Agentic AI insights
Essays on the agentic AI market, paired with AgentWay paths you can use to build the skill behind each idea.
- Jun 23, 2026 · 7 min read
Preparing for Agent/Harness teams: portfolio to experiments
Top Agent and Harness teams want a portfolio of reproducible behavior experiments: task sets, interventions, traces, evals, and failure attribution.
EssayCareerEngineering - Jun 22, 2026 · 5 min read
DeepSeek Harness team's skill tree, decoded
DeepSeek's Harness job descriptions imply a compact agent skill tree: model-external systems, context, tools, multi-agent work, evals, and real feedback.
EssayEngineeringCareer - May 20, 2026 · 10 min read
Build agents with agents: the loop that actually works
Building agents with agents stalls at the one-giant-prompt stage. Working teams run a loop: eval first, narrow scope, subagents, hooks, and trace replay.
EssayEngineeringFrameworks - May 17, 2026 · 7 min read
Using an agent is not the same as directing one
A viral Peking University thread captured a familiar 2026 AI anxiety. Engineers are now sorted by whether they can direct AI, not merely use it.
EssayLaborCareer - May 10, 2026 · 9 min read
Six AI product mindsets: don't stop at wrapping the model
AI features fail when teams apply deterministic product habits to probabilistic systems. The products that keep improving share six mindsets.
EssayProductFrameworks - May 8, 2026 · 9 min read
No AI job apocalypse, but work is being redistributed
The macro data does not support an AI job apocalypse. It does show where work is being reallocated — and the advantage is not going to chatbot-only users.
EssayLaborEconomics - May 5, 2026 · 9 min read
Why Dubai could become an agentic AI testbed
Most cities will adopt agentic AI. Dubai is one of the few that could help agentic companies form, test, and scale locally — it already has the conditions.
EssayPolicyUAE