MATT NIGH'S BLOG
Home
Writing
Home
Writing
Archive
Writing.
Every essay, note, and external piece I've published. Filter by tag.
All
30
engineering
18
ai
11
management
6
copilot
3
favorite
2
personal
2
neurodiversity
2
future-of-work
1
prompts
1
template
1
productivity
1
resources
1
reading
1
chief-of-staff
1
Run multiple agents at once with /fleet in Copilot CLI
↗
APR 2026
ai
engineering
copilot
TL;DR ↗
/fleet dispatches multiple sub-agents simultaneously, with an orchestrator planning work decomposition and coordinating parallel execution
Well-structured prompts with specific deliverables and clear file boundaries enable effective parallelization
Sub-agents share a filesystem but lack direct communication — declaring dependencies and partitioning files prevents conflicts
Custom agents defined in .github/agents/ can be specialized for different tasks with different models and tool access
Fleet maximizes efficiency for multi-file refactors, documentation generation, and feature implementations spanning API, UI, and tests
How to write a great agents.md: Lessons from over 2,500 repositories
↗
NOV 2025
engineering
ai
copilot
TL;DR ↗
A great agents.md file is the key to getting useful output from AI coding agents
Analyzed over 2,500 repositories to find what makes instruction files effective
Successful files are specific about architecture, conventions, and testing expectations
Clarity and structure in your instructions directly correlates with agent output quality
Playbook series: Right-fit tooling
↗
NOV 2025
ai
engineering
TL;DR ↗
The right AI tools match your team's actual workflows, not the latest hype
Evaluate tools against real use cases, not feature checklists
Start narrow with one tool well-integrated, then expand based on feedback
Tool selection should involve the developers who will actually use them
Playbook series: Executive support
↗
NOV 2025
ai
engineering
TL;DR ↗
Executive support is the single biggest predictor of AI adoption success
Leaders must visibly champion AI, not just approve budget for it
Exec sponsorship creates psychological safety for teams to experiment
Top-down and bottom-up alignment accelerates meaningful adoption
Playbook series: Creating clear AI policies and guardrails
↗
OCT 2025
ai
engineering
TL;DR ↗
Clear AI policies remove ambiguity and give teams confidence to move fast
Guardrails should enable experimentation, not block it
Address data privacy, IP, and acceptable use upfront to prevent paralysis
Policies should evolve as the technology and your team's maturity grow
Playbook series: Why you need a DRI for your AI program
↗
SEP 2025
ai
engineering
TL;DR ↗
AI adoption without a dedicated owner drifts and stalls
A DRI (directly responsible individual) keeps the AI program focused and accountable
The DRI bridges the gap between executive vision and on-the-ground execution
Ownership clarity prevents duplication of effort and conflicting initiatives
Playbook series: Fostering AI learning opportunities
↗
SEP 2025
ai
engineering
TL;DR ↗
AI fluency requires ongoing learning opportunities, not a one-time training
Create safe spaces for experimentation where failure is expected
Pair structured programs with organic peer-to-peer knowledge sharing
Learning is most effective when tied to real work, not hypothetical exercises
Playbook series: Activating your internal AI champions
↗
AUG 2025
ai
engineering
TL;DR ↗
Internal champions are your secret weapon for AI adoption at scale
Find advocates who are already experimenting with AI organically
Empower champions with resources, visibility, and a feedback loop to leadership
Grassroots enthusiasm drives adoption faster than top-down mandates
Thoughts on AI Pair Programming
FEB 2025
ai
future-of-work
TL;DR
AI pair programming excels at rapid prototyping, documentation lookup, and pattern recognition
Over-reliance on suggestions is a real risk, especially when tired
Knowing what needs careful verification vs. what doesn't is a key skill to develop
Context limitations have improved dramatically — file limits are mostly gone now
Useful Self-Hosted AI
FEB 2025
ai
TL;DR
Running AI locally with Open WebUI, LM Studio, and self-hosted tools
Web search powered by SearXNG for privacy-respecting AI queries
Whisper for local speech-to-text transcription
Home Assistant integration via Assist for smart home AI control
Using GitHub Copilot to Write Blogs
FEB 2025
ai
copilot
prompts
TL;DR
Copilot can scan PRs, issues, and commits to generate blog ideas and outlines
Includes prompts for brainstorming angles, drafting full posts, and refining tone
Use Copilot to combine multiple PRs/issues into a cohesive narrative
Works for everything from deep technical dives to contributor shout-outs