Modern Creator
Why Modern Creator?
Duncan Rogoff | Learn Claude Code · YouTube

The New Way Top Anthropic Engineers Prompt Claude Code

Andrej Karpathy says the best way to brief an AI coding agent is a messy ten-minute voice ramble — so a creator tests it live by building a naturopath website from a single, uninformed viewer request.

Posted
2 days ago
Duration
Format
Talking Head + Screen Recording
educational
Views
8.8K
228 likes
Big Idea

The argument in one line.

The most effective way to brief Claude Code is not a carefully typed spec but an unstructured ten-minute voice ramble, because the model reconstructs a messy stream of consciousness into a cleaner brief than most people would type by hand.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…
  • You use Claude Code or a similar agentic coding tool and feel you must write a perfect prompt before starting.
  • You want to build something you don't fully understand yet (a niche website, app, or game) and don't know how to scope it.
  • You're willing to talk to your computer via voice/dictation instead of typing out instructions.
  • You want a concrete checklist (desire, quality bar, tools, creative freedom, verification, delivery) for structuring an open-ended ask to an AI agent.
SKIP IF…
  • You already have a tight, well-defined spec and prefer typing precise instructions over talking through an idea.
  • You're using a plain chatbot rather than an agentic tool with file access, subagents, and plan mode — the ramble technique assumes an agent that can research and execute, not just answer.
TL;DR

The full version, fast.

Anthropic engineer Andrej Karpathy tweeted that his favorite way to prompt LLMs is a ten-minute voice ramble — total mess, full stream of consciousness — because models reconstruct incoherent rambles into a cleaner brief than most people would type. The video tests this live: a real viewer comment asking for a 'beautiful naturopath website' becomes the prompt, delivered entirely by voice into Claude Code's desktop app with no typing. Before rambling, six things are worth touching on if you have them: desire, quality bar, tools and discovery, creative freedom, a verification loop, and delivery — Claude Code's own /goal command formalizes this. Claude then researches the unfamiliar topic itself, runs parallel subagents for market research and visual design, generates images, and produces a finished, mobile-checked website end to end.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →
Chapters

Where the time goes.

00:00 – 00:34

01 · Cold open

Karpathy's ramble-prompting tweet is introduced as the topic, with his Anthropic/Tesla/OpenAI background established for credibility.

00:34 – 01:14

02 · Reading Karpathy's tweet

The full tweet is read on screen: ramble for ten minutes, total mess, full stream of consciousness, let the LLM reconstruct it into something cleaner.

01:14 – 01:51

03 · Why voice beats typing

Argues rambling by voice removes the overhead of correct spelling and organized phrasing, making the process faster and the results better.

01:51 – 02:54

04 · Picking a real project

A real YouTube comment is used as the test case: a viewer named Reyna asks for a beginner's tutorial on building a naturopath website, with no reference examples offered.

02:54 – 03:13

05 · The /goal command

Introduces Claude Code's /goal command, which lets you state a target outcome and has the agent work until it's achieved.

03:13 – 04:18

06 · Six-part ramble structure

Lays out what to touch on when rambling: desire, quality bar, tools and discovery, creative freedom, verification loop, and delivery — explicitly optional if unknown.

04:18 – 04:49

07 · Turning on voice mode

Shows where voice input lives in the Claude Code desktop app (a microphone icon) versus the terminal's /voice command.

04:49 – 06:19

08 · Rambling the brief

Delivers the actual ramble into Claude Code by voice: build a naturopath website, admits not knowing what a naturopath is, asks Claude to research audience, visuals, copy, and a booking CTA.

06:19 – 06:41

09 · Plan mode as a safety check

Notes that switching to plan mode before rambling makes Claude produce a reviewable game plan instead of acting immediately, for anyone uneasy handing over full control.

06:41 – 07:08

10 · Claude researches the audience

Two subagents run in parallel: one researching the naturopathy market, audience psychographics, and regulatory language; the other researching visual design direction and choosing an image model.

07:08 – 07:32

11 · Watching subagents work

Subagent progress is inspectable in real time as Claude assembles the project, sets up an image folder, and begins generating visuals.

07:32 – 08:04

12 · Reviewing the finished website

Walks through the completed site: original copy, generated imagery, a four-step 'how care works' breakdown, and a pricing/CTA section, all produced from the single voice ramble.

08:04 – 08:27

13 · Claude verifies its own work

Claude checks its own output for mobile optimization unprompted; the video closes by pointing to the creator's paid resources and a longer $10,000 Websites course.

Atomic Insights

Lines worth screenshotting.

  • Andrej Karpathy, who has worked at Tesla, OpenAI, and Anthropic, says his favorite way to prompt LLMs is a ten-minute voice ramble with zero structure.
  • Typing forces you to overthink spelling and phrasing; talking lets a stream-of-consciousness ramble come out faster and more completely.
  • LLMs are good at reconstructing an incoherent ramble into something cleaner than what the speaker actually said, which is why the technique works.
  • You don't need to understand your own project to brief it — you can ask Claude Code to research the topic, audience, and design language for you.
  • Claude Code's /goal command lets you state a target outcome and the agent keeps working until that goal is met.
  • A useful ramble structure covers six things when you have them: desire, quality bar, tools and discovery, creative freedom, a verification loop, and delivery.
  • If you don't know what tools or references you have (MCPs, repos, existing skills), you can tell Claude to discover them itself rather than leaving the field blank.
  • Plan mode lets Claude produce a full game plan for human review before it touches any files, which is useful when you don't want to hand over complete control.
  • In the Claude Code desktop app, voice input is a microphone icon in the input bar — the /voice slash command is a terminal-only equivalent.
  • Claude Code can run multiple research subagents in parallel — market research and visual design ran simultaneously in the demo, each visible and inspectable while running.
  • The agent chose its own image-generation model (Nano Banana Pro) and resolution without being told which one to use.
  • A single vague voice prompt with no reference site, brand, or copy produced a complete website with generated copy, images, a four-step process breakdown, and pricing.
  • Claude checked its own output for mobile optimization as a final step, without being explicitly asked to.
Takeaway

A messy voice ramble briefs an AI agent better than a careful prompt.

WHAT TO LEARN

Talking through an unfamiliar project for ten unstructured minutes gives an agentic coding tool more to work with than a tightly typed prompt, especially when you don't yet understand the thing you're trying to build.

02Reading Karpathy's tweet
  • Typing forces premature precision — worrying about spelling and phrasing costs you material you'd otherwise include if you just talked.
  • An LLM reconstructing your ramble often produces a cleaner brief than the one you actually spoke, so don't over-edit yourself while rambling.
03Why voice beats typing
  • Removing the mechanics of typing (spelling, structure) is what makes voice rambling faster and more complete than a typed prompt.
04Picking a real project
  • You can brief a project you don't understand: a real, half-formed request is enough to start from if you're honest about not having references.
05The /goal command
  • Stating a target outcome explicitly (via a /goal-style command) lets the agent keep iterating until that specific bar is met, rather than stopping after one pass.
06Six-part ramble structure
  • When structuring an open-ended ask, it helps to touch on desire, a quality bar, available tools, how much creative freedom to grant, how output should be checked, and where it should land.
  • If you don't know what tools, repos, or references exist, tell the agent to discover them itself instead of leaving the request thinner.
09Plan mode as a safety check
  • Use a plan-review step before execution if you want to see the agent's intended approach before it starts changing files.
10Claude researches the audience
  • Ask the agent to research the topic, audience, and design language as part of the same request — expertise you lack becomes a research task, not a blocker.
11Watching subagents work
  • Watching an agent's subagents work in real time is a way to catch a wrong turn early, before a full build is finished on a bad assumption.
12Reviewing the finished website
  • A single spoken, uninformed request can still produce a finished, coherent piece of work — the burden of expertise shifts from you to the tool's research step.
13Claude verifies its own work
  • An agent that self-checks its output against a stated goal (e.g. mobile optimization) closes the loop without needing to be asked twice.
Glossary

Terms worth knowing.

Ramble session
Karpathy's term for briefing an LLM with an unstructured, stream-of-consciousness voice monologue instead of a written, organized prompt.
/goal command
A Claude Code slash command that lets a user state a target outcome; the agent keeps working iteratively until that goal is satisfied.
Plan mode
A Claude Code setting where the agent produces a full plan for a task and waits for human review and approval before making any changes.
Verification loop
The repeated self-check process an agent runs to confirm its own output actually meets the stated goal before finishing.
Subagent
A separate agent instance Claude Code spins up to handle one piece of a larger task (e.g. market research) in parallel with other work.
Nano Banana Pro
An image-generation model Claude selected on its own, at 2K or 4K resolution, to produce the website's visuals.
Resources

Things they pointed at.

00:00linkAndrej Karpathy's tweet on ramble sessions
01:41linkHow to Build $10,000 Websites with Claude Code (2-hour masterclass)
00:00productClaude Code Club (skool community) ↗
Quotables

Lines you could clip.

00:27
“I like to lean back, switch to voice mode, and just ramble for like ten minutes. Total mess. Anything goes. Full stream of consciousness.”
the core claim of the video, in Karpathy's own words→ TikTok hook↗ Tweet quote
01:48
“The LLMs are somehow very good at reconstructing long incoherent rambles, and often their echo of your own tangle of thoughts comes out quite a bit cleaner than what you started with.”
explains the mechanism, not just the claim→ IG reel cold open↗ Tweet quote
03:36
“If there's any information — this is kind of a fancy term — like the verification loop, basically how it's going to check its work over and over again until it achieves your desired goal.”
names a reusable concept for briefing agents→ newsletter pull-quote↗ Tweet quote
The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

metaphor
One of the top engineers at Anthropic just released their favorite way to prompt quad code. And if you've been using quad code at all, especially if you're a beginner, you're gonna wanna pay attention to this. So this is Andre Karpathy, and you might have heard of him.
He's one of the top AI engineers on the entire planet. Right now, he's at Anthropic. Before that, he was at Tesla and even OpenAI.
And he just put out this tweet with one of his favorite ways to prompt Claude code. So I'm gonna break it down for you, and we're even going to test it out on a real world example to see if it's even any good. So he says, one pattern I find useful for working with LLMs is a nice long ramble session.
Sometimes the LLM needs more bits to understand what you're trying to achieve, but you're too lazy to type them. In these cases, I like to lean back, switch to voice mode, and just ramble for like ten minutes. Total mess.
Anything goes. Full stream of consciousness. Sometimes I declare it up top, something like switching to speech recognition.
Sorry for any typos. Sometimes I turn it into a small interview for a few turns, but I find that the LLMs are somehow very good at reconstructing long incoherent rambles, and often their echo of your own tangle of thoughts comes out quite a bit cleaner than what you started with.
The result is that you improve the mind belt and have to correct things less from that point on. So I actually tend to agree with Andre. I've tried this out on a million different projects, and this is one of my favorite ways to work.
It is so much less restrictive than having to type, and you can really truly just have this crazy stream of consciousness ramble for ten minutes, and Claude will figure out what it is that you're trying to achieve. The reason that it's so good is because you're not stuck overthinking, like making sure you're spelling everything correctly or making sure that you're getting all of your thoughts down, especially when you're typing.
It's so much faster to work this way, and honestly, the results are surprising. So I wanted to test this out on a real world example. So in one of my latest videos, I put together, like, a full two hour course on how to build fancy websites with Cloud Code.
I'll probably leave it up top somewhere. But then I had this little comment exchange underneath. I thought this would be a really good example for the video today.
So this woman, Reyna, says, hi, Duncan. Can you please do a beginner's tutorial how to build a beautiful nature naturopath website naturopath website.
Thank you. Okay. I said, I can look into it.
It's a pretty specific request, to be honest, but I have some thoughts. And I asked if she had any references from something that they like. And she said, no.
I don't. So I thought this would be a great example for anybody, especially if you're just getting started where, like, you may not even know what you're working on. So I don't know what a naturopath is.
I don't really know how to build a website for this. So I want to hop into Cloud. I'm just gonna ramble any of my thoughts that I might have, and we're gonna see what Cloud comes up with.
And so it really doesn't matter if you're trying to build this specific website. Like, you can really just ramble the quad even if you're trying to build a game or you're trying to build a microsite or you're trying to build an app. Right?
And so I'm just kinda giving you the tools here no matter what it is that you're actually trying to build. If you wanna learn how to use quad code to build websites, apps, games, more, and get access to all of my resources, just just check the link in the description. Now if you are truly starting from zero, there are a couple things kind of just, like, worth considering, like, when you're rambling, and Quad works really well if you give it a specific goal.
You can actually type slash goal and, like, tell it what you're trying to achieve, and it will work until it achieves that goal. And so when you're talking to Claude, there's just, like, again, a couple of things that you can think about. Like, one, desire.
Like, what is the thing that you are trying to do? Two is the quality bar. Like, what defines the type of quality that you're trying to achieve?
Three is tools and discovery. So tools, like, what does it have access to? Like, any MCPs or GitHub repos or existing skills.
If you have it, great. If you don't, don't worry about it because that's where discovery comes in. Like, you basically tell Quad, like, figure out any of the tools that might exist somewhere out there to figure out how to achieve this goal.
Creative freedom. Like, give Quad, like, the freedom to, like, make its own choices and figure it out. These models are insanely smart now, and they will do a very good job on its own.
There's any information, this is, like, kind of a fancy term, like the verification loop, basically, like, how it's going to check its work over and over again until it achieves your desired goal. Delivery, like, is there a folder that you want this on? Like, do you want it to write, like, something in HTML?
Like, how do you want to give this thing to you? And then at the end, just one line that says, like, what the actual goal is just to restate it so Quad is sure. So, again, if you don't know or you don't have any of this, don't worry about it.
I just want to, like, put it out there for you so you can start thinking about how to, like, really talk to Quad. So I'm going to hop over into Claude code inside of the desktop app. Carpathi did say one thing about using voice mode, and that's really only relevant if you're using the terminal.
You can type slash voice and it'll activate your microphone and you can talk to Claude. But inside of the desktop app itself, you actually have this little microphone down here, so you can do this without the command. So I'm in the desktop app.
I'm inside of Claude Code. I've just selected it in this left panel over here. I'm just going to create a new folder on my desktop.
I'm just gonna call this, like, nature o path website because that's what the ask was, and we'll see what we can come up with. I'm trying to create a beautiful naturopath website.
I don't really even know what that is, so I need you to research what that term is, maybe who the audience might be, what types of visuals or aesthetic or design makes sense for this group of people, like any language they might use, like a type of product that I might sell to them. I think maybe in the end, the CTA is to get somebody to book a call with me, although I'm not really sure.
Is there a design language that we should use, typography, colors? Like, should we generate some type of images? Like, the goal is to really just create this absolutely beautiful website.
And, like, feel free to look to see if we have any skills that you have access to. If there are any, like, web design libraries on the Internet, you can use the Internet. You can use GitHub repos.
You can really use whatever you want. So, again, I'm just trying to create a naturopath website that resonates with this audience that has copy written for them in their own language. And I guess with the CTA to book a call with me to presumably sell them something, although I don't know what that is, and I don't know how to price it.
And the last thing I guess I missed is that I'm just gonna say, I want you to create this website in beautiful HTML, and you can store it in this folder and just so I can open it up immediately, like, on my local machine.
And so you can see, just had this, like, really long ramble session with Claude code. I'm not really even sure what it is that I'm trying to make, but Claude is gonna do all the research for me, figure out about this audience, about what a naturopath is, how to make the website, and I'm gonna send this off, and let's see what we come up with.
One other tip while that's cooking is if you don't want to give Claude complete control over things, you can actually come down here and you can switch this to plan mode before you start your ramble. And that way, Claude is actually going to create an entire game plan for you on how to build this type of website, and it's gonna let you review it as a human before Claude gets to work.
So you can make any changes at that point before Claude builds anything. And so this is pretty cool. Claude has already started working for us and doing some research.
So it's researching the market, the audience, and the type of copy that it should write. It's researching what naturopathy actually is. It's researching any language that the patients use.
It's researching what type of visual design makes sense for this audience. It's set up an image folder inside of the folder on my desktop for any images that were generated, and it's decided that Nano Banana Pro at two k or four k resolution is the right model for this. If you want, you can click in here to see all of the agents that are running and exactly what it is that they're working on.
So Cloud finished, and it put together this absolutely beautiful, clean website. It generated all of the copy for me. It figured out what to talk about.
It created the CTAs. It created all of these beautiful images, all of the sections, like any of the most common pain points that someone might be experiencing.
I love all of these images that it generated. It goes and does a full breakdown of how care actually works. So the four step process that someone might experience when they come in, like, look at the quality of this.
Like, I didn't do any of the prompting or generation. Like, Claude figured out how to take care of it. Then it left a placeholder for, like, a little personal story about, you know, me, a breakdown of some of the tests we might run along with the costs that it found just by researching the Internet, and then, again, the full program and pricing here with a CTA to book a free call.
This is what the rest of the website looks like, and honestly, it's absolutely It even checked its own work to make sure that it's mobile optimized. So, again, ramble to Claude, tell it exactly what you're after, and it's going to figure out how to build it for you.
If you wanna learn how to build anything with Claude code and get access to all of my resources already built for you, just check the link in the description. If you wanna watch the full two hour master class on building beautiful $10,000 websites, just check out this video right here.
I'll see you over there.
The Hook

The bait, then the rug-pull.

Andrej Karpathy — an engineer who has worked at Tesla, OpenAI, and now Anthropic — tweeted that his favorite way to prompt an LLM is to abandon careful typing altogether and just ramble by voice for ten minutes. This breakdown tests that claim live: a real, half-formed viewer comment about a 'naturopath website' becomes the entire brief, spoken cold into Claude Code with no notes and no reference site.

Frameworks

Named ideas worth stealing.

03:13list

Ramble structure (desire, quality bar, tools + discovery, creative freedom, verification loop, delivery)

  1. Desire
  2. Quality bar
  3. Tools + Discovery
  4. Creative Freedom
  5. Verification Loop
  6. Delivery
  7. /goal line

A checklist of things to optionally cover when rambling a brief to Claude Code: what you want, what quality bar defines success, what tools/references exist (or telling Claude to discover them), how much creative latitude to give the agent, how it should check its own work, where to deliver the output, and a closing one-line restatement of the goal.

Steal forbriefing any agentic AI tool on an open-ended build, not just websites
CTA Breakdown

How they asked for the click.

VERBAL ASK
08:13product
“If you wanna learn how to build anything with Claude code and get access to all of my resources already built for you, just check the link in the description.”

Soft CTA to a paid Skool community ($9), repeated twice (mid-video at ~2:30 and again at the close), plus a pointer to a separate free 2-hour masterclass video.

MENTIONED ON CAMERA
00:00productClaude Code Club (skool community) ↗
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
Storyboard

Visual structure at a glance.

cold open
hookcold open00:00
the tweet
promisethe tweet00:37
the /goal framework
valuethe /goal framework02:54
the finished website
valuethe finished website07:13
self-verification + close
ctaself-verification + close08:04
Frame Gallery

Visual moments.

cold open00:00Frame at 00:09 from The New Way Top Anthropic Engineers Prompt Claude Code00:09Frame at 00:18 from The New Way Top Anthropic Engineers Prompt Claude Code00:18Frame at 00:22 from The New Way Top Anthropic Engineers Prompt Claude Code00:22Frame at 00:28 from The New Way Top Anthropic Engineers Prompt Claude Code00:28the tweet00:37
cold open00:00Frame at 00:09 from The New Way Top Anthropic Engineers Prompt Claude Code00:09Frame at 00:18 from The New Way Top Anthropic Engineers Prompt Claude Code00:18Frame at 00:22 from The New Way Top Anthropic Engineers Prompt Claude Code00:22Frame at 00:28 from The New Way Top Anthropic Engineers Prompt Claude Code00:28the tweet00:37Frame at 00:44 from The New Way Top Anthropic Engineers Prompt Claude Code00:44Frame at 00:50 from The New Way Top Anthropic Engineers Prompt Claude Code00:50Frame at 00:57 from The New Way Top Anthropic Engineers Prompt Claude Code00:57Frame at 01:04 from The New Way Top Anthropic Engineers Prompt Claude Code01:04Frame at 01:13 from The New Way Top Anthropic Engineers Prompt Claude Code01:13Frame at 01:17 from The New Way Top Anthropic Engineers Prompt Claude Code01:17Frame at 01:23 from The New Way Top Anthropic Engineers Prompt Claude Code01:23Frame at 01:29 from The New Way Top Anthropic Engineers Prompt Claude Code01:29Frame at 01:35 from The New Way Top Anthropic Engineers Prompt Claude Code01:35Frame at 01:42 from The New Way Top Anthropic Engineers Prompt Claude Code01:42Frame at 01:47 from The New Way Top Anthropic Engineers Prompt Claude Code01:47Frame at 01:54 from The New Way Top Anthropic Engineers Prompt Claude Code01:54Frame at 02:00 from The New Way Top Anthropic Engineers Prompt Claude Code02:00Frame at 02:06 from The New Way Top Anthropic Engineers Prompt Claude Code02:06Frame at 02:13 from The New Way Top Anthropic Engineers Prompt Claude Code02:13Frame at 02:21 from The New Way Top Anthropic Engineers Prompt Claude Code02:21Frame at 02:25 from The New Way Top Anthropic Engineers Prompt Claude Code02:25Frame at 02:31 from The New Way Top Anthropic Engineers Prompt Claude Code02:31Frame at 02:38 from The New Way Top Anthropic Engineers Prompt Claude Code02:38Frame at 02:44 from The New Way Top Anthropic Engineers Prompt Claude Code02:44Frame at 02:47 from The New Way Top Anthropic Engineers Prompt Claude Code02:47the /goal framework02:54Frame at 03:03 from The New Way Top Anthropic Engineers Prompt Claude Code03:03Frame at 03:09 from The New Way Top Anthropic Engineers Prompt Claude Code03:09Frame at 03:16 from The New Way Top Anthropic Engineers Prompt Claude Code03:16Frame at 03:22 from The New Way Top Anthropic Engineers Prompt Claude Code03:22Frame at 03:29 from The New Way Top Anthropic Engineers Prompt Claude Code03:29Frame at 03:35 from The New Way Top Anthropic Engineers Prompt Claude Code03:35Frame at 03:42 from The New Way Top Anthropic Engineers Prompt Claude Code03:42Frame at 03:48 from The New Way Top Anthropic Engineers Prompt Claude Code03:48Frame at 03:55 from The New Way Top Anthropic Engineers Prompt Claude Code03:55Frame at 04:01 from The New Way Top Anthropic Engineers Prompt Claude Code04:01Frame at 04:10 from The New Way Top Anthropic Engineers Prompt Claude Code04:10Frame at 04:17 from The New Way Top Anthropic Engineers Prompt Claude Code04:17Frame at 04:21 from The New Way Top Anthropic Engineers Prompt Claude Code04:21Frame at 04:27 from The New Way Top Anthropic Engineers Prompt Claude Code04:27Frame at 04:33 from The New Way Top Anthropic Engineers Prompt Claude Code04:33Frame at 04:39 from The New Way Top Anthropic Engineers Prompt Claude Code04:39Frame at 04:45 from The New Way Top Anthropic Engineers Prompt Claude Code04:45Frame at 04:52 from The New Way Top Anthropic Engineers Prompt Claude Code04:52Frame at 04:58 from The New Way Top Anthropic Engineers Prompt Claude Code04:58Frame at 05:04 from The New Way Top Anthropic Engineers Prompt Claude Code05:04Frame at 05:10 from The New Way Top Anthropic Engineers Prompt Claude Code05:10Frame at 05:16 from The New Way Top Anthropic Engineers Prompt Claude Code05:16Frame at 05:22 from The New Way Top Anthropic Engineers Prompt Claude Code05:22Frame at 05:28 from The New Way Top Anthropic Engineers Prompt Claude Code05:28Frame at 05:34 from The New Way Top Anthropic Engineers Prompt Claude Code05:34Frame at 05:40 from The New Way Top Anthropic Engineers Prompt Claude Code05:40Frame at 05:46 from The New Way Top Anthropic Engineers Prompt Claude Code05:46Frame at 05:52 from The New Way Top Anthropic Engineers Prompt Claude Code05:52Frame at 05:58 from The New Way Top Anthropic Engineers Prompt Claude Code05:58Frame at 06:04 from The New Way Top Anthropic Engineers Prompt Claude Code06:04Frame at 06:10 from The New Way Top Anthropic Engineers Prompt Claude Code06:10Frame at 06:16 from The New Way Top Anthropic Engineers Prompt Claude Code06:16Frame at 06:19 from The New Way Top Anthropic Engineers Prompt Claude Code06:19Frame at 06:29 from The New Way Top Anthropic Engineers Prompt Claude Code06:29Frame at 06:37 from The New Way Top Anthropic Engineers Prompt Claude Code06:37Frame at 06:41 from The New Way Top Anthropic Engineers Prompt Claude Code06:41Frame at 06:51 from The New Way Top Anthropic Engineers Prompt Claude Code06:51Frame at 06:57 from The New Way Top Anthropic Engineers Prompt Claude Code06:57Frame at 07:04 from The New Way Top Anthropic Engineers Prompt Claude Code07:04the finished website07:13Frame at 07:17 from The New Way Top Anthropic Engineers Prompt Claude Code07:17Frame at 07:23 from The New Way Top Anthropic Engineers Prompt Claude Code07:23Frame at 07:29 from The New Way Top Anthropic Engineers Prompt Claude Code07:29Frame at 07:35 from The New Way Top Anthropic Engineers Prompt Claude Code07:35Frame at 07:41 from The New Way Top Anthropic Engineers Prompt Claude Code07:41Frame at 07:46 from The New Way Top Anthropic Engineers Prompt Claude Code07:46Frame at 07:54 from The New Way Top Anthropic Engineers Prompt Claude Code07:54Frame at 08:00 from The New Way Top Anthropic Engineers Prompt Claude Code08:00self-verification + close08:04Frame at 08:13 from The New Way Top Anthropic Engineers Prompt Claude Code08:13Frame at 08:18 from The New Way Top Anthropic Engineers Prompt Claude Code08:18Frame at 08:24 from The New Way Top Anthropic Engineers Prompt Claude Code08:24
Watch next

More from this channel + related breakdowns.