For teams onboarding engineers onto agent development, this is a free, structured alternative to piecing together scattered blog posts and vendor docs: a five-day curriculum that moves from agent fundamentals through tooling, memory, security evaluation, and production deployment, built by Google's own agent-engineering team and now permanently available on-demand.
What happened
Google first announced the live run of its 5-Day AI Agents: Intensive Vibe Coding Course on April 27, 2026, scheduling it for June 15-19, 2026 as a follow-up to a November 2025 course that drew more than 1.5 million learners. i-Programmer reported July 1, 2026 that the course materials are now published as a free, self-paced Learn Guide on Kaggle. Each day pairs a whitepaper with a 45-60 minute podcast and hands-on codelabs, requiring roughly 1-2 hours per day; access requires a free Kaggle account and a free Google AI Studio account with an API key.
Technical context
The five days move through: agent fundamentals and "vibe coding" workflows, where natural language becomes the primary programming interface (Day 1); tool integration and agent-to-agent interoperability via the Model Context Protocol (Day 2); persistent "Agent Skills" organized around a SKILL.md file for managing long context (Day 3); security and evaluation, including a 7-pillar "Effective Trust" framework, sandboxing, defenses against hallucinated ("slopsquatting") package dependencies, and human-in-the-loop review (Day 4); and Spec-Driven Development for moving prototypes to governed, observable, cloud-deployed production systems (Day 5). Labs use Google's Antigravity 2.0 agentic IDE and CLI; the final capstone and cloud-deployment exercises are optional and may need a billing-enabled Google Cloud account.
For practitioners
The security and evaluation module is the most directly useful section for teams already shipping agents: it names concrete, current failure modes (slopsquatting, context rot, non-deterministic evaluation) rather than generic AI-safety advice, and pairs them with specific mitigations (sandboxing, an "Effective Trust" scoring model, OpenTelemetry-based trajectory evaluation).
What to watch
Whether Google keeps this Learn Guide updated as Antigravity and MCP tooling evolve, and whether teams report meaningfully faster agent-development ramp time after completing it.
Key Points
- 1Google's 5-Day AI Agents course, first run live in June 2026, is now a free self-paced Kaggle Learn Guide for building production AI agents.
- 2The curriculum covers agent tooling, MCP interoperability, persistent Agent Skills, a 7-pillar security and evaluation framework, and Spec-Driven deployment.
- 3For practitioner teams, it offers a structured, Google-vetted onboarding path to agent engineering with concrete security failure modes and fixes.
Scoring Rationale
A legitimate, technically substantive free educational resource from Google/Kaggle covering agent tooling, MCP, security evaluation, and production deployment - genuinely useful for practitioner upskilling, corroborated by Google's own blog and the Kaggle Learn Guide page. It is an education resource rather than product or research news, so it stays in the minor band despite good sourcing and real depth.
Sources
Public references used for this report.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems
