link>https://www.fabswill.com/blog/gettingstartedm365dev/</link><pubDate>Fri, 21 Oct 2022 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/gettingstartedm365dev/</guid><description>Answering the quesiton... How do I start learning Microsoft365 Dev</description></item><item><title>Student Loan Debt Forgiveness Survey Results - 1 Week</title><link>https://www.fabswill.com/blog/studentloanreliefresults-1week/</link><pubDate>Fri, 13 May 2022 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/studentloanreliefresults-1week/</guid><description>Sentiment Analysis Charts and Graphs from the Survey Respondents after 1 Week</description></item><item><title>Student Loan Debt Forgiveness Survey Results</title><link>https://www.fabswill.com/blog/studentloanforgiveness_24hrin/</link><pubDate>Sun, 08 May 2022 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/studentloanforgiveness_24hrin/</guid><description>Sentiment Analysis Charts and Graphs from the Survey Respondents</description></item><item><title>Mask Mandate Survey Results - 1 Week</title><link>https://www.fabswill.com/blog/maskmandatesresults-1week/</link><pubDate>Fri, 29 Apr 2022 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/maskmandatesresults-1week/</guid><description>Sentiment Analysis Charts and Graphs from the Survey Respondents after 1 Week</description></item><item><title>Mask Mandate Survey Results</title><link>https://www.fabswill.com/blog/maskmandatesresults/</link><pubDate>Thu, 21 Apr 2022 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/maskmandatesresults/</guid><description>Sentiment Analysis Charts and Graphs from the Survey Respondents</description></item><item><title>ErgoChair 2 Survey Results</title><link>https://www.fabswill.com/blog/surveyergowfhresults/</link><pubDate>Sat, 26 Mar 2022 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/surveyergowfhresults/</guid><description>Sentiment Analysis Charts and Graphs from the Survey Respondents</description></item><item><title>Cultivate a healthy Work from Home - WFH</title><link>https://www.fabswill.com/blog/cultivateahealthwfh/</link><pubDate>Sat, 19 Mar 2022 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/cultivateahealthwfh/</guid><description>If you work from home hence remote you should create an environment to enhance your physical well being</description></item><item><title>Understand and Improve our Website value through Microsoft Clarity &amp; Bing</title><link>https://www.fabswill.com/blog/clarityandbing/</link><pubDate>Sat, 22 Jan 2022 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/clarityandbing/</guid><description>Using the FREE Microsoft Clarity and Bing Webmaster Tools you can gain better understanding of your Website and enhance its value</description></item><item><title>2022 Top Three Journey</title><link>https://www.fabswill.com/blog/2022top3journey/</link><pubDate>Sat, 01 Jan 2022 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/2022top3journey/</guid><description>A journey list not a goals list or a retropsective. A reminder list for when days go sideways</description></item><item><title>Gotcha with CICD and Azure Static WebApps</title><link>https://www.fabswill.com/blog/cicdgotchastaticwebapps/</link><pubDate>Thu, 30 Dec 2021 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/cicdgotchastaticwebapps/</guid><description>If you lost your token on an Azure Static WebApp integrated with GitHub you could be SOL</description></item><item><title>Take your outdoor walk to the next level</title><link>https://www.fabswill.com/blog/mindfulwalk/</link><pubDate>Tue, 28 Dec 2021 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/mindfulwalk/</guid><description>Bionic ears are here in the form of Transarency mode in Earbuds and Headphones</description></item><item><title>Manage your Time with Graph in Calendar Events</title><link>https://www.fabswill.com/blog/blockyourtimewithgraph/</link><pubDate>Mon, 24 May 2021 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/blockyourtimewithgraph/</guid><description>Creating a Calendar event with Microsoft Graph that repeats an existing recurring event adhoc</description></item><item><title>Azure Web App DNS Gotcha</title><link>https://www.fabswill.com/blog/azurestaticwebappdnsgotcha/</link><pubDate>Sat, 15 May 2021 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/azurestaticwebappdnsgotcha/</guid><description>Creating an Apex Domain registration will error out if your DNS Zone is in a different subscription from your Web App by default</description></item><item><title>Microsoft Graph and Python</title><link>https://www.fabswill.com/blog/graphwithpython/</link><pubDate>Mon, 10 May 2021 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/graphwithpython/</guid><description>Using Microsoft Graph and Python to access Microsoft 365 Data and Services</description></item><item><title>A New Day</title><link>https://www.fabswill.com/blog/anewday/</link><pubDate>Sun, 02 May 2021 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/anewday/</guid><description>My 1st Blog on my new Static Site Platform</description></item><item><title>Tweets and Vids of Fabian</title><link>https://www.fabswill.com/blog/gettingstarted/</link><pubDate>Sat, 01 May 2021 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/gettingstarted/</guid><description>Highlight of some of Fabians work</description></item><item><title>Markdown Syntax Guide</title><link>https://www.fabswill.com/blog/fabianmarkdowncheatsheet/</link><pubDate>Fri, 30 Apr 2021 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/blog/fabianmarkdowncheatsheet/</guid><description>Fabian Cheat Sheet &amp;amp; Sample article showcasing basic Markdown syntax and formatting for HTML elements.</description></item><item><title>Contact</title><link>https://www.fabswill.com/contact/</link><pubDate>Mon, 01 Jan 0001 00:00:00 +0000</pubDate><guid>https://www.fabswill.com/contact/</guid><description>How to reach Fabian G. Williams</description></item></channel></rss>