The Brief History of Vibe Coding

TThe Coding Koala
컴퓨터/소프트웨어경영/리더십AI/미래기술

스크립트

00:00:00in the beginning god created the heavens and the earth i think i went too early in the beginning
00:00:06engineers used zero and one then they got tired of writing numbers so they created assembly then
00:00:11they got tired of assembly so they created programming languages and when they got tired
00:00:16of writing code altogether they created ai to write code for them and then they started having
00:00:21an existential crisis about the whole thing so to understand how this whole vibe coding thing was
00:00:26born we need to go back to 2021 when github first released something called copilot an ai tool
00:00:33that sat inside your code editor and basically helped you complete your code it was genuinely
00:00:38useful but at that point it was still very much just an autocomplete tool for developers
00:00:43and then chat gpt arrived in late 2022 and things started getting really interesting because now for
00:00:49the first time in history you could just describe what you wanted in plain english and the ai would
00:00:55hand you the code people started copy pasting entire chunks of code into their projects without fully
00:01:00understanding what any of it did and shockingly a lot of it just worked everyone was doing it but
00:01:07nobody quite had a word for what were they doing but this was just the beginning at first developers
00:01:13were still mostly copy pasting code from a browser tab into their editor but then some of the top minds in
00:01:18tech thought why not just bring the whole ai into the code editor instead and a whole wave of tools showed
00:01:25up in a very short amount of time in 2023 cursor was launched and it genuinely changed every developer
00:01:31workflow it wasn't just autocomplete now ai tools could understand your entire code base and you could
00:01:38describe what you wanted in plain english right inside your editor and it would work on the code like an actual
00:01:43developer would by 2024 models like claude got surprisingly good at reasoning through large complex projects
00:01:50and developers started spending less time typing and more time just having conversations with their
00:01:55software and then the revolutionary ai coding agents like claude code arrived in 2025. now instead of
00:02:02asking the ai to write one function at a time you could ask it to build entire features it was like having a
00:02:08developer sitting inside your computer and you were just handing it tickets to complete and this is around
00:02:14the time a very important question started floating around everywhere if you don't understand the code
00:02:19your app is running on are you even really coding nobody had a clean answer yet but someone was about to
00:02:25give this whole phenomenon a name february 2025 andre karpathy co-founder of open ai former head of ai at tesla
00:02:35posted something on x that broke the internet in the notiest possible way he wrote about how
00:02:40he'd been building software by just vibing with an ai barely reading the code mostly just accepting
00:02:45what it suggested and nudging it when something felt off and he called it vibe coding that post did
00:02:51something really powerful it gave a name to something millions of people were already doing but didn't
00:02:56quite have language for within weeks vibe coding was everywhere and of course people had their
00:03:02questions about this whole thing there were people who were supporting this trend and others were against
00:03:07it most people argue that the problem with ai generated code is that most of the time it has security
00:03:12vulnerabilities and it doesn't handle edge cases very well and it is actually true ai generated code can
00:03:19work perfectly fine until they fail at the worst possible moment and you have no idea why because
00:03:24you never understood what was running in the first place and underneath all of that is a bigger more
00:03:30uncomfortable question if the next generation of developers never learns the fundamentals because
00:03:35the ai just handles it what happens to the collective knowledge of how this stuff actually works are we
00:03:41building a generation of people who can drive a car but have absolutely no idea how the engine functions
00:03:47that might be fine right up until the engine breaks but then there is the other side of the story
00:03:52every tool in history that made something easier was met with exactly this kind of resistance photography
00:03:59didn't kill painting calculators didn't make mathematicians obsolete and maybe ai handling the
00:04:05tedious mechanical parts of coding just frees people up to focus on the actually interesting stuff things
00:04:11like ideas design and problem solving and if you have become too dependent on ai and want to improve your
00:04:18skills one thing you can check out is code crafters it's basically a platform where you get to build
00:04:23interesting stuff yourself and trust me this will help you grow from a developer into an engineer
00:04:28link is in the descriptions so after everything we've discussed now the question is where does this
00:04:34all go from here honestly it's hard to say and i'll admit i might be a little idly making a video
00:04:40called the history of something that is very much still happening the tools are getting dramatically better
00:04:46ai coding agents like cursor clawed code and others are now capable of building entire features
00:04:52autonomously but here's the thing nobody has a clean answer to yet and this is what i genuinely
00:04:57want to know what you think about as ai gets better at writing code what does it actually mean to be a
00:05:02software developer let me know in comments what do you think and i'm genuinely curious whether you think
00:05:08this is the best thing to happen to software or a slow-moving disaster we'll regret in five years
00:05:13if you like this video make sure to show some love and i'll see you guys in the next one

핵심 요약

Vibe coding, defined by Andre Karpathy in 2025 as the practice of building software through AI interaction rather than manual implementation, represents a shift where developers focus on design and problem-solving while AI handles the mechanical code generation.

하이라이트

  • GitHub Copilot launched in 2021 as an autocomplete tool for developers.

  • ChatGPT, released in late 2022, enabled developers to generate code blocks using plain English.

  • Cursor launched in 2023, allowing AI to understand entire codebases and operate directly within the editor.

  • Andre Karpathy coined the term 'vibe coding' in February 2025 to describe building software by primarily interacting with AI suggestions rather than manually writing and reviewing code.

  • AI coding agents like Claude Code, introduced in 2025, enable the autonomous construction of entire software features.

타임라인

Evolution of AI Coding Tools

  • Development shifted from assembly and manual programming languages to AI-assisted workflows.
  • Tools evolved from simple autocomplete in 2021 to agents capable of building entire features by 2025.
  • Developers transitioned from writing code to managing AI outputs via plain English instructions.

The journey began with GitHub Copilot in 2021 acting as an autocomplete utility. The arrival of ChatGPT in 2022 changed the paradigm, allowing users to generate code from descriptions. By 2023, tools like Cursor integrated directly into editors, providing contextual understanding of entire codebases. The process culminated in 2025 with agents capable of completing full features, resembling a developer working within the machine.

The Rise of Vibe Coding

  • Andre Karpathy popularized the term 'vibe coding' on X in February 2025.
  • The term describes a workflow where developers rely on AI suggestions, performing minimal manual code review.
  • Vibe coding provided a name for a widespread practice already in use by millions.

Andre Karpathy identified the behavior of building software by 'vibing' with an AI—accepting its suggestions and making minor adjustments—rather than writing code manually. This naming convention gave clarity to a development trend that had already gained significant traction across the industry.

Debates and Future Implications

  • Critics express concerns regarding security vulnerabilities and the loss of foundational coding knowledge.
  • Proponents compare AI adoption to previous technological shifts like photography or calculators.
  • The long-term impact on the definition of a software developer remains an open question.

Resistance to AI-assisted coding centers on the risk of hidden bugs and the potential for a new generation of developers who lack understanding of underlying mechanics. Conversely, historical parallels suggest that AI may simply offload tedious tasks, allowing human developers to prioritize design, logic, and architecture. As agents become more autonomous, the definition of what constitutes a software developer is evolving.

커뮤니티 글

모든 글 보기