🐳 Why DeepSeek caused stock markets to crash worldwide

Hello AInauts,

Welcome to the latest issue of your favorite newsletter. Today with a closer look at the AI madness and lots of practical tips, from simple to advanced.

This is what we have in store for you today:

• 🐳 DeepSeek causes stock markets to crash worldwide

• 🔥 Scraping: How to save any website without any technical effort

• 💭 ChatGPT Canvas gets a great update

Here we go!

🐳 DeepSeek causes stock markets to crash worldwide

Yes, this is now the third issue in a row with DeepSeek as the topic. (Sorry...)

But of course we have to talk about the mini stock market crash from earlier this week.

If you also hold an MSCI World ETF, or even NVIDIA individual stocks, you may have seen it anyway:

NVIDIA alone lost over 500 billion in market value in one day! And all because of a new model from China?

Let's take a moment to share a few thoughts about it in simple terms.

What has happened?

We have already written about DeepSeek's new R1 reasoning model.

Let's take another quick look at R1:

• In the benchmarks, it is almost on a par with OpenAI's o1,

• but had only about 5% of the development costs compared to o1.

• And this despite the fact that, due to the U.S. chip export ban only inferior chips were available for training!

This fact has led to a small panic here in the West:

• NVIDIA shares lost 17% in one fell swoop - a historic one-day loss!

• The CEO, Jensen Huang, had to watch his fortune shrink by 21 billion. (He will cope)

• The Nasdaq fell by over 3%, while many AI stocks lost 8% or more.

• The DeepSeek app overtakes ChatGPT in the App Store and lands at #1. 😉

To top it all off, DeepSeek is now also launching a new image model, Janus-Pro. It is said to be at about the same level as DALL-E 3 and Stable Diffusion.

But more on that another time. Back to the stock market.

The panic was caused by the fact that many people realized for the first time that they might not just need more computing power and more money to win the AI arms race.

And that the sales of the big AI players, above all NVIDIA and its enormous share valuation, might not be so sustainable after all...

Since China is involved, the rumor mill naturally started again right away 😉.

Our take: Don't panic, innovation stimulates the market

If you look at the market, the whole thing is already recovering a little and looks like a typical overreaction.

We don't want to go into extreme depth now - but if you are interested in the background and theories, read the following article, this X post and perhaps this one.

For us, the following points are crucial:

Innovation > Regulation and tariffs

The DeepSeek team has impressively shown how to overcome disadvantages and lack of resources through innovation to create great things.

Because they only had access to inferior chips, the team was forced to work more efficiently and better. To innovate.

Too much regulation, too many sanctions just don't seem to work in the long term (Hi, 🇪🇺, we are looking at you). We are pro-competition, pro-innovation.

More efficiency in AI is great

It's a great development that new models are becoming more and more efficient. This makes so much more possible for users and companies, and AI can make things better in more areas of our lives. We will all use the possibilities even more as a result.

Big tech and consumers should continue to win

Finally, a look into the crystal ball.

More efficient and cheaper models should be more of an advantage for the big techs like Apple and Meta. They already have a direct line to us consumers and can therefore integrate new features more and more cheaply or even for free. Innovation revitalizes the field!

Personally, we are also less concerned about NVIDIA. Voice models are a big issue. But with images, video, autonomous vehicles, robots, etc., there should still be enough demand for fast chips. 😉

So it's not that bad.

(Not an investment recommendation, just our opinion).

🔥 Scraping: How to save any website without any technical effort

Let's move on to another interesting topic.

A regular favorite, because it is often not easy to implement, as it is rather technical. But it's hugely important for everyone - not just those who work a lot with AI.

We are talking about web scraping!

By this, we mean the backing up of (structured) data from websites. We have often written about this, using tools such as Apify or Jina, etc.

This is also a recurring topic in our automation community. We are currently working on a comprehensive deep dive, which we will soon publish for all Premium members.

Today, however, we want to briefly introduce you to a new feature of the Firecrawl Scraper.

Why is this important?

There are an incredible number of reasons for scraping. But here are three simple practical examples:

• You want to build a chatbot for your company and make all blog articles and all subpages of the website available as knowledge.

• You want to do a market research and save and analyze all products of your Amazon competitors including price and rating in a list.

• You want to pull down the ratings of your company or competitors from Trust Pilot and analyze them with AI.

And to avoid having to do this manually, you use a web scraper. However, this can quickly become technical and time-consuming.

Until now. Because thanks to Firecrawl /extract, you simply describe what you want to do and Firecrawl scrapes the data for you.

This is how it works:

We once wanted to scrape the titles of our individual newsletters and the respective publication date from our website.

To do this, we create an account with Firecrawl.

In the Playground, we then go to Extract (/extract) Beta:

Here we simply enter what we want to do, i.e. collect the date and title of the posts from our website.

Firecrawl then automatically creates a schema for us and starts collecting the data.

You can then copy the formatted result directly and have ChatGPT create a table from it, for example:

As a little intro, we are really impressed. It works much easier than any other tool we've used so far.

P.S. For all automation experts: Firecrawl is of course also available via API and even has a direct no-code connection to Zapier.

💭 ChatGPT Canvas gets a great update

After all the DeepSeek talk, let's end the issue with a cool ChatGPT update. (No, nothing about the operator - you can read about that here).

The Canvas feature in ChatGPT now FINALLY also has the possibility to use the o1 model (or vice versa: o1 can now use Canvas).

This allows you to display your code properly in a live preview.

To be honest: ChatGPT is still our most-used AI app. The team is also busy shipping new features.

We wanted to test this in practice right away.

Use case: As always at the beginning of the year, the German part of the AInauts are upset about the Pre-determined Flat Rate Tax (taxation of funds & ETFs).

Here ChatGPT also proves its good humor and helps to cope with all this:

And hey, the CIA has had all our data for years anyway. So we prefer sending data to the US instead of China. 😉

We made it! But no need to be sad. The AInauts will be back soon, with new stuff for you.

Reto & Fabian from the AInauts

P.S.: Follow us on social media - that motivates us to keep going 😁!
Twitter, LinkedIn, Facebook, Insta, YouTube, TikTok

Your feedback is essential for us. We read EVERY comment and feedback, just respond to this email. Tell us what was (not) good and what is interesting for YOU.