Agentic Engineering: Working With AI, Not Just Using It — Brendan O'Leary

By ai.engineer

Categories: AI, Tools

Summary

AI coding agents represent a paradigm shift from tools to collaborators—engineers who work *with* AI gain 30% time savings by strategically deciding what to hand off versus keep. The critical skill isn't prompting; it's context engineering: loading just enough information to avoid degrading model quality beyond 50% context capacity.

Key Takeaways

1. Reframe AI agents as 'confidently wrong junior developers'—fast, tireless, broad knowledge, but lacking business judgment and architectural context. Success requires directing their work, not blindly accepting suggestions.
2. Context engineering is the core skill: more context doesn't mean better results. Quality degrades above 50% context fill, and bad context (outdated comments, mixed tasks) actively poisons model outputs. Every token has cost implications.
3. The evolution is real: 2020 (line completion) → 2022 (full function suggestions) → 2025 (task breakdown, file changes, test execution, pull requests). This isn't fancy autocomplete—it's collaborative execution.
4. Only ~50% of engineers using AI do so regularly; most can't articulate their actual workflow. The gap between 'using AI' and 'working with AI' determines whether you gain real productivity or just autocomplete through your day.
5. Arman (Flask creator) gained 30% daily time by knowing what to hand off and what to keep—this requires explicit judgment calls, not passive acceptance. This directional discipline is what defines agentic engineering.

Topics

• Agentic Engineering
• Context Engineering for AI
• AI-Assisted Code Generation
• Mental Models for AI Collaboration
• LLM Token Optimization

Transcript Excerpt

Let's talk a little bit about what I mean by agentic engineering. And let's maybe start with a question. If I were to ask you right now, how are you using AI in your work? Could you actually really explain it? Not just, you know, it helps me code faster, it can write code really fast, but like the real workflow, what you hand off, what you keep, how you decide in between. Most engineers can't. And that's a little wild to me because 90% of engineers are already using AI tools or have used them. M...