How To Learn To Code In 2026 — Transcript
Full transcript
- 0:00Learning to code is so different in
- 0:012026. So, in this video, I shall explain
- 0:04to you how to learn to code in 2026.
- 0:07Starting off with what coding actually
- 0:08looks like today, it looks very
- 0:09different. Then I'll explain what you
- 0:10should be learning that is still
- 0:11relevant right now and for the future.
- 0:13And finally, some tips on how to
- 0:15supercharge your learning. Oh, by the
- 0:16way, just in case you have any doubt
- 0:17about whether you should still learn how
- 0:19to code in 2026, the answer is yes.
- 0:21There are still so many opportunities
- 0:23that are only available to people who
- 0:25know how to code. Plus, learning to code
- 0:27in 2026 is so much easier and faster
- 0:30than it used to be. So, it's kind of a
- 0:31no-brainer if you want to build stuff.
- 0:33Now, with that out of the way, in case
- 0:34you don't know me, hi. My name is Tina.
- 0:36I used to be a data scientist at Meta
- 0:38and I have been coding for 10 years.
- 0:40Without further ado, let's go. A portion
- 0:43of this video is sponsored by Warp. On
- 0:45February 3rd, 2025, Andrej Karpathy, the
- 0:47previous director of AI at Tesla and the
- 0:49founding member of OpenAI, made a viral
- 0:52tweet defining a new term called vibe
- 0:54coding. He explains, "There's a new kind
- 0:55of coding I call vibe coding where you
- 0:57fully give in to the vibes, embrace
- 0:58exponentials, and I forget that the code
- 1:00even exists. It's not too bad for
- 1:02throwaway weekend projects, but still
- 1:04quite amusing." Then, almost exactly a
- 1:06year later, on February 5th, 2026, he
- 1:09made another post. "Today, 1 year later,
- 1:11programming via LLM agents is
- 1:12increasingly becoming a default workflow
- 1:14for professionals, except with more
- 1:15oversight and scrutiny." Now, to
- 1:17differentiate this much more
- 1:18sophisticated integral workflow from
- 1:21vibe coding, he said his current
- 1:22favorite term for it is agentic
- 1:24engineering. You have your terminal/IDE
- 1:26open where you have multiple coding
- 1:28agents coding furiously away for you and
- 1:30you're like, "Build this. Fix this. No,
- 1:32don't do that. Do this instead."
- 1:34Orchestrating your AI agents like a
- 1:36puppet master. But, do not be fooled. It
- 1:39may look like you're just speaking
- 1:40English to the AI agents and just like
- 1:42orchestrating them, but you need to be
- 1:44an experienced software engineer to even
- 1:46do that. You need to know how to
- 1:47structure what it is that you want to
- 1:48build in a way that the AI agents can
- 1:50understand, be able to provide context
- 1:51and documentation. And in order to
- 1:53correct the AI and tell it that it's
- 1:55wrong, you need to actually know what is
- 1:56right. To be a good manager, you need to
- 1:58know what it is that you're managing.
- 2:00So, that is why if you want to learn how
- 2:02to code in 2026, you still need to know
- 2:04the fundamentals of coding and
- 2:05engineering. But, instead of focusing so
- 2:07much on the details of the code
- 2:09implementation itself, it's about
- 2:11understanding big picture software
- 2:12design, systems, and architecture. And
- 2:15particular emphasis on security and
- 2:17privacy, because these are the things
- 2:18that AI coding agents tend to struggle
- 2:20the most with. Then, you can actually
- 2:21move on to take advantage of the AI
- 2:23coding tools that we have available
- 2:25today and really do agentic engineering.
- 2:27So, should I tell you what it is that
- 2:28you need to learn? All right, I'm going
- 2:29to list out the topics in a roughly
- 2:31chronological order that I think you
- 2:32should be learning, assuming you don't
- 2:34know anything. And also suggest some
- 2:35resources on these topics.
- 2:38Okay, so first thing that you want to
- 2:40learn is the basics of coding itself.
- 2:42Stuff like variables, types, [music] if
- 2:44statements, loops, object-oriented
- 2:46programming, and APIs. Also, I'm using
- 2:49Python as the demo language to learn,
- 2:51which is particularly good if you want
- 2:52to be building AI agents and doing data
- 2:54stuff. But, if you want to be doing
- 2:55something more like web stuff, like
- 2:56developing web apps, you might want to
- 2:58start off with a language like
- 2:58JavaScript. It's okay though, because
- 3:00I'm going to give you guys a prompt at
- 3:01the end of the section that you can use
- 3:02to customize this learning plan specific
- 3:04to the language of what you want to be
- 3:06learning. Okay, great. So, these are the
- 3:07basic basics. Like, you really need to
- 3:09know these, otherwise you wouldn't even
- 3:11be able to read the code that the AI
- 3:13generates.
- 3:13>> That sign can't stop me, because I can't
- 3:16read.
- 3:17>> And I'm going to put on screen now some
- 3:18of my favorite resources on this topic.
- 3:20I personally still recommend going with
- 3:21these human-made resources, uh because
- 3:24we know that they're really good.
- 3:25They've been vetted by thousands of
- 3:26engineers. Besides, all AI-generated
- 3:28code is based upon code that was written
- 3:30by human developers for decades. Next
- 3:32up, you need to learn about software
- 3:33architecture. What that means is how is
- 3:35it that projects are structured? How do
- 3:37you choose a tech stack? The system
- 3:39design. What kind of APIs should you be
- 3:40using? And how the data flows through
- 3:42your software, and where is it being
- 3:43stored? What kind of databases are you
- 3:45using? And testing. Testing is so
- 3:46important, especially if you're using AI
- 3:48coding agents. You need to understand
- 3:50how to write test cases and how to test
- 3:51different of your software. And finally,
- 3:53how does deployment work? Where are you
- 3:54going to be hosting your software? This
- 3:56is like the bread and butter of software
- 3:58engineering and you really need to
- 4:00understand these high-level things to be
- 4:02able to direct your AI to be building
- 4:05the things that I should be building and
- 4:06redirecting it when it's doing something
- 4:07wrong. You need to be able to
- 4:09conceptualize what the end product
- 4:10should look like based upon the
- 4:12requirements that you have. For example,
- 4:14at work, you might want to build a
- 4:15software um that stores a massive amount
- 4:17of information and have lots of people
- 4:18querying it all the time. So, then you
- 4:20need to be thinking, what sort of
- 4:21database should I be using? That is
- 4:22cheap, safe, and fast. And how do I
- 4:24design the user interface that allows
- 4:26for multiple people querying at the same
- 4:27time? These are all like design
- 4:28questions that you need to be able to
- 4:30understand. I'm going to put on screen
- 4:31now some resources that you can check
- 4:33out to learn about these topics. Next up
- 4:34is version control and GitHub. By
- 4:36version control, we mean a system that's
- 4:38able to track the changes in your files
- 4:40and code over time. So, you can track
- 4:42the history, revert back if you need to,
- 4:44and collaborate simultaneously with
- 4:46other people as well. There are actually
- 4:47a lot of different types of version
- 4:48control software out there. Um the most
- 4:50popular one is called Git. So, people
- 4:52would use Git in order to do version
- 4:53control and then put their code on
- 4:55something called GitHub. GitHub is the
- 4:57industry standard for displaying your
- 4:59code, sharing it, and collaborating with
- 5:01other people. This is especially
- 5:02important if you're intending to do
- 5:03agentic engineering because you want to
- 5:04be tracking the changes that your AI
- 5:06coding agent is doing, right? And in
- 5:07case it makes a mistake, you can revert
- 5:09it and redirect it without like losing
- 5:11your entire code base. Which by the way,
- 5:13this happened to a lot of vibe coders
- 5:15who did not know the fundamentals of
- 5:16coding and engineering. I'll put on
- 5:18screen now some resources for you to
- 5:19learn about version control and GitHub.
- 5:21All right, next up is security and
- 5:23privacy. So, this is a topic that was
- 5:25not particularly important when somebody
- 5:28is learning how to code um like a few
- 5:30years back. Like obviously security and
- 5:32privacy is important, right? But it's
- 5:33sort of something that just gets taught
- 5:34like within the scope of other things.
- 5:37Like how does authentication work and
- 5:38how do you like integrate authentication
- 5:40into your apps? But the reason I'm
- 5:41separating this out as its own like
- 5:43little topic here is because I I it's
- 5:45particularly important in this day and
- 5:47age to learn about security and privacy
- 5:49because it's usually one of the blind
- 5:50spots of AI coding agents. So you as the
- 5:52human can't just kind of like hope that
- 5:54the AI agent is just going to take care
- 5:56of it. You do need to be quite explicit
- 5:57in incorporating these principles. Don't
- 5:59worry, this is honestly not like a
- 6:00really big topic. It's just really
- 6:02important to know. And as they say, I
- 6:05think, paranoia makes a good engineer,
- 6:07maybe. I'm going to put on screen now a
- 6:09few different resources that I suggest
- 6:10you take a look at and learn about
- 6:12privacy and security. All right, so at
- 6:13this point you would have covered pretty
- 6:15much the fundamentals of coding and
- 6:17software engineering now. You have now
- 6:19earned your stripes to be able to
- 6:21harness the power of agentic engineering
- 6:25by using AI coding agents. Yay! But with
- 6:28that being said, I do want to throw in a
- 6:30bonus little topic, which is debatable
- 6:33whether it's actually like necessary
- 6:35necessary for you to know this topic,
- 6:37but I personally think it's pretty
- 6:38important and you're probably going to
- 6:39have to learn about it anyway. And that
- 6:40is microservices, otherwise known as the
- 6:42concept of containerization. This is a
- 6:44software deployment method that packages
- 6:46an application's code together with all
- 6:49of the files, libraries, and
- 6:50dependencies that it needs in order to
- 6:51run that piece of code. So it's sort of
- 6:53like little packages, kind of little
- 6:54containers of pieces of software. The
- 6:56purpose of this is to isolate the
- 6:58application from its surroundings so
- 6:59it's not interacting with the rest of
- 7:01your computer, um and it's able to run
- 7:03consistently across different computer
- 7:04environments, like different computers,
- 7:06different computer systems, different
- 7:07types of clouds. This is usually more of
- 7:09an advanced topic and people usually
- 7:10learn about this when they're building
- 7:12stuff or like on the job. But the reason
- 7:13why I'm putting this here as like an
- 7:15optional but maybe you should learn
- 7:17about this kind of topic is as another
- 7:19precaution if you're using AI coding
- 7:21agents. You can imagine if you're
- 7:22letting your coding agent or even like
- 7:23multiple coding agents just do whatever
- 7:25it wants, right? It can end up messing
- 7:26up your development environment entirely
- 7:28or like building things in a way that
- 7:30cannot be properly deployed and used by
- 7:32other people. And you'll be just left
- 7:33like really frustrated cuz you can't
- 7:35figure out why it's happening. So if you
- 7:37do take the precaution to containerize,
- 7:39you're going to end up with something
- 7:39more reliable. And if your AI agent does
- 7:42run amok and starts doing crazy things,
- 7:45it's going to be in its own little
- 7:47environment doing it. So, worst case
- 7:48scenario, just shut off that
- 7:49environment. Let me know in the comments
- 7:51of these topics I have just covered, how
- 7:53many of these topics do you already
- 7:54know? Which ones do you still need to
- 7:56learn? Curious where you guys at on the
- 7:57learning journey. As promised, I'm also
- 7:59going to put a prompt on screen. Also,
- 8:01I'll probably link in the description,
- 8:02too. You what you can put into your
- 8:03favorite AI chatbot, ChatGPT, Gemini,
- 8:05Claude, Quen, Deep Seek, whatever you
- 8:07want, and it will help adapt this
- 8:09learning plan and resource list to fit
- 8:11the goals of your learning and also help
- 8:13you determine which coding language
- 8:14makes the most sense for you to learn.
- 8:16Yay, amazing. You have now earned your
- 8:18stripes to be able to tackle this new
- 8:21world, this brave new world of agentic
- 8:23engineering.
- 8:25Now, agentic engineering itself, gosh,
- 8:28it does agentic engineering just not
- 8:29sound as cool as vibe coding. It's just
- 8:31not as catchy.
- 8:33>> [gasps]
- 8:33>> It's fine. Agentic engineering, okay.
- 8:36Agentic engineering itself is also a
- 8:38skill. It involves how to set up your
- 8:40project and how to set up your AI coding
- 8:41agents, how to communicate what it is
- 8:43that you wanted to build to your coding
- 8:45agents, and how to monitor them
- 8:46correctly among many other things. For
- 8:48agentic engineering, there are different
- 8:50options for which coding agents to use
- 8:52and how to use them and orchestrate them
- 8:53together. Me personally, I really like
- 8:55using Warp, who is also the sponsor of
- 8:57this portion of the video. Because it's
- 8:59a pro developer tool that covers the
- 9:01full spectrum of AI coding. Getting
- 9:03started is so simple. I built this
- 9:05entire project using Warp. But, as it
- 9:07scaled, I needed to move faster, and
- 9:09that's where the magic begins. Instead
- 9:11of tackling one thing at a time, I
- 9:12deployed multiple coding agents
- 9:14simultaneously. Here's one that's adding
- 9:16a back-end feature, one on
- 9:17documentation, and one that's writing
- 9:18tests. This is all powered by Oz, which
- 9:21is a cloud-based agent orchestration
- 9:23platform built on top of Warp. The Oz
- 9:24dashboard gives you a real-time view of
- 9:27all your agents. At a glance, you can
- 9:28see what they're working on, their
- 9:29status, and any issues that they may
- 9:31have. And because it's cloud-based, each
- 9:33agent runs in its own isolated
- 9:35environment, meaning that they won't
- 9:36interfere with each other and my poor
- 9:38MacBook Air can stay alive. That is not
- 9:41all. You can actually incorporate Oz
- 9:43within your project itself. For example,
- 9:45this is a scheduled agent that updates
- 9:46my microsite every day. It goes in
- 9:47there, takes the AI news of that day,
- 9:49which is what I care about, and
- 9:50summarizes it and displays that
- 9:51information for me. With Oz, you can
- 9:53also convert skills into AI coding
- 9:55agents, integrate them directly with
- 9:56Slack and Linear, and tap into a ton of
- 9:58other features that support team
- 9:59workflows. 700,000 engineers are already
- 10:02using it and 97% of code diffs get
- 10:04accepted. Which, honestly, is pretty
- 10:06impressive and tracks from my
- 10:07experience, too. So, if you want to get
- 10:09started, you can get started for free at
- 10:10the link linked below. Thank you so much
- 10:11to sponsoring this portion of the video.
- 10:13Now, back to the video. Unfortunately,
- 10:15there are not that many good resources
- 10:17on agentic engineering, probably because
- 10:18it's such a new topic. But, I'm going to
- 10:20put on screen now a couple videos that
- 10:22I've done and some other resources as
- 10:23well. If you guys want, I can also make
- 10:25a video that specifically covers how to
- 10:27do agentic engineering. You know what's
- 10:29crazy is that to learn all of these
- 10:31topics that we just discussed prior to
- 10:33AI, it would have taken you years. go to
- 10:35school for this and then work multiple
- 10:37years to get to this level. Or, even if
- 10:38they're attending like a fast-track boot
- 10:40camp, we're still talking like at least
- 10:4210 to 12 months completely full-time.
- 10:44But, with good learning techniques and
- 10:45with AI to help with your learning, you
- 10:47can reduce the learning time down by so
- 10:49much. I think that if you really focus
- 10:51and if say you're doing it like
- 10:52completely full-time, I would estimate
- 10:54maybe like 6 months to be able to learn
- 10:56all these topics very solidly. So, in
- 10:58the next section, I'm actually going to
- 10:59share some of my favorite learning
- 11:00techniques that can help you learn
- 11:01faster. For those of you who have
- 11:03watched my channel for a while, you
- 11:04probably also know that learning things
- 11:06fast is like my favorite activity. So,
- 11:08very exciting. Let's now talk about tips
- 11:10to supercharge your learning.
- 11:12I'm going to divide these tips up into
- 11:13two different categories. The first is
- 11:15how to use your resources better, like
- 11:16the lectures, the books. And the second
- 11:18is how to use projects for learning and
- 11:20specifically how to use a AI coding
- 11:21agent to help you build a project and
- 11:24learn at the same time. Two birds with
- 11:26one stone. Okay, let's first talk about
- 11:27how to use resources. If your resource
- 11:29happens to be like a video or like some
- 11:31text or like readies and stuff like
- 11:33that, I would really recommend using
- 11:35something like Notebook LM and put all
- 11:37that information in there and ask it to
- 11:39summarize things for you and also to
- 11:41generate questions that will guide you
- 11:43along the work. This is called primate.
- 11:45It can really speed up your learning
- 11:46process because you understand what are
- 11:49the things that you need to be learning
- 11:50and how they fit together with each
- 11:51other. As you're going through the
- 11:52lectures, the text, and like code
- 11:54snippets and and whatever, um if there's
- 11:56anything you don't understand, you can
- 11:57also just copy paste that and put that
- 12:00into AI to help you explain what's going
- 12:02on. When it comes to code and logic, at
- 12:04least of the filming right now, the best
- 12:06AI model to do this would be Claude. I
- 12:08personally default using Claude's on it
- 12:09for anything that happens to do with
- 12:11code. But if you don't want to pay for
- 12:12it and you run out of the free trial,
- 12:13honestly most modern AI models would be
- 12:16able to help explain code. You can even
- 12:18use the free ones, the open source ones,
- 12:19like Gwen, Kimmy, or Deep Seek. Another
- 12:21way that I really like using AI models
- 12:23to help me learn is by taking a concept
- 12:25and ask it to give me examples and
- 12:26analogies because this will really allow
- 12:28it to solidify my mind to understand the
- 12:30concept. Like for example, if you're
- 12:32learning about object-oriented
- 12:33programming, you can ask it explain
- 12:35object-oriented programming to me using
- 12:36analogy and provide examples. Much
- 12:38clearer, I think. And finally, often
- 12:40times the best way to learn how to code
- 12:42is to take the code and the snippets
- 12:44from the resources that you have, like
- 12:46other people's code, and then ask AI to
- 12:48help you understand the code that's
- 12:50there, to explain to you what the code
- 12:52is doing. This is so powerful and I'm
- 12:55kind of salty that I did not have this
- 12:58option when I was learning how to code.
- 13:01To do this, you can also just like copy
- 13:02paste or like upload the code into an AI
- 13:04model, or this could be a good
- 13:05opportunity for you to set up an AI
- 13:07coding agent. Just give it the code and
- 13:09it would explain to you the structure of
- 13:10the code, how the pieces are fitting
- 13:11together, um what different lines of
- 13:13code mean, line by line if you want. And
- 13:15speaking of projects, this is a good
- 13:16transition to the second category of
- 13:19tips that I have, which is for projects.
- 13:21First of all, I want to just emphasize
- 13:22that projects are the best way for you
- 13:25to learn how to code, hands down. You
- 13:26can watch like so many lectures, read so
- 13:29many things and whatever, and you think
- 13:30that you know these things. Yeah, but
- 13:31you don't unless you actually start
- 13:33building with it. And the way you do
- 13:34that is by doing projects. So, I
- 13:36encourage you to do projects
- 13:37continuously as you're learning. Like if
- 13:39you're learning about how to use an API,
- 13:40actually make a project in which you use
- 13:42an API or multiple APIs. If you're
- 13:44learning about testing, write some
- 13:46tests. Try to understand the differences
- 13:47between different tech stacks. Try them
- 13:49out yourself. Especially with the help
- 13:51of AI coding agents, it is so easy to do
- 13:53this. You can ask the coding agent
- 13:55something like build me a music storage
- 13:57software with like XYZ specs built with
- 14:00a Django back-end and Next.js front-end
- 14:03and store the data in a SQLite database,
- 14:05for example. This is how you're going to
- 14:06learn faster, gain more experience, and
- 14:07also work towards using AI coding agents
- 14:10with agentic engineering. I previously
- 14:12mentioned other people's projects as
- 14:13well. Another great way for you to learn
- 14:15is to take other people's projects and
- 14:18then adopt them. Add on more stuff to
- 14:20them. Change around the code and see how
- 14:22the results change. This is such a great
- 14:24way to learn and it's people amongst
- 14:25engineering students everywhere. Another
- 14:27tip, as you're coding, you can also ask
- 14:29AI to give you multiple suggestions for
- 14:31how to implement something. This way you
- 14:33can expand the scope of your knowledge,
- 14:35which is going to make you a much better
- 14:36coder. And final little tip, ask AI to
- 14:38help you understand documentation. I
- 14:40hate reading documentation. Like I hate
- 14:43it so much. I don't think I've read any
- 14:45documentation since like 2024 cuz I will
- 14:48ask AI to help me understand
- 14:49documentation. To be clear, I still need
- 14:52to understand the documentation. I just
- 14:53don't like reading it. So, this is a
- 14:55great way to understand documentation
- 14:56faster and incorporate them into your
- 14:58projects. I'm not going to go into Vibe
- 15:00Coding side agentic engineering specific
- 15:02tips right now cuz there's just like way
- 15:04too much and there's a whole other video
- 15:06that I can make on this topic. So, I'm
- 15:08just going to put on screen some of my
- 15:10favorite little Vibe Coding tips. If
- 15:13that's helpful for you guys, just know
- 15:14that this is not everything and it might
- 15:16be confusing if you never had any
- 15:17experience at all. Let me know in the
- 15:19comments if if want me to actually make
- 15:21like a fully fledged video uh for that
- 15:23topic. All right, we have now come to
- 15:25the end of this video. Thank you so much
- 15:27for watching until the end. I know that
- 15:29was a lot. I really hope this is helpful
- 15:31to you. I'm so glad that you're learning
- 15:33how to code in 2026. Best of luck in
- 15:35your coding journey and I will see you
- 15:36guys in the next video or live stream.
About this transcript
This page contains the full transcript of How To Learn To Code In 2026 by Tina Huang, generated from the public captions YouTube serves with the video. The transcript has 3,655 words across 531 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.