Vibe Coding Part 1
Vibe Coding and the Skynet Prototype: Is This the Future of Coding (and Humanity)? (Part 1)
Introduction
Alright, buckle up, folks, because we’re about to dive into some seriously futuristic stuff! This isn’t just another tech blog; it’s an experiment to see if this “Vibe Coding” thing could actually work.
Vibe Coding is where we ditch the old-school, solo coding grind and throw a party where human developers and AI agents team up!
It’s all about letting AI handle the boring stuff, like those repetitive compliance checks, so we humans can focus on the cool, creative coding that really matters. But beyond just making our lives easier, we’re also kinda wondering: what does this mean for developer jobs?
For the future of software development? Heck, for humanity itself?!
In this Part 1, we’ll focus on setting the stage: explaining Vibe Coding and diving into the Skynet prototype’s tech. In Part 2, we’ll get into the nitty-gritty of how to structure and govern these AI agents. And finally, in Part 3, we’ll put on our futurist hats and try to predict the grand finale: the implications for developers and humanity.
The Skynet Prototype
Enter Skynet (yeah, we know it sounds intense, but trust us, it’s the good kind of AI!). This prototype is our attempt to build an AI orchestra, where “Minions” (our AI agents) monitor, audit, and generally keep an eye on things to make sure our software is top-notch.
Technology Stack: Why We Chose Our AI Dream Team (And Our Sanity Check)
My journey into the world of AI protocols and Agentic Frameworks felt less like a smooth ascent and more like repeatedly tripping up the learning curve. Just when I thought I had a handle on things, the goalposts would merrily relocate. So, after enough research to warrant a PhD in “Keeping Up,” I opted for Google’s framework. This video pretty much sums up why my brain needed a break and a reliable solution.
- Coding Assistant: Gemini Pro 2.5: This thing is basically a coding rockstar! I picked it because it’s a genius at spitting out clean, efficient code for our Minions, understands the coding vibe, and helps us code faster.
- Agent Development: Google Agent Development Kit (ADK): Think of this as our Minion-building toolkit. It’s got everything we need to create and manage our AI crew.
- Communication Protocols: A2A and MCP: Okay, this is where it gets a bit philosophical.
These protocols are the secret languages our AI agents use to chat with each other (A2A) and with other tools (MCP).
But they’re not just any languages – they’re *open* protocols.
Why’s that important?
Well, if we ever get into a situation where our AI creations start getting a little too big for their digital britches (you know, the whole “evil Skynet” scenario), having them communicate in open, well-understood ways could be our way of…
- Understanding them: We can analyze their communications to figure out what they’re up to.
- Controlling them: It gives us a potential “hook” to intervene or limit their actions.
- Preventing vendor lock-in: Open protocols prevent us from being completely reliant on any one company’s proprietary technology, which could become a single point of failure or control. It’s like making sure everyone speaks a language we can all understand, even if they start having ideas of their own! So, yeah, open protocols are a big deal – they’re our way of being both innovative and (hopefully) not accidentally creating our own digital overlords.
Environment Setup: Let’s Get This Coding Party Started!
Alright, here’s how to get your coding dojo ready:
-
Prerequisites: Make sure you’ve got Python, Git, Docker (or Podman), AWS CLI, Google Cloud CLI, Terraform, and a Google AI API Key installed.
-
Repository Cloning: Grab the code and jump into the directory:
git clone ... cd ...
-
Virtual Environment: Create a safe space for your Python stuff:
python3 -m venv .venv source .venv/bin/activate
-
Dependencies: Install the goodies:
pip install -r requirements.txt
-
Environment Config: Create those
.env
files to keep your API key and agent details safe:minions/basic_audit_minion/.env minions/basic_execution_minion/.env
Toss in your secrets:
GOOGLE_API_KEY=YOUR_API_KEY ADK_AGENT_NAME=agent_name ADK_AGENT_MODEL=gemini-pro
-
Terraform Backend (Optional): Set up remote storage for your Terraform state (if you’re into that):
infra/backend.tf terraform init
-
PYTHONPATH (Optional): Tell Python where to find your code:
export PYTHONPATH="${PYTHONPATH}:$(pwd)"
Or just tweak your
activate
script. -
Dockerfile: Double-check that Dockerfile is hanging out in your project.
-
Terraform Infrastructure: Give your Terraform code a once-over in
infra/
.
Running the Prototype: Let the Minions Loose!
-
Local Execution: Fire up the ADK API server, then unleash the Skynet client:
adk api_server minions --port 8000 --log_level debug python skynet_stub/main.py
-
Deployed Execution (Fargate): Build and push your Docker image to the cloud, tell Terraform to do its thing, and then make sure everything’s running smoothly:
podman build --platform linux/amd64 -t your-image-name:tag -f Dockerfile . podman login -u AWS -p $(aws ecr get-login-password --region your-region) your-aws-account-id.dkr.ecr.your-region.amazonaws.com podman push [your-aws-account-id.dkr.ecr.your-region.amazonaws.com/your-image-name:tag](https://your-aws-account-id.dkr.ecr.your-region.amazonaws.com/your-image-name:tag) terraform init -upgrade terraform apply -var="google_api_key_value=YOUR_API_KEY" -var="image_tag=your-image-tag"
Then, grab your Fargate task’s public IP, update
AGENT_BASE_URL
inskynet_stub/main.py
, open up the security group gates, and runpython skynet_stub/main.py
.
So there you have it! With this setup, you’re all set to build some seriously cool AI agents that’ll whip your software into shape. But more importantly, you’re joining us in exploring a whole new world of coding – a world where AI is our partner, not our replacement (hopefully!).