~/prompt-to-productionon» dev✓

# understanding AI tools for better outcomes

$ls posts/ --sort=recent

2026-09-04

Same Mechanic, Different Shop: Why AI Code Quality Is a Tooling Problem

The posts about hallucinated APIs and bad AI code describe something real. They just blame the wrong thing.

2026-08-13

The Limits of Telling: Why Instruction Files Don't Do What You Think

An instruction file feels like enforcement, but it only ever asks. Five quiet ways CLAUDE.md, AGENTS.md, and system prompts fail to do what you told them — and what to reach for when a rule actually has to hold.

ai-tooling workflow meta

2026-07-23

Other Handoff Situations: Getting What You Need From Others

The best handoffs to AI-driven development aren't longer explanations — they're structured enough to leave nothing to guess. A finance consultant's posting diagram, and the ledger it produced, made the difference visible.

workflow ai-tooling meta

2026-07-21

Moving Between Different Tools: Hitting the Capability Wall

Every AI tool is good at something specific, which means you're always eventually moving work from one to another. The thing that makes that move clean isn't a better prompt — it's a document.

workflow ai-tooling meta

2026-07-20

Knowing Where to Spend AI's Time: Building the DGX Spark Control Plane

I had Claude Code manually swapping models on my DGX Spark every session — performing the same task by hand every time instead of solving the actual problem. I stopped wasting Claude's time and finally had it build a proper tool.

ai-collaboration local-llm dgx-spark case-study

❯ view all posts →