State of Agentic Coding

In this session, we are going to revisit the major developments in agentic coding from over the past few months, from MCP to skills and beyond; discuss how to improve trust in your agents and take a tour of the new Google Antigravity ecosystem. Every technique discussed will be paired with a live demo using practical examples of day to day development so we can see the coding agent in action.

Supporting materials for this session can be found at: https://github.com/danicat/state-of-agentic-coding

Transcript

1. 2026 State of Agentic Coding
2. Google Cloud 02 Daniela Petruzalek Developer Relations Engineer, Google Cloud
3. Google Cloud 3 Agenda Antigravity 2.0 Warm-up Exercise Tools and Techniques 02 01 03
4. The 7 Stages of Agentic Coding
5. Google Cloud 05 You treat AI as a separate entity Ask Gemini questions in a browser tab and manually copy-paste the occasional snippet back into your editor Low trust, low integration Zero integration Stage 1: Using Gemini in the browser
6. Google Cloud 06 Check every single step Stage 2: Coding agent with manual approval You use Antigravity, Gemini CLI, Cursor, Windsurf, Copilot, Claude Code (...) for coding You manually approve every single tool call and carefully review every file edit and shell command
7. Google Cloud 07 You begin to loosen the reins. You allow file edits to happen automatically but still manually approve other tool calls You start learning the art of creating the agentic loop: Making the agent iterating through errors with automated feedback Selective trust Stage 3: The Guarded Collaborator
8. Google Cloud 08 Focus on end results Stage 4: The Hands-Off Director Trust goes up. You stop reading every line of code as it’s generated. You switch to "Full Screen" mode. You are no longer the writer; you are the reviewer.
9. Google Cloud 09 You disable all restrictions. Diffs scroll by faster than you can read, and you don't try to. Your cognitive load goes down, as your brain shifts from implementation details to high-level architectural guidance. Complete trust on the agent Stage 5: YOLO Mode
10. Google Cloud 010 You scale horizontally. You run 3 to 5 agent instances in parallel Your focus shifts from "how to code" to "how to manage a fleet of agents." Coordinate multiple agents Stage 6: The Orchestrator
11. Google Cloud 011 You are no longer writing software; you are guiding a system that writes software You define the what and the why, and the swarm handles the how entirely in the background A new way of working Stage 7: The Architect
12. How do we increase trust?
13. Google Cloud 13
14. Google Cloud Antigravity
15. Google Cloud 015 Your command center to manage multiple local agents in parallel. Group conversations into Projects, operate across multiple workspaces, and automate routine tasks with scheduled messages. Antigravity 2.0 Antigravity App
16. Google Cloud 016 Antigravity IDE Antigravity CLI Antigravity SDK
17. 017 Google Cloud Prompting Hooks Rules* Subagents Agent Skills MCP 01 03 02 04 05 06 What is in our toolbox? * aka AGENTS.md
18. Google Cloud 018 Understanding the model families Prompting techniques Extensibility (tools, skills, subagents, hooks, etc.) Workflow Agent type: CLI, IDE, orchestrator/manager Understanding the model families Prompting techniques Extensibility (tools, skills, subagents, hooks, etc.) Architecture (single agent, multi-agent,…) Frameworks (ADK, Antigravity SDK, Genkit,...) Using Agents to Build Things Building Agents to Do Things Skill 1 Skill 2 The Agentic Future
19. Google Cloud 019 Build an Arcade Game with Gemini and Antigravity Option 1 Option 2 Choose your own adventure Photo Restoration App with Genkit and Nano Banana
20. Build the future builders.google
21. 021 Thank you