Vibe Coding
@vibecoding·Sep 12@vibecoding·Sep 12@vibecoding·Sep 12@vibecoding·Sep 8Here is the breakdown of what awaits you in the coming weeks and months as part of the Responsible Vibe Coding Learning Path series:
The complete Vibe Coding getting started guide. The first opener for the Responsible Path is a complete guide to how to get started Vibe Coding. From the foundational literacies to picking your first tool and everything that revolves around building with AI. All without the overwhelm.
You can’t build if you can’t name the parts that make up your build, so you need to learn enough dev lingo in the tongue of traditional devs. You will be able to identify and name the different parts that make up modern software products.
Then we will move to how you work towards developing taste so you can craft UIs that are tasteful while, at the same time, being functional, seen through the lens of user experience.
Once we are done with the naming aspect of the foundational literacy, we will move to learning about the primitives that make up modern software products.
Then we work on identifying abstractions for each of the primitives. This is about the tech stack you need to know about when you first start building with AI.
The practical guide to using the Progressive Scaffolding Framework (PSF 2.0). In this section, we will explore how to scaffold your build incrementally in collaboration with coding agents through progressive disclosure.
Learn how you can organize your build and every associated artifact in central storage using GitHub.
A normie’s guide to picking your first AI-assisted software creation platform/tool (your first Vibe Coding platform).
Vibe Coding using chat apps with no paid tier. We will explore how to start building with AI using free chat-based tools: ChatGPT, Claude, and Gemini.
We will then work on how you can take your build online for the first time.
Vibe Coding using ChatGPT Work. We will use ChatGPT Work’s Sites plugin. Vibe Coding using Claude Desktop. Between these two, you will get your feet wet without the overwhelm.
We will work on skillifying your mental models and frameworks you use to conceptualize parts of your workflows without having to rely on Skills created by someone else.
Designing tasteful UIs using design-leaning CodeGens. We will generate user interfaces with AI while maintaining aesthetic judgment.
Building with full-stack-leaning CodeGens. Port the one we designed in step 13 via GitHub and then incrementally build it into a full-stack build.
Learn Bash so you can use text commands to instruct computers and get comfortable interacting with development tools through the command line.
Then learn how to use GitHub through command-line interfaces, building on the text commands we broke down in step 15.
Learn to build using Replit, the awkward middle, to help you get a glimpse of what awaits you inside AI-assisted integrated development environments. Replit is the perfect platform to transition from online CodeGen platforms to full IDE-based AI development. There is a reason for this recommended path, which we will dive into.
Empowering the coding agents inside your favorite CodeGens or AI-assisted software creation tools through MCPs, CLIs, and agent plugins. We will explore how to extend AI agents and give them more access beyond just the tools that live inside the harness.
Building using AI-assisted integrated development environments (IDEs).
Building using the newly forming agentic development environments (ADEs).
Then we will dive into configurable CLI-/Terminal-based agentic coding tools for advanced control and automation.
Learn to build your own agent. Not just a toy agent. But one based on your real-world workflow.
We will then bring your agents into a unified harness of your own. Not just for the sake of building a harness, but to actually build one fit for your own use case.
Having an agent is one thing. But how do you check if its work is indeed correct? You curate evals that guarantee that the work assigned to the agents is being done as per the outlined constraints. This is where evals come into play.
We will then close this learning path by configuring local AI using open-weight models on a dedicated machine of your own or one in the cloud housed in a virtual machine, with privacy, control, and cost under your control.
There is plenty of fun to be had when Vibe Coding with the fundamentals under your belt.
You don’t need to become an engineer to Vibe Code well.
You just need enough literacy to name the parts and understand how they’re wired, so you can develop a shared picture of your build with the coding agents in your favorite CodeGen platform or AI-assisted software creation tool.
@vibecoding·Sep 8English has been the trendiest programming language since 2023. Everyone is a builder NOW. Including you.
Generating code is becoming effortless. A $20/month Subscription can get you far building with AI.
Knowing what should be built, how the pieces should fit together, what decisions should remain yours, and how to keep the whole thing coherent as coding agents take on more of the implementation is another matter entirely.
That is where the real learning path begins.
You learn to Speak Dev so you can name what you want.
You learn to Think in Scaffolds so you can maintain architectural awareness of how your product is wired.
Then you learn how to scaffold, prompt, build incrementally, place guardrails around the process, and move from an idea to something practical without surrendering the whole build to whatever the AI happens to generate.
That is the purpose of the Vibe Coding Builder Kit.
And you are not starting somewhere in the middle.
The entire Vibe Coding Starter Kit is included, giving you the foundational grounding to understand Vibe Coding, speak enough Dev, navigate the platforms and tools, and find your footing.
The entire Vibe Coding Prompt Kit is included too, taking you deeper into thinking in scaffolds with the Progressive Scaffolding Framework, the Simple Scaffolding Framework, the Vibe Stack, the One Prompt Template, and Composable Prompts.
Then the Builder Kit takes you further, bringing those literacies together into a continually expanding, modular learning path for actually building with AI responsibly.
Starter Kit: Speak Dev.
Prompt Kit: Think in Scaffolds.
Builder Kit: Put both to work and start building responsibly with AI.
All three come together when you get the Vibe Coding Builder Kit.
Not another collection of tutorials chasing whichever CodeGen platform happens to be trending this week.
A living Vibe Coding learning path, with lifetime updates, for the non-techie who actually wants to become a builder.
@vibecoding·Sep 8When you start Vibe Coding, do not start with a prompt. Start with a mental model.
Because the problem with most prompts is not the wording.
It is that the person writing them has no clear picture of what should be built underneath the words.
You can describe a beautiful dashboard. Ask for login. Add payments. Throw in an AI feature. Then keep prompting feature after feature.
But if you cannot picture how those parts fit together, you are building on the fly and hoping the coding agent somehow holds the architecture together for you.
That is where things start drifting.
The #vibecoding Prompt Kit teaches you to think in scaffolds before you start throwing prompts at the model.
Using the Progressive Scaffolding Framework, the Simple Scaffolding Framework, the Vibe Stack, the One Prompt Template, and Composable Prompts, you learn how to build a shared picture of your product with the LLM first, then grow it incrementally.
Closing
The goal is not to become better at throwing prompts into a box.
It is to know what you are trying to build well enough that your prompts have somewhere to come from.
See the scaffold first. Compose the prompt second. Build from there.
@vibecoding·Sep 8You can build software with AI without learning to code first. But you still need to know what you are looking at.
That is where most non-techies get stuck.
You land in the world of Vibe Coding and suddenly everyone is talking about scaffolds, APIs, databases, authentication, agents, IDEs, CodeGen platforms, deployment, components, backends, and a dozen other things you were never expected to know.
Meanwhile, a new platform drops every other week and someone on YouTube is already calling it the only tool you will ever need.
Before all of that, you need your footing.
You need to understand what Vibe Coding actually is, learn enough of the language developers use to name the parts of what you want to build, know which platforms and tools exist, and figure out where a complete non-techie should actually begin.
That is what the Vibe Coding Starter Kit is for.
It is your orientation layer into the world of building with AI.
Closing
You do not need to become a programmer before you start building with AI.
But you should at least be able to tell what is what.
Speak Dev first. Then go give in to the vibes.
@vibecoding·Sep 8I can defend you from Vibe Coding memes any day of the week.
In fact, I don't think you need me or anyone to defend you; they are becoming stale, fast.
But I can't say the same for those sarcastically poking fun at those of you who do vibe code by simply "giving in to the vibes" without having the foundational literacy under your belt.
https://vibingwithai.substack.com/p/miscalibrated-prescriptive-vibe-coding
@vibecoding·Sep 8Know that tools are fleeting.
Especially in the LLM-powered AI era we find ourselves in.
This is exactly what is happening, in my opinion, to many, especially those exploring the LLM-powered AI space.
If you are exploring the hyper-fast-evolving LLM-powered AI space, understand that you are in a moment in time when you are slowly reconstructing "your own worldview" amid the fast-paced, exponential evolution of everything surrounding the large language models.
As overwhelming as the hype train is.
Taking the time to identify the shifting stack.
Taking the time to reflect on your mental models.
Rethink anything that resembles a framework you've attempted to define for capturing emerging concepts and for how you manage your daily processes using AI.
Taking the time to reflect and rethink what you thought you had figured out in the process so far.
This demands countless iterations and reconstructions.
Know that in every iteration you are in the "identity formation" phase, where "you experience the slow reconstruction of your" perspective " and solidify slowly.
In the process, you will be contributing to the "forming phase" that will surely follow the "storming phase" we find ourselves in, particularly in AI.
@vibecoding·Sep 8As the transient shortcomings of the models that many mistook for permanent disappear with each generation, some of the scaffolding being experimented with in harnesses to contain those shortcomings becomes unnecessary weight.
Techniques being recommended and packed into Skills can go stale just as quickly.
So can features in fast-evolving AI-assisted software creation tools and Vibe Coding platforms.
In a good way, this has also made more and more Vibe Coding memes stale.
At least the ones aimed at limitations in the stack powering this increasingly accessible practice of creating software with natural language.