• AInauten.net
  • Posts
  • πŸ‘¨β€πŸš€ This is the most ingenious AI browser!

πŸ‘¨β€πŸš€ This is the most ingenious AI browser!

PLUS: Strategic move in the Valley – Zuck wants to give it another go

Hello AInauts,

Welcome to the latest issue of your favorite newsletter!

We only wanted to touch on this new AI browser "briefly" - and then we went down the πŸ•³οΈπŸ‡ … But almost like Alice, we discovered new worlds.

Here's what we have in store for you today:

  • πŸ”₯ Ingenious! New AI browser finally makes AI-first workflow a reality

  • πŸ’° Fully hijacked: Meta snatches up a data pipeline

  • 😁 AI-Fun: Prompt Engineering 2.0

Here we go!

πŸ”₯ Ingenious! This new AI browser finally makes the AI-first workflow a reality

Our work at the computer often consists of research, content generation, communication - and of course lots of practical tests. And we want to live "AI-First!" as much as possible.

But if we are honest: despite agents, automation and MCPs, there are still many times we have to copy-paste and switch context...

Genspark aims to solve this problem. If you don't know Genspark yet, it's an "AI Agentic Engine" that lets agents do the work for you.

The latest highlight: the Genspark browser has just been released. Instead of just surfing passively, an AI actively works for you in the background. We took a closer look at the browser and subjected it to an in-depth practical test.

In a nutshell, it's about the future of human-computer interaction. Whoever comes out on top will control how we interact with the internet!

The battle for the AI browser, or the future user interface for all of us, is raging.

  • The browser company released the Dia browser last week (currently free, for Mac only),

  • Google introduced Project Mariner (and started the rollout as part of the Ultra plan starting at $125),

  • Microsoft has integrated Copilot into Edge,

  • Perplexity teased us with Comet (we've been on the waiting list for months - hopefully the Genspark release will ensure a public beta soon),

  • OpenAI has released the Operator (in the $200 plan),

  • Anthropic provides Claude Computer Use,

  • ChatLLM has the Deep Agent,

  • there is the Strawberry browser, etc.

Genspark has the advantage of having an arsenal of really helpful tools built-in - and with this release, it has a first mover advantage over other providers that are either not as good, not as fast or much more expensive.

It's an AI-Swiss Army knife!

Eric Jing, the Genspark CEO, puts it in a nutshell: "Speed is everything!" and "...we believe the winner won't be determined by who has more money - it'll be who innovates fastest." We love this spirit!

What can the Genspark browser do with an AI-first workflow?

With Genspark, the integrated "Super Agent" acts as the interface between you and the browser. This AI chatbot offers context-related help on every website and can also perform actions directly for you via the Playwright framework (see screenshot) from Microsoft and various MCP servers.

MCP is the new standard for linking AI with all kinds of applications. There are lots of ready-made integrations waiting to be tried out.

Genspark Agent works in the background

You can also install and use any Chrome extensions. This makes switching from other browsers very easy.

The browser is currently only available for Mac OS (Windows version coming soon), and you need a paid subscription starting at 25 dollars a month to use the full functionality.

Genspark Browser is privacy-first and blocks ads automatically. The privacy-oriented Brave search engine is also used in the background (it can be customized if desired - we have switched to Perplexity 😎).

While you are on a website, the integrated AI continues to work in the background in other tabs and completes tasks for you. By the way: the browser was developed by 1 employee in 3 months!

The browser can perform all kinds of tasks autonomously:

  • Carry out searches,

  • create a presentation,

  • analyze Gmail and prepare emails,

  • create tables,

  • generate images or videos,

  • make summaries from X or YouTube,

  • create a podcast from it,

  • compare products and prices or find the best deals, …

Our practical tests: Everything works as promised, very nice πŸ‘Œ

The setup is straight forward - install the browser, log in and get started. You can install your password manager extension so that you have all your logins available. We would have liked an automated import workflow for browser settings - but that's not so bad, you can do it manually via the URL genspark://settings/importData.

Then the tests started straight away:

  • Create a presentation in German from a YouTube video βœ…
    ("10 versteckte ChatGPT Features, die du nutzen solltest")

  • Make a summary of your favorite tweets βœ…

  • A research with a website for Art Basel 2025 βœ…

  • Find cheaper products at Amazon Save & Subscribe (βœ…)
    (the products were cheaper, but not close enough to the original)

  • Analyze product evaluations and background information on the effect βœ…,

  • create a review page from it βœ…

  • Find older conversations in Gmail based on an email and research what the person is currently doing βœ…

  • Compose a reply to an email and save/send it directly in Gmail βœ…

  • Set up a Google Analytics view based on instructions (βœ…)

  • Research a topic and write an article about it βœ…

  • Check and supplement facts βœ…

But things got really cool when we logged in to Lovable. Then we gave Genspark the task of further developing a mini-app that had us going round in circles a few weeks ago. The agent started the conversation straight away! 😍

It is important that you tell the agent exactly what to do ("Navigate to https://... - you are already logged in. Then follow the instructions from ... step by step.")

The Genspark agent interacts directly with the Lovable Chat to fix our app!

Little things that can be improved

Of course, there are some minor things that can be improved:

  • One issue was that the context-sensitive chat was occasionally lost on some pages when we navigated to a different page. However, the conversation itself could be found at any time via the super agent interface and continued there or shared with others.

  • We were also unable to log in directly to ChatGPT and Claude - that would of course be nice, although not mandatory (the models themselves are also available in Genspark, just not our history).

  • Since the agent has to analyze every single action that it performs autonomously with a screenshot, it can take longer for something to be done, depending on the complexity. We had to abort our attempt to unfollow inactive X users at an early stage.

  • And we would also have liked a hint when a tab/agent needs help from us.

  • Genspark can be tested for free, but if you want to use the browser, you have to budget $25 dollars per month - and power users will probably not quite make ends meet with the 10k credits. Get 1000 extra credits here.

But all in all, these things are really minor - Genspark has once again impressed us!

Our take: Agents in the browser are the next logical step...

The Genspark browser shows where the journey is heading: away from passive consumption and towards an active AI partner that thinks and works with you! Instead of juggling browser tabs, you have an intelligent assistant that optimizes your online activities and makes you more productive.

Genspark is therefore a veritable alternative to ChatGPT and the like: a chatbot with the best models, image, video and audio generation, presentations at the touch of a button, an AI secretary that coordinates appointments or drafts emails, an AI drive that searches for or compiles files, a sheets function that researches information and packs it into tables, and much more.

And thanks to the MCP integration and the use of browser extensions, practically all tools can be docked directly and controlled via an interface. Amazing, isn't it?

Under the hood, there is a collection of different tools and models that are nothing special on their own (see also the leaked prompts from Manus and Co.) - but in its entirety as an integrated application, they represent a truly productivity-enhancing tool.

For $25 a month, you get an AI assistant that noticeably increases your productivity. The learning curve is minimal, the gain immense - the only sticking point: you have to rethink your habits and be prepared to actually hand over tasks to the browser.

πŸ’° Fully hijacked: Meta snatches up a data pipeline

Zuck is serious and is acquiring exactly 49% of Scale AI for 14.3 billion dollars - the data labeling king of the industry, which has so far fed OpenAI, Google and co. with clean training data. Meta thus secures access to the most important raw material of the AI era: first-class annotated training data!

This gives Meta the say without officially taking over the majority: 49 percent is enough influence for a board seat and strategic control, but too little to alarm the regulators.

Wang in, LeCun out?

But the data is only one side of the coin. Probably the most interesting move: Alexandr Wang, the 28-year-old founder of Scale, is moving to Meta!

His new mission: to set up a "Superintelligence Lab" with a direct line to Zuck and GPU blank checks included. The background: the race towards AGI, the general superintelligence.

Chief AI Scientist Yann LeCun is staying on, but his FAIR lab is bleeding dry: 11 out of 14 Llama first authors have left for Mistral, Anthropic, Microsoft & Co.

The internal power struggle between LeCun's open source vision and Wang's "data is power" mantra is imminent, and we are excited to see which way it goes!

via GPT-Image in Genspark

LeCun has been preaching the open source gospel so far. But Scale's customers naturally don't like the fact that Scale is now so close to Meta. Google was Scale's largest customer in 2024 with 870 million dollars - and is already planning to exit, while OpenAI and Microsoft want to wait and see.

Our take: Zuck’s move shakes up the Valley!

Meta has a trust issue at the moment. In April, they were caught scoring points in benchmarks with a non-public "Maverick" version. In the community, this smelled of cheating. The denial came promptly, but was hardly convincing. And they are once again in the headlines because of privacy issues with their Meta app...

This deal buys three things at once:

  1. Data: Scale's premium pipeline, fed by its competitors

  2. Talent: Wang and his team, now working exclusively for Meta

  3. Narrative: Superintelligence instead of social network

If Wang delivers and LeCun pivots his tweets- sorry, threads - towards the future instead of the past, Meta could actually make up ground.

Until then, it remains the best reality show of the AI summer. On that note, here's a fun fact: Scale AI co-founder Lucy Guo left the company back in 2018, but kept 5 percent. At 30, this makes her the youngest self-made billionaire.

via ChatGPT

😁 AI-Fun: Prompt Engineering 2.0

True ...

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 - it motivates us to keep going 😁!
X, 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.

🌠 Please rate this issue:

Your feedback is our rocket fuel - to the moon and beyond!

Login or Subscribe to participate in polls.