AI Subjects Learning

AI Proficiency Requirements

Proficient integrating AI coding assistants, CLIs, and MCPs into daily development workflows.
Uses AI effectively across the SDLC - planning, design, implementation, debugging, deployment, and observability.
Builds reusable AI tooling (agent rules, skills, internal MCPs) that compounds team productivity, with sound judgment on when AI is (and isn’t) the right tool.
Balances AI-driven velocity with engineering rigor - automated testing, structural guardrails, and CI checks to validate AI-generated code.

Anthropic - Claude Code: Best practices for agentic coding - the canonical field guide for CLI-driven agentic coding: CLAUDE.md conventions, workflow patterns, multi-agent orchestration.
Claude Code Docs - Best Practices - living doc, updated frequently; covers plugins, skills/subagents/hooks/MCP decision-making, and codebase-onboarding patterns.
Anthropic - Writing effective tools for AI agents - how tool/MCP description quality directly affects agent reliability (SWE-bench data included).
DeepLearning.AI - MCP: Build Rich-Context AI Apps with Anthropic - free, ~2 hrs, taught by Anthropic’s head of technical education; builds a real MCP server/client end to end (FastMCP, resources, prompts, remote deployment).
Simon Willison - MCP tag archive and his Agentic Engineering Patterns - practitioner notes on what actually works day-to-day, including where MCP helps vs. where plain CLIs/libraries beat it.
Firecrawl - 10 Best MCP Servers for Developers in 2026 - practical map of which MCP servers are worth wiring into a real workflow right now.

Martin Fowler - martinfowler.com AI-assisted programming articles (search “AI” tag) - his pieces on LLM-augmented programming, spec-driven development, and “on the loop” human roles are the most rigorous non-vendor writing on this.
Kief Morris / Martin Fowler - Where Do Humans Fit in AI-Assisted Software Development (InfoQ summary) - on shifting from reviewing every artifact to designing specs/tests/feedback loops that govern agents.
Google Cloud DORA - Balancing AI tensions: Moving from AI adoption to effective SDLC use and the 2025 DORA State of AI-assisted Software Development report - the best evidence-based look at where AI speeds up the SDLC and where it just relocates the bottleneck (“verification tax”).
The New Stack - Martin Fowler on Preparing for AI’s Nondeterministic Computing - his “treat every AI diff like a PR from a productive but untrustworthy collaborator” framing, plus tolerance/structural-engineering analogy for judging AI fit.

Given what’s already in your Copilot agent infrastructure work, these map directly onto it:

Anthropic - Equipping agents for the real world with Agent Skills - the foundational explainer: skills as folders (not just markdown), when code beats tokens, how discovery works.
Anthropic - Lessons from building Claude Code: How we use skills - internal Anthropic playbook: only write skills for things that push the model off its defaults, “Gotchas” sections as highest-signal content, iterate from real failures.
Anthropic - Building effective agents - the widely-cited framework for choosing workflows vs. autonomous agents, and when not to reach for agentic complexity - directly useful for your “sound judgment on when AI is/isn’t the right tool” goal.
shanraisshan/claude-code-best-practice GitHub repo (referenced in the mcp.directory writeup) - community-maintained, heavily-starred consolidation of commands/agents/skills/hooks/MCP patterns.

minware - Why the Test Pyramid Matters More Than Ever in AI-Assisted Development - concrete argument for fast local checks as the safety net against AI-inflated PR size and defect rate.
Martin Fowler’s site - pieces on using test suites as regression sensors for AI agents, and static-analysis “maintainability sensors” (linting, dependency rules, coupling) to keep AI-touched codebases clean - both referenced in the AllDevBlogs Fowler roundup.
DORA - ROI of AI-assisted Software Development report - introduces the “verification tax” and J-curve model; useful for arguing to a team/manager why review and CI investment must scale alongside AI adoption, not lag behind it.
InfoQ - New DORA Report: Strong Engineering Foundations Drive AI ROI - summary of why platform quality and workflow clarity (not tool choice) determine whether AI velocity gains survive contact with production.

Simon Willison - Engineering practices that make coding agents work (The Pragmatic Summit, ~30 min, Feb 2026) - fireside chat covering red-green TDD with agents, his “Showboat” manual-testing tool, the lethal-trifecta security framework, and which model he actually trusts for professional work. Companion write-up with extra notes: simonwillison.net/2026/Mar/14/pragmatic-summit.
Agent Skills or MCP in the era of Claude Code? - Tim Berglund (Confluent/DevRel veteran) lightboard explainer on how Skills changes the calculus for when you reach for MCP vs. skills vs. plain tool use.
AWS re:Invent 2025 - Accelerate multi-step SDLC with Kiro (DVT321) - live demo building an app end-to-end with a custom agent config using the “AI-DLC” methodology; useful even if you don’t use Kiro, since the phase breakdown (spec → design → tasks → build) generalizes to any CLI agent.

The New SDLC: From Vibe Coding to Agentic Engineering (Google × Kaggle AI Agents Intensive) - walks through how AI touches each SDLC phase differently, with a clear framing of what changes at planning vs. implementation vs. review stages.
GitHub Universe 2025 Announcements + Reactions - GitHub staff reacting live to Agent HQ, Copilot’s move into planning/orchestration/review, and the shift toward what GitHub calls the “agentic development lifecycle.” Good for seeing where platform vendors are pushing observability and control planes.

Máster Claude Code 2026: Agentes IA, MCP, Skills y Comandos (Spanish) - long-form, hands-on walkthrough of building custom agents, skills, and slash commands; skip if you need English, but it’s one of the more thorough free build-alongs if Spanish works for you.

Most of the good material here (Fowler, DORA, minware) is written, not recorded. If a talk surfaces later, the QCon AI Boston 2026 session “Prompt to Prod: Engineering an Autonomous SDLC at Scale” (Roblox’s engineering-acceleration lead, on agentic guardrails and self-healing codebases) may get a recording posted post-event - worth checking QCon’s YouTube channel.