What Developers Actually Need to Know Right Now

2026-02-19 · Source: AI & ML – Radar · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Emerging Technologies & Innovation · Depth: Intermediate, medium

Summary

Addy Osmani, formerly of Chrome's developer experience team and now with Google Cloud AI, discusses the current state and future trajectory of AI in software engineering. He emphasizes that the primary challenge with AI agents is coordination and orchestration, not code generation itself, highlighting frameworks like Google's Agent Development Kit (ADK) and protocols like A2A and MCP. Osmani also addresses the "Something Big Is Happening" debate, cautioning against overstating AI's current production readiness versus its prototyping capabilities. He notes a shift where planning and defining constraints are becoming more critical than coding, and that code review is increasingly complex due to AI-generated pull requests. Despite these challenges, he strongly encourages new entrants into software engineering, viewing the current era as an unprecedented opportunity for innovation. Osmani also advises careful experimentation with token costs and observes that many 2028 AI predictions are already technically feasible, with adoption being the slower factor.

Key takeaway

For AI Architects evaluating agent-based development, you should prioritize robust orchestration and coordination strategies over raw generation capabilities. Focus on defining clear planning phases, establishing precise constraints, and implementing rigorous quality gates for AI-generated code. Your ability to manage agent interactions and ensure traceability will be crucial for successful enterprise-scale deployments, rather than merely deploying numerous agents without oversight.

Key insights

AI's true frontier lies in orchestrating agents to solve real problems with control and traceability.

Principles

• Coordination, not generation, is the hard problem for AI agents.
• Planning is the new coding in an AI-assisted development workflow.
• The gap between AI capability and adoption is where innovation will occur.

Method

Utilize frameworks like Google's Agent Development Kit (ADK) to integrate deterministic workflow agents and dynamic LLM agents, balancing predictability and flexibility for specific tasks.

In practice

• Codify team best practices for agents in Markdown or MCP tools.
• Define clear quality bars for merging AI-generated code.
• Experiment with token costs to assess productivity gains.

Topics

• AI Agents
• Agent Orchestration
• Software Engineering
• LLM Productivity
• AI Adoption

Code references

• google/A2A

Best for: AI Architect, Software Engineer, Machine Learning Engineer, AI Engineer

Related on AIssential

• Agentic AI’s challenge is getting agents to act like a team, not a crowd — Effective multi-agent AI systems require dedicated coordination infrastructure to ensure agents work as a team, not in isolation.
• From IDEs to AI Agents with Steve Yegge — AI is fundamentally reshaping software engineering, shifting focus from manual coding to agent orchestration and rapid prototyping.
• AI Dev 26 x SF | Diamond Bishop: The Next 100 Agents. Building the Agent Native Office — Scaling enterprise AI agents requires agent-first design, proactive operation, continuous evaluation, and model/framework agnosticism.
• The Emerging "Harness Engineering" Playbook — Effective AI agent integration requires dedicated "harness engineering" to structure environments and manage agent workflows.
• Presentation: Directing a Swarm of Agents for Fun and Profit — Orchestrating AI agent swarms enables rapid, cost-effective, and lower-carbon software development, shifting focus to platform engineering.
• Inside Multi-Agent Orchestration: State, Security, Scaling and Future — Multi-agent orchestration coordinates specialized AI agents to achieve complex shared goals, mirroring project management for AI workflows.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI & ML – Radar.