It's Not The Tokens Honey It's The Model
Cloud AI loves to burn through tokens and we mistake that for speed and productivity but it ain't all about the tokens baby...
One Intentionality To Rule Them All...
FOMO, novelty and marketing are three forces making the adoption and implementation of new tech like AI confusing. But there's a way out.
Musnad Middle Ground — Phase 5 Progress and the Agentic Next Step
The first Musnad records cleared full human review and scholarly verification, a shared review workspace went live, and the next milestone is a deterministic, agent-operable review layer.
Building Musnad Middle Ground: Phases 1–4
Four infrastructure phases down: canonical data model, historical candidate extraction, human review and promotion, and scholarly verification — the challenges, the fixes, and what's still deliberately undone.
Middle Ground Musnad — Project Overview
A structured repository connecting canonical aḥādīth to every khuṭbah, class, post, and podcast where Middle Ground has used them — architecture, data model, guardrails, and project phases.
Remember to Get Out of Your Shell
There's a world beyond the shell, beyond the screen. It's good to enjoy it.
Vander-Mattox Omarchy Security Audit: Outcome v1
A security audit of the vander-mattox MacBook Air running Omarchy — closing the two urgent findings and tracking what's still open.
senior-moments Is Live — What It Took to Actually Ship It
senior-moments went from a working pipeline to a live site this week — the stack, two real deploy failures, a Cloudflare wall that wasn't fixable, and what it took to actually filter events for relevance.
How Senior Moments Actually Works (And Why It's Not on Telegram Yet)
A walk through the senior-moments scraper pipeline — how it collects, filters, and publishes local events, why Telegram isn't wired up for it yet, and what it'll take to add more sites.
Omarchy Quattro - New Life For An Oldie But Goodie
Omarchy, Linux, and new life for an old machine.
We're Still Hallucinating And It's Okay
The reports of my death are greatly exaggerated. - Mark Twain
Delegating the Feed, Part 4: The Agent That Fixed Its Own Bug
An unattended OCR pipeline for a nonprofit's flyer hits four separate bugs in one afternoon — including an agent that quietly patched around its own permission error instead of fixing it.
Delegating the Feed, Part 3: Twingate, Tailscale & Telegram - A Tale of Three T's
A laptop's changing IP breaks a home AI pipeline, Tailscale looks like the fix, and then a healthy-looking status line hides the one environment variable that actually mattered.
Delegating the Feed, Part 2: The Bug That Wasn't My Fault (Mostly)
Two and a half hours to click one button — a documented Meta platform bug, a wrong instruction I gave in good faith, and an Instagram settings menu that hid the fix one level too deep.
Delegating the Feed, Part 1: The Paperwork Before the Post
Turning newsletter automation into Instagram posting sounds like reusing a pattern, until Meta's Business Verification and App Review turn one button into a paperwork exercise.
Zero-Config, Zero-Server: How This Blog Builds and Deploys
No YAML workflows, no Dockerfiles, no edge adapters — how a pure static Astro build and a plain git push replace a conventional CI/CD pipeline.
Domain-First Guardrails: Building al-Khizanah for Classical Scholarship
A fully offline RAG search engine over classical Arabic and Islamic texts, built with hand-authored guardrails against the silent errors a domain like scholarship can't tolerate.
The Hybrid Local AI Rig: Running Hermes and Qwen on Apple Silicon
What it's actually like running a local model as your daily driver, with cloud APIs kept around as a fallback, across two very different Apple Silicon workloads.
Debugging the Invisible: Silent Failures in Agentic Automation
Three ways an AI-driven pipeline can look like it's working — clean exit code, calm chat transcript — while quietly doing nothing, or the wrong thing.
Welcome
First post — more to come on the projects I'm working on.