AI that works where you work.
InnerStack tracks every workflow, prompt and tool we test in the field, then records what we actually adopted and why. Read the journal, copy the stack, or have us install it in your office — where your customer data never leaves the building.
What we found, and where it went
Every entry is tagged to one of our workflows and carries a status: watching, evaluated, or adopted. Nothing here is auto-generated — the agents gather, a person decides.
LiteSpeed Advanced over Aggressive
The host offers three caching presets. The fastest one is also the one most likely to break a page you have not built yet.
Post types in a code snippet, not a plugin UI
Custom post types and taxonomies registered in one PHP snippet, instead of assembled by clicking through CPT UI.
Skipping the Kadence starter templates
The starter templates are the fastest way to a finished-looking site and the slowest way to the one you designed.
Splitting the pipeline: Hermes gathers, Claude decides
Two agent platforms, split by the kind of work rather than by preference: parallel gathering on Hermes, judgement on Claude.
Seven stacks we run and revise in public
Each workflow is a versioned page. Journal entries attach to it; adopted changes bump the version and land in the changelog. What our agents follow is exactly what you read here.
What changed, and the reason
Every adopted entry becomes a diff against the workflow it improved — what it replaced, and why. Reviewable by anyone.
SEO
First published version. Covers the technical baseline that is live, and marks the answer-engine checklist as still being evaluated rather than claiming it runs.
Agentic Structure
First published version. Written from the pipeline as it actually runs: gathering on short-lived Hermes workers, judgement on Claude with a person in the loop.
Website Development
First published version. Documents the stack this site is built on, including the two places we deliberately went against the conventional answer.
Four hubs, written for the person who has to make it work
Plain over clever, specific over hype, honest about tradeoffs. Some reviews tell you not to buy the thing.
AI Tools
Reviews, roundups and head-to-heads with real prices, real model names and what we would pick.
Browse reviews → HUB 02 · 2–3 / WEEKAI for Your Business
AI for dentists, restaurants, realtors, contractors — one trade at a time, with the exact stack.
Find your trade → HUB 03 · 2 / WEEKLocal & Private AI
Ollama, LM Studio, hardware picks, privacy setups. The guides behind everything we install.
Start here → HUB 04 · DAILYNews & Updates
Short, dated, sourced. What shipped, what it costs, whether it matters to a small office.
Today’s brief →Private AI, installed in your office.
We set up a local model on hardware you own, wire it to the tools you already use, and hand you the same workflow pages you have been reading. Nothing your customers tell you is sent to a cloud provider.
- A fit callTwenty minutes. We tell you if we are not the right fit.
- Site surveyWhat you run, what your data is, what has to stay inside.
- Install dayHardware, a local model, and your tools wired together.
- Monthly supportUpdates, new workflows, and a person who answers.
One email a week: what we adopted, what we dropped, and why.
The changelog in your inbox, with the two guides worth your time. Written by a person; the agents only gather.