VIBE CODING is Eating the World...

Okay, let's break down what's happening with this whole "Vibe Coding" trend based on the video!

There's this new buzzword, "Vibe Coding," coined or popularized by Andrej Karpathy, which basically means letting AI do most of the heavy lifting when writing code. The idea is you guide the AI, sometimes even just by speaking, rather than painstakingly typing out every line yourself. This concept is causing a stir, with some people jumping on board enthusiastically and others pushing back hard, sparking a pretty interesting debate about the future of software development.

Here are the key points and details from the video:

• What is Vibe Coding? It's a new way of coding where you primarily rely on AI to generate the code for you. You kind of "go with the flow" of what the AI produces, guiding it rather than writing everything from scratch yourself. The ideal vision even involves using voice commands instead of typing.
• It Started with Andrej Karpathy: He's credited with popularizing the term and kicking off the discussion.
• It's Polarizing: People have taken extreme stances – either fully embracing it or being strongly against it.
• Real-World Examples & Success:
  • Peter Levels, creator of the FlyPeter game, is highlighted as a successful Vibe Coder. His AI-built game is reportedly making around $87,000 a month from sponsorships, showing that you can build substantial projects this way.
  • He built FlyPeter using AI for most of the code and graphics, including features like multiplayer (though he later removed this based on feedback).
  • He mentioned using tools like Claude 3.5 Sonnet Max via Cursor.
• Building a Resource Hub: The presenter is creating a free resource page on his website (natural20.com/vibe-coding) to collect tutorials, games, and tools related to Vibe Coding, aiming to be a central place for people interested in getting started. He's even using AI agents (like Manus AI) to research and gather information for this site.
• Personal Vibe Coding Experiments: The presenter has tried Vibe Coding himself, creating small projects like a hand-tracking app that plays music and an idle incremental game.
  • He notes his incremental game took about 2.5 hours of focused time just using natural language prompts via AI (Mid Journey for art, Cursor for code), without writing any code himself initially.
  • He experienced issues where the AI (Claude) would sometimes mess things up, stressing the importance of using checkpoints or version control like GitHub.
• The Vibe Coding Game Jam: Peter Levels is organizing a game jam for 2025 where people can submit games built with AI assistance.
  • Amazingly, Andrej Karpathy (the term's popularizer) and legendary game developer John Carmack (creator of Doom) are involved, providing input or judging.
  • Carmack's involvement is seen as a significant sign of interest from a major figure in the industry.
  • The jam requirements were adjusted (like removing the multiplayer requirement) to make it more accessible.
  • A key point for the jam is that prior coding knowledge isn't required – just a willingness to learn and be flexible.
• More Cool Projects Being Built: People are submitting impressive games to the jam, including a first-person game with great physics, clones of popular games like Fortnite/Minecraft and Bomberman, a Hexen/Heretic style game, and a Vampire Survivors-like game (Mech and Cheese, which allegedly impressed Elon Musk).
• Tools and How to Start:
  • Examples of tools used include Cursor, Grok, Replit Agent, and Claude.ai.
  • The easiest way to start, according to the presenter, is using Claude.ai (Anthropics' AI). He recommends Claude 3 Sonnet or Opus, especially with Extended mode.
  • Claude has a great feature called "artifacts" that lets you see and run the code (like a simple game) directly within the browser window, which makes testing easy.
  • He demonstrated building a simple 3D racing game ("Claude Cart") in 10-20 minutes using Claude, showing how to prompt it, run the code, and even troubleshoot errors by pasting them back into the AI.
• The Big Debate: Traditional vs. Vibe Coding:
  • Critiques (Traditional Developers): Many experienced developers dismiss Vibe Coding, saying it's not a serious way to code, results in messy code, and people without real skills are creating low-quality work. They feel their jobs are safe because AI can't yet produce the sophisticated, high-quality code needed for complex projects.
  • Peter Levels' Defense: He argues that some critics are acting like "gatekeepers" who feel threatened. He believes coding should be accessible to everyone and that this allows exploring new creative areas.
  • Jonathan Blow's Perspective: Acknowledges that AI generating some code is cool, but insists it's nowhere near the quality needed for a real game. He stresses that getting things on the screen is easy; making a game good is incredibly hard and requires deep skill. He doesn't see his view as gatekeeping, but as recognizing the complexity and skill involved in professional game development.
• The Digital Camera Analogy: The presenter draws a strong parallel between Vibe Coding and the transition from film to digital cameras in the late 90s.
  • Initially, professional photographers dismissed digital cameras as inferior and requiring less skill (compared to mastering film types, exposure, and darkroom techniques).
  • However, digital cameras rapidly improved, became accessible (especially on phones), cratered the film market, and dramatically increased the total number of photos taken globally because anyone could point and shoot.
  • Professional photographers didn't disappear but adapted and integrated digital tools.
• Predictions for the Future of Coding:
  • Based on the analogy, the presenter believes AI coding will follow a similar path: starting less skilled than pros but improving rapidly.
  • It will significantly reduce the need for manual typing, freeing up experienced developers to focus on higher-level tasks and potentially 10x their output.
  • Crucially, it removes the skill barrier to entry, allowing a massive number of new people (especially "time rich" individuals like kids and students) to start creating software and games.
  • This will lead to an explosion in the amount of code produced worldwide, particularly for simple, custom, or niche applications (personal tools, home automation, etc.) that we haven't even imagined yet – you could potentially conjure software into existence just by speaking a few sentences.
  • A future "creation ecosystem" or app store is envisioned where complex tasks like deployment, networking, and security are abstracted away, letting creators just focus on their idea and publish.
  • Great software engineering talent will remain valuable, just like great musicians or writers are still valued despite anyone being able to publish online.
  • However, those with only average skills might find it harder as basic coding tasks become automated and accessible to everyone.
  • The overall cost of software/code is likely to decrease, with more free and custom options available.
• Actionable Takeaway: Don't "sleep on" Vibe Coding. Even if you've never coded, try making a simple game or tool with AI (like using Claude.ai). It's surprisingly easy and a great way to see the potential. Keep practicing and stay updated, as the tools and processes will only get simpler (like one-click deployment eventually).
• Conclusion on the Debate: The presenter thinks both sides of the debate are simultaneously right: software engineering is a hard, skilled discipline where experienced professionals are invaluable, and the AI coding wave is real, massive, and will completely change the landscape by lowering the barrier to entry and increasing the sheer volume of code created for everyday uses.