Free training for the way - MindStudio Workshops & Videos

Learn how to embed your MindStudio apps, launch custom functions, and deploy to Slack with free videos

We’re pleased to confirm all known quirks post-launch have been fixed, and our users are shipping like crazy! MindStudio recently surpassed 50,000 apps on the platform, nearly 40,000 in March alone, marking an incredible process in the last 3 months.

This week, we witnessed the debut of a new LLM open-source model by Databricks, an emotionally intelligent model by Hume, and the imminent arrival of Grok 1.5. Additionally, the collaboration between Microsoft and OpenAI on a $100 billion supercomputer continues to unfold.

What a week! Keep reading to learn more about the latest in tech and to find out more about our new pricing & packaging.

AIs of the Week

Submit your AI by replying to this email!

New Guides for Pros

🗞️ Industry news

  1. Hume AI releases first emphatic AI
    Hume is an innovative new model built to understand and convey human emotions. It’s able to understand your tone of voice, reply with natural sounding voices, and help with all sorts of tasks.

    For the first time, Hume’s focus wasn’t on raw IQ levels or completion time, but rather on how helpful the model feels when used for emphatic interactions. This might be the highest EQ model together with Pi by Inflection AI (now mostly owned by Microsoft).

    You can try out the model here for free.

  2. Databricks’ DBRX might be better than Mistral (and is open source)

    Databricks released DBRX, a model that should surpass GPT-3.5 and rival Gemini 1.0. Inference is up to 2x faster than LLaMA2-70B, and DBRX is about 40% of the size of Grok-1 in terms of both total and active parameter-counts. When hosted on Mosaic AI Model Serving, DBRX can generate text at up to 150 tok/s/user.


    The company open sourced both the foundational model and the weights. We’re considering adding DBRX to MindStudio!

  3. Grok 1.5 to release soon, Musk claims Grok 2 will beat all current LLM models in benchmarks
    Elon Musk has a long history of promising groundbreaking innovation at breakneck speed, and this week was no outlier.

    X AI revealed Grok 1.5 together with industry-leading benchmark results. The model beats Claude 3 Sonnet, Mistral Large, and even GPT-4 for one of the tests.


    Grok 2 is supposedly in training. This comes just a couple of weeks after Grok 1.0 went open source, which means 1.5 and 2 might also be open sourced in the future.

  4. The Information reports $100b supercomputer project “Stargate” by Satya Nadella (Microsoft) and Sam Altman (OpenAI)
    AI requires an absurd amount of compute. At least for today. GPT 4 cost as much as $100M to train, using over 8,000 NVIDIA GPUs.

    Also, it takes months.

    Microsoft is looking to up their ante on AI development and invest $100b in building Stargate, one of the largest datacenter projects in the World. A spokesperson for Sam Altman reported the news to The Information, but there’s no official confirmation from the two yet.

🔥 Product Updates

Last week, MindStudio focused on fixing post-launch quirks. These issues should now be fixed:

  • Claude 3 issues: We got reports of people being unable to use Claude 3 Haiku, Sonnet, and Opus. The models should now work as expected, but please keep reporting if that’s not the case;

  • Data sources are back at work: All your RAG processes are now back to normal;

  • Access issues: your apps no longer need to be public to be embedded or used through API. As long as your app is published, that’s enough.

We also shipped a plethora of quality of life improvements to meet the requirements of our customers:

  • Compact tabs: our new tabs are smaller, more modern, and fluid;

  • Budget transparency: the app will now show how much money you have in the workspace at all times during the building process (in the editor), the workspace selector (dropdown menu), and in the usage tab;

  • Bug bounties: if you find a bug and report it with the app ID, video of the issue with console and debugger open, and link to the app, our team might consider a bug bounty. Please note the payment amount highly depends on the severity of the issue and the number of people impacted by it;

  • Logging: we brought back logging! Global and one-shot logging are now shown in the app dashboard;

  • Remixing: we brought back remixing, but it’s now “copy”. When you open the link of a public AI, you can click on the question mark at the top right corner → copy to have it on your account. Please note the app will start consuming your workspace credits once cloned.

In the coming weeks, we will also released a desktop app, budget control features, and improve our invoice generation. We’re working hard on our certifications as well to launch the Enterprise plan ASAP - stay tuned!

💡 Tip of The Week

Remember RAG does NOT use AI by default. When you add a query data source block in your workflow, this is what it looks like:

It’s calling a data source, saving the result in a variable, and retrieving a certain number of results, usually 1 to 3.

The query template is what instructs the search. Picture this as a google search box facing your datasource.

If your datasource is a list of 50 places in Italy, your query data source has to be a search for something in those 50 places and their metadata. For example, the query could be “pizza”, and that would find all the pizzerias out of the 50 places.

The query data source looks for contextually relevant results in a given database using the query template at starting point. It does not use the LLM, hence you can’t instruct it to do anything other than retrieving a set number of results and saving them in a variable.

Do not add instructions like “do this with ”. It won’t do anything, just like instructing Google’s search box doesn’t do much either.

🤝 Community Events

If you’re looking to meet our Product Manager and group up with community members, you can join our Discord Hangout on Friday, April 5th @ 3pm EST. Link in Discord.

Our in-house MindStudio expert Giorgio Barilla will host a series of workshops over the next 3 weeks.

Register to the ones that look interesting on the blog. We’ll upload the recordings for all of these in a future cohort.

🌯 That’s a wrap!

AI is moving faster than usual, and we’re happy to be part of this groundbreaking revolution. Stay tuned to learn more about what’s next and get tips & tricks for your MindStudio build.

You saw it here first,

Giorgio Barilla
MindStudio Developer & Project Manager @ MindStudio

How did you like this issue?

Login or Subscribe to participate in polls.