Экстракт Знаний

Claude Code 2.0 + Claude Sonnet 4.5 is Finally Out!

7:51

Claude Code 2.0 + Claude Sonnet 4.5 is Finally Out!

Ray Amjad 30.09.2025 6 751 просмотров 223 лайков обн. 18.02.2026

Смотреть на YouTube

Поделиться Telegram VK Бот

Транскрипт Скачать .md

Анализ с AI

Описание видео

Level up with my Claude Code Masterclass 👉 https://www.masterclaudecode.com/ Learn the AI I'm learning with my newsletter 👉 https://newsletter.rayamjad.com/ Got any questions? DM me on Instagram 👉 https://www.instagram.com/theramjad/ 🎙️ Sign up to the HyperWhisper Windows Waitlist 👉 https://forms.gle/yCuqmEUrfKKnd6sN7 Since I've never accepted a sponsor, my videos are made possible by... —— MY CLASSES —— 🚀 Claude Code Masterclass: https://www.masterclaudecode.com/?utm_source=youtube&utm_campaign=wGQFUBFlB8Q - Use coupon code YEAR2026 for 35% off —— MY APPS —— 🎙️ HyperWhisper, write 5x faster with your voice: https://www.hyperwhisper.com/?utm_source=youtube&utm_campaign=wGQFUBFlB8Q - Use coupon code YEAR2026 for 35% off 📲 Tensor AI: Never Miss the AI News - on iOS: https://apps.apple.com/us/app/ai-news-tensor-ai/id6746403746 - on Android: https://play.google.com/store/apps/details?id=app.tensorai.tensorai - 100% FREE 📹 VidTempla, Manage YouTube Descriptions at Scale: http://vidtempla.com/?utm_source=youtube&utm_campaign=wGQFUBFlB8Q 💬 AgentStack, AI agents for customer support and sales: https://www.agentstack.build/?utm_source=youtube&utm_campaign=wGQFUBFlB8Q - Request private beta by emailing r@rayamjad.com ————— CONNECT WITH ME 🐦 X: https://x.com/@theramjad 👥 LinkedIn: https://www.linkedin.com/in/rayamjad/ 📸 Instagram: https://www.instagram.com/theramjad/ 🌍 My website/blog: https://www.rayamjad.com/ ————— Links: - Changelog: https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md - Announcement: https://www.anthropic.com/news/claude-sonnet-4-5 - Imagine with Claude: https://claude.ai/imagine/ - Context Editing: https://docs.claude.com/en/docs/build-with-claude/context-editing - Benchmarks: https://www.vals.ai/models/anthropic_claude-sonnet-4.5-20250929-thinking Timestamps: 00:00 - Intro 00:19 - Updating 00:28 - New Design 00:51 - /rewind 01:47 - /usage 01:54 - Tab for Toggle Thinking 02:13 - Ctrl+R for Search 02:31 - Claude Claude SDK Rename 02:42 - Adding Subagents 03:09 - VSCode Extension 03:59 - Claude-Sonnet-4.5 04:50 - Research Preview (Imagine with Claude) 05:18 - Claude.ai Usage 05:43 - Context Management 06:55 - Pricing 07:15 - Conclusion

Оглавление (16 сегментов)

0:00 Intro 74 сл.
0:19 Updating 45 сл.
0:28 New Design 79 сл.
0:51 /rewind 217 сл.
1:47 /usage 32 сл.
1:54 Tab for Toggle Thinking 87 сл.
2:13 Ctrl+R for Search 59 сл.
2:31 Claude Claude SDK Rename 42 сл.
2:42 Adding Subagents 88 сл.
3:09 VSCode Extension 188 сл.
3:59 Claude-Sonnet-4.5 195 сл.
4:50 Research Preview (Imagine with Claude) 83 сл.
5:18 Claude.ai Usage 89 сл.
5:43 Context Management 247 сл.
6:55 Pricing 75 сл.
7:15 Conclusion 154 сл.

Intro

All right, so a couple hours ago, Claude Code released their biggest update yet, which is 2. 0. And this includes all these updates over here, and we're going to be going over them and talking about what's changed, and also talking a little bit about Claude Sonnet 4. 5, which is now the best in the world coding model. And you can see that right over here. But let's focus on Claude Code

Updating

first. So firstly, if you want to update Claude Code to make sure you're on the latest version, you can run claude update, and that will then update to the latest version for you. And running Claude, you should see a fresh coat of paint

New Design

throughout the whole application. So the design is a little bit different from before. I do like the little droid figure that they have. And of course, if you run /model, then you can see that Sonnet 4. 5 is now the smartest model and used for daily use. But because of the way that I've set up my terminal and theme, you can't see the new background color, but you may be able to on other devices. Anyways,

/rewind

a pretty big change they added is this new /rewind command. So you can go back to any point in the conversation history where any changes may have happened. So it's like making checkpoints for the conversation history as you're making changes. So, for example, over here, I got it to make some changes to colors for the application. And if I do /rewind, and you can see I can go back just before I entered that command, "Change the landing page colors of the nextjs application. " So, I think a checkpoint is basically made throughout the codebase whenever you give it a command, and those are like the checkpoints. So, you can see this is current, this is /rewind where there would be no code changes, and this is like when I entered this message. So, if I press that, it will ask me if I want to restore the code and conversation, restore just a conversation, or restore the code. So, I think this option is like really handy, really nice. It means I probably don't have to use git as much to like go back and forth between changes when I'm implementing a feature. And it seems more flexible as well because you can go back in terms of conversation history too. You

/usage

can also do /usage to see your plan limits, and it takes you to this new tab of settings that shows you your usage for the current session. You can also now

Tab for Toggle Thinking

use Tab to toggle thinking, so if you press Tab on your computer, it should turn it off and on. And this will turn on thinking for whichever model that you're using because Anthropic ship a non-thinking and thinking version of their models. And of course, if you do want to see how the non-thinking and thinking models compare, then you can go on this website over here, and it should list out the comparisons and benchmarks for you. You can also now do Ctrl-R to search

Ctrl+R for Search

your entire chat history. So, if I do Ctrl-R, then search "page," then press Tab, it jumps back to like that particular prompt and that particular moment. So, I can make a modification to a prompt and then press Enter again. Unfortunately, I have no idea what this "unshipped claude config command" is. But for the Claude Code SDK,

Claude Claude SDK Rename

they now have renamed it to the Claude Agent SDK, probably because a lot of people are using it to do things that weren't coding related, and they just wanted to make that clearer. They also now allow you to add subagents

Adding Subagents

dynamically with the --agents flag. So, if you have a file that looks kind of like this, and it's a JSON file, then you can go to your terminal and run claude --agents. /subagent. json and then reference the JSON file. And that should create both of those agents within Claude Code. And I'm actually not sure if I used that command properly because I don't see the agent, despite it being described as this. So, like, maybe I'll come back to this when I figure this out.

VSCode Extension

They also now have a brand new native VS Code extension. So, if you go to VS Code and then press install, you can see that this should be on version 2. 0. 0, 2. 0. 0 over here. And after it's installed, then you can press the Claude button in the top right. And it should look kind of like a pretty nice chat. "Can you change the color of the website landing page to green? " And now it's going to do everything that it does. And you can see it has a bunch of commands if I zoom in a second, then I can select any of these commands over here when it comes to attaching a file, for example, by doing / and then I don't know, /mention and then choose a file from this dropdown list over here. And this is probably preferred for people who like having more of an interface and being less in the terminal, maybe. But yeah, it's interesting. Yes, I'll leave a comment down below once I figure out how to use this particular command. But that was

Claude-Sonnet-4.5

everything to do with Claude Code 2. 0. Anyways, moving on to Sonnet 4. 5. You can see right over here, it's now the best code like software engineering agent in the world. So, it gets a high score on the SWE-bench Verified benchmark over here. You can see some of the scores over here, and it seems it's got about a 3% - 5% jump between Opus 4. 1 and Sonnet 4 when it comes to agentic coding in this benchmark. So, I'm not sure if that will be noticeable when you're actually coding day-to-day, but I will make a follow-up video about it if it actually is. They have their own misaligned behavior scores, and it seems to be their most aligned model. And you can look at the system card as well if you're more interested in all these scores, which is like a super long document which I won't be reading. As I said earlier, they renamed the Claude Code SDK to be the Claude Agent SDK. So, you can now basically build agents for any other task quite easily with this SDK. And they also released this new research preview

Research Preview (Imagine with Claude)

called "Imagine with Claude" that if you go to, then you can basically come up with some random thing with your Claude Code account. So, if I press like "Take me to NYC in the 1980s," let's see what it does. And you can see it's done right over here. I'm not exactly sure what this is used for, but maybe it's pretty cool. And then it makes me this like random thing for NYC. And this is a bit naughty, so anyways,

Claude.ai Usage

let's go back. You can now track your usage in real-time across Claude apps and Claude Code by going to settings in claude. ai. And you can see this is my current usage right over here for my five-hour block. And this is exactly the same thing that if you do Claude and then /usage, it is the same amount as you can see over here. Oh, actually it's 1% off. I'm not exactly sure why. That's quite interesting. Maybe it's the way they're rounding it. Something else they

Context Management

also added is context management. And you can see right over here in a small example, where it's playing a Catan game, which is like a popular board game. And what happens is that once Claude realizes it doesn't need a particular thing in its context anymore, then it removes it, as you can see over here. So, it crosses it out, and then the amount of tokens in its context decreases, and it also removes stale context as well. So, I think that's particularly useful when it comes to coding because there may be a particular terminal output that happened in Claude Code that you no longer need, and that can be removed, and then you don't have to compact the conversation quite as much. There is some documentation about it over here. So, it says that when activated, "The API automatically clears the oldest tool results in chronological order, replacing them with placeholder text to let Claude know that the tool result was removed. " But doing this context editing invalidates the cached prompt. And because cached tokens are cheaper, to account for this, they recommend "clearing enough tokens to make the cache invalidation worthwhile. " Use the clear_at_least parameter to ensure a minimum number of tokens is cleared each time. And then it's supported on these models over here, and they give an example of how to use it that you can read through using the link down below. Another thing worth noting is

Pricing

that the pricing for Sonnet 4. 5 is exactly the same when it comes to input tokens and output tokens as Sonnet 4 was right down over here. And I think the general statement that people are having is that Anthropic don't really care about making their models cheaper, they just care about having the best models overall. But we'll see what happens when Sonnet 5 and Opus 5 come out. Anyways, if you do

Conclusion

want to see more Claude Code related news in the future, then do subscribe to the channel and like the video because it lets me know that I should make more of this stuff and also lets the algorithm know to recommend this to more people. Anyways, if you are interested in improving your vibe coding and vibe marketing skills and learning some advanced techniques that I cover in my community over here, then there will be a link down below. I also teach people how to build profitable web applications and mobile applications, and a bunch of people have had success as you can see right over here. If you click the link down below, there will be a lot more information, but if you do join, then you get personal help from me for anything that you're stuck with, and I think that's one of the biggest value adds for many people.

Ещё от Ray Amjad

How to Make Anki Flashcards 10x Faster with AI (for free!)

Ray Amjad | 23.04.2025 | 6 сегм. | 267 982

Staying Organised for A-levels // Cambridge Student

Ray Amjad | 05.09.2021 | 11 сегм. | 117 382

How I Got an A* in Physics A-level (Cambridge Student)

Ray Amjad | 28.09.2021 | 14 сегм. | 68 131

Mistakes I Made in Sixth Form

Ray Amjad | 17.09.2021 | 10 сегм. | 68 002

Brutally Honest Replit Review (Not Sponsored)

Ray Amjad | 12.09.2025 | 16 сегм. | 49 405

How I Got an A* in Further Maths A-level (Cambridge Student)

Ray Amjad | 02.10.2021 | 14 сегм. | 47 938

Ctrl+V

Экстракт Знаний в Telegram

Транскрипты, идеи, методички — всё самое полезное из лучших YouTube-каналов.

Подписаться