Resources
Links to keep going after the workshop.
Workshop
- AI-native Discord Free community for everyone who took the workshop — share what you built, ask questions when you get stuck, swap notes with others on the same path. The cheapest way to keep momentum after today.
- Slides The presentation deck — opening keynote, theory, quizzes, recap.
- NotebookLM Curated notebook with the workshop's source material — research deeper on your own time.
- Workshop Gem The Gemini Gem you used in step 3 — keep it bookmarked to re-spec future projects.
- Feedback form Two minutes of your honest feedback shape the next workshop.
Google products
- Antigravity IDE Google's AI IDE; the editor we used to build the site.
- Stitch AI design tool for sketching app and site layouts from a prompt.
- Gemini / Gemini Gems Consumer Gemini app; build a Gem to reuse the persona-interview prompt.
- Vertex AI Gemini API Reference for calling Gemini from Google Cloud, including current model names.
- Cloud Run docs Serverless container hosting; the platform we deployed to.
- Google Fonts The font CDN powering this site (Google Sans family).
Codelabs to explore
What is a codelab? Google Codelabs are free, step-by-step guided tutorials hosted at codelabs.developers.google.com. Each one is scoped (typically 30–90 minutes), written by Google engineers, and updated as products evolve. They are the highest-signal way to go deep on a single Google tool. Now is the right moment — you have credits, a working setup, and momentum.
Antigravity
- Getting Started with Google Antigravity The canonical first codelab for Antigravity. Beginner-friendly. Recommended if you want to retread the IDE setup with more depth.
- Build and Deploy to Google Cloud with Antigravity Closest to what you just did, formalized. A good sanity check that you understood the pattern.
- Authoring Google Antigravity Skills Write your own skills. The natural next step once you've tried one of the spec-driven frameworks below and want to bake your team's conventions in.
- Spec-Driven Development in Antigravity Maps closely to a structured spec → plan → implement loop, applied directly inside Antigravity.
MCP servers
- Getting Started with Google MCP Servers Google's official starting point for MCP. The natural next step from this workshop if you want to extend an agent with custom tools.
- Deploy a Secure MCP Server on Cloud Run Once you've built an MCP server, this is how you put it on the cloud, with auth.
Agents (ADK)
- Building AI Agents with ADK: The Foundation First codelab in the ADK series. Sets up a single conversational agent on Gemini and explains the core ADK loop.
- Tools Make an Agent: From Zero to Assistant with ADK Build a software-bug assistant by progressively adding tools, including MCP. Concrete and practical.
- From Prototypes to Agents with ADK Bridges the gap from "I have an idea" to "I have a deployable agent."
- Getting Started with MCP, ADK and A2A Combines all three concepts in one project — build a currency agent that talks to other agents over the A2A protocol.
Frameworks & protocols
Protocol
- Model Context Protocol Spec, SDKs, and example servers for the protocol that lets coding agents talk to external tools.
Spec-driven AI development frameworks
The workshop ran a lean version of the spec → plan → execute → review → finish workflow — one good prompt was enough for a one-evening site. For week-long features and team work, several open-source frameworks codify the same shape with installable skills, templates, and review gates. Pick one when a single prompt stops being enough.
- Superpowers A curated bundle of skills (TDD, planning, code review, debugging, sub-agents) that drops the named-stage workflow straight into Antigravity / Claude Code / Gemini CLI. The closest match to what you ran conceptually today.
- GitHub Spec Kit GitHub's official toolkit for spec-driven development with AI. Four stages — Specify → Plan → Tasks → Implement — backed by templates and a CLI. Tool-agnostic; works with Copilot, Claude, Gemini.
- BMAD-METHOD "Breakthrough Method of Agile AI-Driven Development." A more opinionated, agile-flavoured framework with role-based agents (PM, architect, dev, QA) handing artifacts to each other. Heavier than Superpowers, lighter than full enterprise process.
Alternative agentic coding tools
Antigravity and Gemini CLI aren't the only games in town. If your team already lives in a different editor, these slot into the same role.
- Aider Terminal-based AI pair programmer with strong git integration. Mature, model-agnostic (Claude, Gemini, GPT, local). Good if you live in the shell.
- Cline Popular open-source VS Code extension. Plan/Act mode mirrors the spec → execute discipline. Good for teams already standardized on VS Code.
Staying current with AI tooling
The hardest part of staying relevant in this market is the cadence — new tools ship weekly, half are noise, the other half matter for six months and then get displaced. Don't try to follow everyone. Pick two sources from the list below and read or watch them every week. Two sources consumed consistently beats twenty sources skimmed.
Our pick for everyone: Simon Willison for daily technical signal, plus Latent Space for weekly ecosystem context. About three hours per week and you'll be ahead of 90% of working developers.
Deep explainers — how things actually work
- Simon Willison A running, opinionated journal of working LLM techniques — MCP, tool use, prompt injection, model releases the day they ship. Highest signal-to-noise on this list. Start here.
- Andrej Karpathy (YouTube) — plus the "Software 2.0" / "Software 3.0" essays Long-form deep dives that set the conceptual frame. His "Software 3.0" thesis is what this workshop is built on.
- Anthropic engineering blog + their YouTube channel Primary source for Claude Code internals, MCP design, and prompt engineering practice. Worth checking when a new feature lands and you want the official "why."
Hands-on builds — show me the agent doing the thing
- IndyDevDan (YouTube) Practical builds with Claude Code, MCP servers, agent workflows. Sets up real workflows on screen — copy the pattern.
- Cole Medin (YouTube) Agentic systems, RAG, MCP, context engineering with Claude Code. Less polished than mainstream channels, more useful for actual builds.
- AI Jason (Jason Zhou, YouTube) Builds AI products end-to-end, often shipping the result. Good for "how does this become a real app I could charge for."
Ecosystem & business — what's actually working in production
- Latent Space (podcast — Swyx + Alessio Fanelli) Interviews with people building production AI tooling. Strong on MCP, agent infrastructure, what's working in production. The single best "ear to the ground."
- The Pragmatic Engineer (Gergely Orosz, newsletter) How engineering teams actually adopt AI tools and what the hiring market is doing. Crucial for the "make myself relevant" angle.
- AI Engineer (YouTube) — AI Engineer Summit + World's Fair talks Practitioner conference talks. Less hype, more "we ran this in production for a year — here's what broke."
Fast pulse — what shipped this week
- Theo — t3.gg (YouTube) Opinionated, fast. Covers Claude Code, Cursor, MCP regularly. Skim at 1.5×.
- Fireship (YouTube) "This thing exists now" overviews in 100 seconds. Low depth, high coverage. Use for awareness, not understanding.
- Matthew Berman (YouTube) Daily-ish AI news and demos. Breadth over depth — useful for not missing announcements.
Long-form thinking — for the train ride
- Lex Fridman Podcast (YouTube) — AI episodes Long interviews with AI researchers (Karpathy, Dario Amodei, etc.). Variable quality, but the AI episodes are usually worth the ride.
- Acquired — selected AI episodes Strategic and business framing of AI companies. The "where's the money" lens that pure-tech sources skip.
Specific topics worth bookmarking
- Model Context Protocol — Introduction If you only read one MCP doc, read this one before writing your second server.