Why Claude 3.5 Sonnet is Worth the Hype

AI with ALLIE

The professional’s guide to quick AI bites for your personal life, work life, and beyond.

July 16, 2024

I’ve been exploring Anthropic’s latest AI features and, if you couldn’t tell from my social media, I’ve been going a bit wild with the prompts.

So, in this newsletter, we’ll be talking about how this 'Radiohead' of AI is making waves, Claude 3.5 Sonnet, the new Artifacts feature, and how Anthropic stacks up against giants like OpenAI.

Let’s go.

Jump ahead:

• Winner takes… some.

• What should you know about Anthropic?

• Why did I choose to pay for Claude?

• Claude 3.5 Sonnet

• Claude Artifacts

• Claude Projects

• The type of person that will succeed

• I’m hiring: wanna join the AI fun?

• Tools, courses, and blogs that caught my eye

Winner takes… some.

At this point in the AI age, there is no one single runaway AI leader.

And that’s a good thing.

Companies like OpenAI, Anthropic, Meta, Mistral, Perplexity, Amazon, Microsoft, Nvidia, and Google are all releasing incredible models and AI applications month after month, and if you can believe it, there's still a lot more to come.

In fact, we are ridiculously early in areas like 3D spatial video (Luma is doing cool stuff), audio (Google is doing cool stuff), multilingual capabilities (ElevenLabs is doing cool stuff), structured data (Hebbia AI is doing cool stuff), and agentic AI (OpenAI is supposedly doing cool stuff).

But this newsletter is going to focus on Anthropic.

What should you know about Anthropic?

Anthropic was founded in early 2021 by a team of ex-OpenAI employees (three of whom I previously lived with in SF, all of whom are better cooks than I am).

The company describes themselves as an “AI safety and research company” with customers including LexisNexis, BCG, Slack, Asana, Jane Street, Perplexity, Quora, SAP, Scale AI, SK Telecom, Brave, Bridgewater, and more. They have 109 open roles, mostly in Research and Sales, across SF, NY, London, Dublin, and more.

When it comes to funding, brand awareness, and community size, Anthropic is still “behind” OpenAI.

Anthropic is valued at approximately $18 billion, whereas OpenAI is ~$90 billion. Anthropic’s team is considerably smaller (reportedly, Anthropic has <500 employees and OpenAI has ~2500), and the team has raised “up to” $7.6B compared to OpenAI’s $13B. OpenAI has 3.6M followers on X versus Anthropic with 300K (and LinkedIn is about 5.5M vs 400K). Anecdotally, whenever I ask my audiences or clients who has heard of ChatGPT or Claude, the hands are usually >90% versus <5%.

Their market positioning statements say the most to me—Anthropic’s heaviest marketing message is on safety, reliability, and control, whereas OpenAI’s focuses on AGI and prosperity.

One friend described Anthropic as the "Radiohead" to OpenAI's "Coldplay" - smaller, more experimental, and more creative.

I sort of loved that.

Why did I choose to pay for Claude?

Truth be told, I never used to pay for Claude.

I paid for ChatGPT but Claude was always this “other” app that didn’t have any of the features I wanted or needed (like voice dictation, mobile app, usability of output, file upload).

Then, in March 2024, Claude 3 Opus was released. And minutes after testing it, my team decided to upgrade and subscribe to the Professional Plan.

Yes, it was that fast.

Since then, and this is me speaking from experience not benchmarks (benchmarks are notoriously tricky in generative AI), Claude has always been better at capturing my tone of voice and writing style.

So when it came to crafting emails, newsletters, subject lines, product names, product descriptions, or website descriptions, Claude was my bestie.

It feels like Claude gets me. My first drafts with Claude are closer to my voice, mimic my examples more accurately, and require less editing. And compounding those gains over a week or month, it can easily save you hours.

Then, I was in the early testing group for Claude 3.5 Sonnet.

And that’s when Anthropic kicked it up a notch.

Introduction to Claude 3.5 Sonnet

Anthropic dropped Claude 3.5 Sonnet last month, and it's been shaking up the tech world (and most of my X feed).

I've been testing it extensively since before its public release, and I'm blown away by what it can do.

This particular release is my favorite.

Here's everything you need to know.

Claude 3.5 Sonnet is part of the latest model family released by Anthropic in 2024. It’s more powerful, more intelligent, and faster than its previous model (Claude 3 Opus).

But, while Sonnet itself is great, they've also released two new features that have stolen the show. One of these features, called Artifacts, has been one of my favorite AI releases this year.

The epic new feature: Claude Artifacts

I've said for years that AI is not only about chat and chatbots.

The future of AI is considerably more multimodal and interactive than what we have today, and hey, Claude Artifacts is giving us an early glimpse into that future. This new offering inside Anthropic’s assistant has:

1. A code window that shows the code for whatever it's writing

2. A preview window that shows animations, dashboards, or React components/applications from the code

There is also a history of created items, so if you are editing your prompt back and forth, you can still refer back to earlier versions of the app.

As of last week, you can also publish (available for free and pro users) and remix (available for free, pro, and team users) Artifacts.

The UX of this feature is pretty darn great, though there are a few buttons that could be improved (you can only download certain file types, you can’t mass export docs). But this feature alone makes it worth upgrading to a paid account if you haven’t already.

Looking for some Claude inspiration? Here are 10 epic ways people are using Claude Artifacts like data visualizations, games, sensitivity analyses, or animations.

AI assistants are not just chat—now they’re also enabling creation.

The second feature release: Claude Projects

Along with Artifacts, Anthropic also launched "Projects," which semi-competes with OpenAI's GPTs.

It’s sort of like a question-documents bot. Those with a Pro or Team account can save style guides, codebases, interview transcripts, work documents, past work, or any context you want Claude to know with custom instructions for repetitive prompting.

For more advanced AI users, it feels like a RAG set up to me, with the ability to hold about 500 pages of information in its context window (200K tokens).

However, it seems to have lower functionality at launch. OpenAI’s GPTs have function calling and a free GPT store with a creator fund. Anthropic hasn’t signaled a marketplace or monetization for Projects.

I also think the UX of Projects is much harder to navigate than the UX of ChatGPT custom GPTs. For beginners wanting to build repetitive prompting, ChatGPT might be easier. But for those focused on writing and willing to navigate a more slightly more complex UI, Claude might be the better choice.

The type of person that will succeed with these releases

If you believe the things around you are fixed—that a textbook is just a stack of papers with some graphics slapped on—then you will struggle with AI releases like this.

You need to go Fairy Godmother mode and add glitter to the world around you.

You need to go MacGyver mode and turn a cargo net into a rappel line.

You need to be the person to look at a textbook and go “ooo, this is inspiration! I could make a game based on this! I could turn this into a story! I could ask it questions! I could animate these graphics!”

This is the opportunity to test out more interactive elements and see what they can do.

Everything around you, you need to start thinking of it all as being 70% more interactive. You can now:

• Take a chemistry book page and spin it into a collaborative educational tool (here’s my demo)

• Turn data from invoices, earnings reports, or marketing readouts into sensitivity analyses

• Create a quick animation of a key concept in your exact brand colors for a social media post

• and 10 more amazing examples!

Things that were once inaccessible (like code) are now more accessible. Things that were once static are coming to life. And those among you with a creative eye and growth mindset are going to surge ahead.

Where we go from here

Human-AI collaboration will continue to evolve. We can expect:

- More interactivity and multimodality in our creations

- More robust decision-making capabilities

- Increased accessibility to complex tasks like coding

The key is to approach these tools with a growth mindset, challenge our existing ways of working, and be ready to explore new possibilities.

As always: stay curious, stay informed,

Allie

I’m hiring: wanna join the AI fun?

Do you excel at managing KPIs, supporting clients, creating automations, and driving continuous improvement? If so, I want you on my AI-first team.

Here's what working with us looks like:

💡 Using AI to enhance workflows—sometimes a little, sometimes a lot

💡 Choosing the best approach for the task, even if it's not AI

💡 Redefining how work gets done from the ground up

If efficiency, customer support, AI, and teamwork are your strengths, we want to hear from you. Junior candidates are welcome to apply.

Tools, courses, and blogs that caught my eye

Over the last few weeks, I’ve pulled together some of the top releases, and my take on each one. Check it out.

1. Anthropic launches Claude Artifacts — Anthropic launched Artifacts on Claude, letting users take on a more collaborative approach to AI. I've already used it for creating learning material, coding games, and WebGLs. Others have created charts, calculators, websites, and more. In my opinion, it's one of the best AI products of the year (try it)

2. Hebbia AI raises $130M at a $700M valuation — If you’re looking for AI agents to augment your work, especially in document processing and analysis, check out Hebbia, which just raised $130 million at a $700 million valuation led by a16z (read it) (my thoughts)

3. My upcoming ‘How to Use Generative AI’ course — The cat’s out of the bag: I recorded a LinkedIn Learning course on how to use generative AI and its coming out very soon. Follow me on LinkedIn and hit the 🔔 on my profile to get notified when it’s out (my profile)

4. Stanford researchers create mini WALL-E for $200 — Axel Peytavin and his team at Stanford University created a mini WALL-E for $200 that uses a camera, GPT-4o, and custom code to take actions like picking up trash, shaking hands, and chatting adorably (watch it)

5. Luma AI launches free text-to-video generation — Luma AI's new Dream Machine lets anyone create short AI videos for free, albeit with lower quality and longer generation times compared to Runway Gen 3 Alpha (try it) (my thoughts)

6. Leading legal AI research tools hallucinate… a lot — Despite their promise, new research shows leading legal AI research tools still frequently hallucinate, with error rates of 17-33% compared to GPT-4 error rate of 43% (read it) (my thoughts)

Feedback is a Gift

I would love to know what you thought of this newsletter and any feedback you have for me. Do you have a favorite part? Wish I would change something? Felt confused? Please reply and share your thoughts or just take the poll below so I can continue to improve and deliver value for you all.