🧙🏼‍♂️ The AI browser reinventing search

This week’s issue is brought to you by:

Stay up-to-date with AI.

The Rundown is the world’s fastest-growing AI newsletter, with over 500,000+ readers staying up-to-date with the latest AI news, tools, and tutorials.

Our research team spends all day learning what’s new in AI, then distills the most important developments into one free email every morning.

Sign up today.

 Dario’s Picks

I. AI-native web browser Arc gets two powerful new features

Arc is a browser with several innovative AI features that is taking the world by storm. It recently made Perplexity its default search engine. And a couple of new powerful features are launching in the next week. Watch the demo or read my breakdown below.

Instant links: Tell the search bar what you’re looking for, and it instantly launches relevant tabs (without overloading you) directly in the browser. Imagine Google but you don’t have to go through the “middle man” page of selecting sites. You can also ask it to compile results into a folder, right inside the prompt, e.g. “folder of Apple Vision Pro reviews”.

Live folders: a live-updating stream of data from wherever you want. Sounds like an RSS feed with superpowers.

Why it matters: First of all, kudos for featuring Todd Terje - Inspector Norse in their demo video at 07:48. How good is that song!?

Arc and its maker The Browser Company is a force to be reckoned with: a visionary AI-native, alternative to Google Chrome with great UI and onboarding, and also brilliant marketing. To me, it seems like their product is gradually turning into what Google has been trying, but lacks the vision to do, with SGE.

If the search results (the stuff that opens) in Arc when using instant links actually makes sense, this could be a big improvement to the user experience of search, both in terms of speed and taking some of the decision making out of the picture.

II. Bard now has Gemini Pro globally and can generate images

In case you missed it, Gemini Pro is Google’s new language model and is equivalent to about the performance of GPT 3.5. The model now powers their chatbot Bard globally, meaning you can get a solid alternative to ChatGPT in more languages.

Bard can now also use Google’s Imagen 2 model to generate images, similar to how ChatGPT has DALL-E integrated. The images are also watermarked using a tool named SynthID. Haven’t tested the image gen feature myself yet, but apparently it’s decent, offering higher image resolution and a less rendered look than DALL-E.

Apparently, Bard is also in for a big rebrand very soon: an early changelog reveals that it will rebranded to just “Gemini” next week, and also a new Android app will be launched.

Why it matters: The competition to stay relevant in the chatbot world with the fast pace of innovation in the cross-section of AI and UX is intense to say the least. Google is gradually catching up to ChatGPT, but is still way less performant, so it will be interesting to see what Gemini Ultra will be like (it’s claimed to beat GPT-4 and will be part of a paid version of Bard).

 In Focus 

@-mentions 101

Dan Shipper published a neat walkthrough on how to use the @-mentions inside ChatGPT last week. He used an example of writing a journal entry in ChatGPT and, maintaining the context of the conversation, calling a GPT that connects to Notion and save the entry there.

I’d say @-mentions are really a solid improvement when it comes to the user experience of GPTs. The biggest differences for me so far has been that now I don’t have to “plan” on using a GPT. I can just call it in the middle of the chat and it already knows the chat context. And after I’ve used it, I can just exit and return to default ChatGPT (or switch to another GPT). As the video also points out, one of the annoying things currently is all the “confirm” steps that keep appearing, but this is likely to get improved over time.

GPTs  

💬 = total chats, ↑ = chats last 7 days

Editor’s Choice

Weekly picks from me to you

Explains and visualizes concepts with Whimsical Flowcharts, Mindmaps and Sequence Diagrams.

❞ This GPT quickly makes a chart about any concept – making it potentially easier to understand for you and easier to communicate to others. Try it the next time you’re stuck on understanding something inside ChatGPT; @-mention Whimsical Diagrams and ask for a visual.

Tries to predict future stock market prices. Not financial advice.

❞ There’s a lot of interest in using LLMs to predict the stock market. The rapid popularity of this GPT is a case in point. Is it totally accurate? Probably not. Would I trust it blindly for an investment? Definitely not. Could it be helpful as a supplement to other research and information sources? Quite possible.

Your expert frontend developer sidekick. Provides advice to coders of all levels: writing, explaining, debugging and reviewing code.

❞ Neat coding helper for frontend development. Lays out concepts and instructions succinctly.

Trending GPTs

Popular GPTs with big increases in chats in the last 7 days

1. Diagrams: Show Me ↑200K    Data Analysis

2. Cartoonize Yourself ↑200K Image Generation

3. Write For Me  ↑100K Writing

4. Professional Coder (Auto programming)  ↑100K   Coding

5. Humanizer Pro  ↑100K Writing

Up and coming

Lesser known (not featured in the GPT store) but rapidly growing GPTs

1. Ebook Creator  ↑300    Writing

2. PC Builder GPT  ↑300    Miscellaneous

3. Ranko Football  ↑200    Gaming

4. Job Description Generator  ↑200    Writing

5. WP Code Wizard  ↑200    Coding

Featured resources

• If you want to make the process of building an elaborate GPT super easy, check out Skillfusion.ai. You provide a description of what you want, and it generates instructions, knowledge base, starter questions, etc. It lets you generate 2-3 GPTs for free.

Bytes 

• AI chatbots are becoming available on Apple Vision Pro, here’s what ChatGPT looks like. I’m almost surprised it’s flat and not 3D 😄 

• HuggingFace launches assistants for their chatbot HuggingChat. The resemblance to GPTs is striking, but it can use any of the LLMs available on HuggingFace such as Llama and Mistral to make bots.

• New AI features coming soon to Google Maps, with the goal of making it more powerful for discovering new places. It’ll be able to use AI for analysing photos, ratings and reviews. Suggestions will be categorised, with photo carousels and review summaries.

• Meta released CodeLlama 70B, beating GPT-4 on the HumanEval benchmark with 67.8 (GPT-4 has 67). CodeLlama is part of an LLM series for code generation.

• Microsoft releases new Future of Work report, and it’s all about AI.

• The tiny island of Anguilla is making $30 million per month from AI domains. It got “.ai” assigned as their two letter domain in the late 80s, and it’s now paying dividends.

• OpenAI is building an early warning system LLM-aided biological threat creation.

*if you sign up for these newsletters I may earn a commission at no cost to you. If you do, THANK YOU, it will make it possible for me to continue to do this.