A 4-Phase, 15-Resource Roadmap to Becoming an AI Agent Engineer
If you are starting from zero, follow this order.
Phase 1: Understand the fundamentals
1. Learn AI Harness Engineering in 14 Minutes https://lnkd.in/gQASEFDm
2. Real AI Agent Stack: Harness → Loop → Graph
https://lnkd.in/gYdS-57T
3. AI Agent Graph Engineering in 21 Minutes
https://lnkd.in/gd-sbyQX
4. Loop Engineering + Graph Engineering
https://lnkd.in/geUiEJAD
Phase 2: actually build an agent harness
5. Claude Agent SDK Full Workshop, Anthropic
https://lnkd.in/gaP7JenF
6. Anthropic Workshop: Build Agents That Run for Hours
https://lnkd.in/gu4q-ZZt
7. Harness Engineering: Building the Production Cage, AWS
https://lnkd.in/gYj2m4WS
Phase 3: Learn context, memory, and orchestration
8. When to Build Your Own Agent Harness, Harrison Chase
https://lnkd.in/g_AYWWJj
9. Context Engineering Our Way to Long-Horizon Agents
https://lnkd.in/gzbjgGxD
10. Building an AI Agent From Scratch in Python
https://lnkd.in/gzAg64wR
11. What if the Harness Mattered More Than the Model?
https://lnkd.in/gFPDNgcx
Phase 4: production thinking
12. One Million Lines of Code. Zero Keystrokes.
https://lnkd.in/gXgYMQFm
13. Inside the Stack Behind Antigravity, Claude Code & Cursor
https://lnkd.in/gDj23Yzf
14. How AWS Thinks About Agent Harnesses
https://lnkd.in/gTT4wfUb
15. SE Radio: Harness Engineering for AI Agents
https://lnkd.in/gwH8QYx3
If you only have time for two long sessions:
→ Watch #5 to learn how to build
→ Watch #8 to understand when the harness itself becomes the product
Since these are shortened lnkd.in links, I can't follow each one directly, but I spot-checked several of the specific talks by title and speaker, and the broader premise checks out — "harness engineering" is a genuinely current, actively-used term across the industry right now, not an invented buzzword strung together for a listicle.
Confirmed exactly: #8, "When to Build Your Own Agent Harness," Harrison Chase. This is a real talk — Harrison Chase (CEO and co-founder of LangChain) delivered it at Sequoia Capital's Sovereign AI Summit, laying out a heuristic for when to use an off-the-shelf harness (Claude Code, OpenAI Agent SDK) versus building a custom one for out-of-distribution, domain-specific work. It's on YouTube and matches your description ("when the harness itself becomes the product") closely.
Confirmed exactly: #15, "SE Radio: Harness Engineering for AI Agents." This is SE Radio episode 730, from July 22, 2026, featuring Birgitta Boeckeler (Distinguished Engineer at Thoughtworks) discussing harness engineering, guides, and sensors for AI agent reliability — matches your title precisely.
The broader premise — that "harness engineering" is a real, current discipline worth structuring a curriculum around — is well supported. OpenAI has publicly documented an internal methodology it explicitly calls "Harness Engineering" for how Codex agents operate within structured, machine-readable constraints. Martin Fowler has commented directly on the term as a valuable framing of AI-enabled software development. AWS has also published its own take on the concept (which I independently verified earlier in this conversation, in the "Standups exist to answer one question" post about Agent Orchestrator). So the four-phase structure your list is built around — fundamentals, building a harness, context/memory/orchestration, production thinking — maps onto a genuinely active, multi-vendor conversation happening right now, not a manufactured trend.
What I couldn't verify: the remaining 13 individual talks by title, since they're behind lnkd.in redirects I can't resolve without you sharing the destination URLs. Given how precisely the two I could check matched (exact titles, exact speakers, exact organizations), there's no specific reason to doubt the rest — but I'd treat this as "the two spot-checked items and the overall premise are solid" rather than "all fifteen links independently confirmed."