The ATD journal
Practical field notes on AI, systems integration, cloud engineering, and compliance in regulated environments.
Explore the field notesWhat we learn.
What we build.
What comes next.
From the studio
Decisions, discoveries, and lessons
from building things that work.
11 field notes / Latest first
The latest field note Applied AI
Most AI consulting pitches fall apart the moment a client asks, 'just tell me the number.' Here's how to quantify AI-accelerated engineering with three metrics a client can verify in their own systems — cycle time, AI-authored code share, and defect rate — and how to price the speed so the savings still hold up past the second invoice.
5 min read
Read the field noteContinue exploring 10 more field notes
Cloud engineering
The hard part of enterprise AI in 2026 is the plumbing and not the model. We present a field guide to VPCs, peering, Transit Gateway, and the NAT Gateway bill that will blindside your AI workload.
5 min read
Healthcare & governance
The FDA premarket guidance gets the headlines, but for small device manufacturers the bigger surprise is what hospital security teams ask for. A practical playbook for the MDS2, SBOM, HDO questionnaire, and HSCC contract language.
6 min read
Systems & integration
Most AI coding tools write fluent code with the architectural awareness of someone who started yesterday. Here's how Model Context Protocol changes that — and why smaller engineering teams stand to benefit the most.
5 min read
Applied AI
When RAG works, it's the closest thing to an expert in a box that anyone has shipped. When it doesn't, it fails confidently, with footnotes. A breakdown of the patterns that work in 2026, the failure modes nobody talks about, and how to tell the difference.
6 min read
Applied AI
Most companies skip the build, buy, or partner question and head straight to procurement. That's why 95% of AI pilots produce no measurable ROI. Here's the four-question framework that actually works.
6 min read
Systems & integration
The hard part of putting AI on top of old systems isn't the model. It's the wiring. Here's the bridge pattern that keeps legacy systems productive without a rip-and-replace.
6 min read
Systems & integration
The Model Context Protocol is becoming the default integration layer for agentic AI. Here's how it works, why 2026 is the inflection year, and what it means for small and mid-size teams building AI-powered tools.
7 min read
Healthcare & governance
70% of hospital AI pilots never make it past limited deployment. The problem isn't the technology — it's the missing governance layer between 'this model works' and 'we trust it enough for patients.'
7 min read
Applied AI
A plain-English guide to agentic AI — and how small businesses can use it today without a tech team.
5 min read
Applied AI
How a growing business can pick a first AI project, check it works, and keep automation accountable. One lab client cut hands-on reporting time from 20+ hours a week to under one.
2 min read
From thinking to making
Explore the systems, products, and experiences behind the field notes.
Explore selected work