Technical Articles
Practical Wear OS guides and deep dives into AI engineering, software development, and infrastructure.
RAG Permissions: Stop Your Chatbot Leaking Documents Users Should Not See
An internal chatbot that can read everything will eventually quote something the person asking was never allowed to open. The fix is document-level permissions enforced at retrieval time, plus tests that prove it.
MCP Security Checklist: Letting AI Agents Touch Your Infrastructure
Giving an agent MCP tools for AWS, Kubernetes or Terraform is useful right up to the moment it runs the wrong command. This checklist covers scoped tools, dry-run-first changes, approvals, audit logs and secrets.
Hybrid LLM Routing: Send Each Request to the Cheapest Model That Can Handle It
Most production LLM traffic does not need the most expensive model. A router that sends each request to the cheapest model that passes your evals, with tested fallbacks, usually cuts cost without users noticing.
What Does a Private LLM Actually Cost? GPUs, APIs and the Ops Bill Nobody Quotes
A private LLM is rarely cheaper than an API on tokens alone. It pays off when your data cannot leave, when volume keeps GPUs busy, or when you need to pin a model. Here is how to run the numbers honestly.
Small Business AI That Pays for Itself: A One-Page Test Before You Buy
Most small firms now use some AI, but few can say whether it paid off. A one-page business case, a named human check, and a short list of boring, repetitive tasks beat any tool list.
How to Trigger a Webhook from a Wear OS Watch
Yes, you can fire a webhook from Wear OS. The useful setup is to configure it on a phone, keep the watch as the deliberate trigger, and be strict about connectivity and credentials.
How to Run Multiple Timers on Wear OS
Running one timer is easy. Running several without mixing them up is mostly a labeling and feedback problem, which is where a dedicated Wear OS multi-timer earns its place.
How to Use a Tally Counter on Wear OS
A watch is a surprisingly good tally counter because the action can stay local, fast, and one-handed. The main thing to get right is what one tap means and how you avoid accidental counts.
How to Convert Units from a Wear OS Watch
Fixed unit conversions are a good fit for a watch because they can stay local and offline. The real UX win comes from fast input, favorites, and not digging through the same unit list every time.
How to Use a Wear OS Watch as an Ambient Light Meter
A Wear OS watch can report ambient light only when the hardware exposes a light sensor to apps. If it does, the watch is useful for relative lux readings, not as a substitute for a calibrated meter.
How to Read Barometric Pressure on Wear OS
A watch with a real barometer can track local pressure without a phone or cloud service. The useful part is the trend over time, not overinterpreting one number.
Deploying Llama 3 8B on Consumer Hardware
On consumer hardware, local LLM reliability is mostly a memory, context, and concurrency problem. This article keeps the deployment shape deliberately small so failures are easy to see.
Optimizing Recomposition in Wear OS Compose
On a watch, sloppy state handling shows up fast as jank, extra work, and battery cost. The goal is not “zero recomposition”; it is keeping fast-changing state from invalidating more UI than necessary.
RAG Architecture Patterns for Enterprise
Most RAG failures are not model failures. They come from retrieval, permissions, stale content, or a pipeline nobody can inspect when an answer goes wrong.