Project Strawberry Unleashed

+ Your own notes, the hottest new podcast?

Mindstream Banner

PRESENTED BY

Ah! It’s Friday the 13th! But why do we freak out over 13? It could be down to Norse mythology. Supposedly, 12 Norse gods were having a dinner party in Valhalla, when the trickster god Loki turned up as a 13th guest, and tricked Hoder into killing Baldur with a mistletoe-tipped arrow, symbolising 13 as unlucky forever after.

What’s in store:

  • OpenAI finally unleashes project strawberry

  • NotebookLM, Google’s tool is here to make your notes into a podcast.

  • Around the web: Steam Families lets up to six members share and play games simultaneously.

  • Why Guidde can increase your productivity!

Read Time: 5 minutes

OPENAI

OpenAI has introduced its latest AI model family, OpenAI o1, available in two versions: o1-preview and o1-mini.

Code named Project Strawberry, these models are accessible to ChatGPT Plus and Team subscribers, with enterprise and education users getting access soon.

At present, o1 is limited in functionality but promises supreme reasoning skills.

It can't browse the web or analyse files yet, and image-analysing features are still being tested.

Users are restricted to 30 messages per week for o1-preview and 50 for o1-mini.

Costs are also high: $15 per 1 million input tokens and $60 per 1 million output tokens, significantly more expensive than earlier models like GPT-4o.

Here’s what you should know:

  • OpenAI o1 is designed for better reasoning, excelling at complex tasks like coding and legal analysis.

  • Its high cost—$15 for input and $60 for output per 1 million tokens—is a notable drawback.

  • o1 can be slower and sometimes produces errors or "hallucinates" answers.

Trained harder than rocky

OpenAI has trained o1 with reinforcement learning, allowing it to handle more complex tasks by "thinking" before responding.

This makes it useful for things like legal briefs or brainstorming strategies.

However, it's not without issues. Due to more reasoning, o1 can take over 10 seconds to answer, and it can still make mistakes - more so in fact.

OpenAI admits it tends to "hallucinate" more than earlier models like GPT-4o.

The competition is fierce, with Google DeepMind showing improvements in reasoning with similar techniques.

OpenAI plans to keep developing o1, aiming for models that can "think" over longer periods to improve accuracy.

10 seconds to answer, it’s like it thinks before it talks.

Is ChatGPT o1 a step forward?

Login or Subscribe to participate in polls.

Vote for live results and see results + opinions from yesterday at the bottom of the email.

PRODUCTIVITY HACKS

Ever wish you could clone yourself to answer the neverending repetitive questions?

Well, with Guidde, you can!

Guidde is an AI-powered tool that can turn your instructions and explanations into neat, easy-to-follow videos in seconds.

It’s like having a digital assistant who never gets tired of explaining the same old thing for the nth time.

Whether it’s showing your team how to navigate a new software or helping customers get the most out of your product, Guidde makes it happen in record time.

Here’s how Guidde gets it done:

  • Turn boring explanations into dynamic video tutorials with just a few clicks

  • Keep your team on the same page with branded, consistent content

  • Update and edit videos with brand-new features like intros and outros

Whether you’re training new hires, or keeping your customers in the loop, Guidde does the hard work so you can focus on other things.

Trusted by 20,000+ users and rated 5 stars on the Google Chrome store, it’s the go-to tool everyone should have.

Why not give it a try? It’s free!

On This Day

September 13, 1845 - English chemist Michael Faraday discovers the 'Faraday effect', demonstrating the influence of a magnetic field on polarised light, a foundational experiment in the field of electromagnetism.

Riddle Me This

I’m light as a feather,

Yet even the strongest person can’t hold me for more than 5 minutes.

What am I?

Can you guess it right?

Login or Subscribe to participate in polls.

Nifty Numbers

From Cranks to Touch Screens

Telephones have come a long way since the boxy, wooden devices of 1876. From the rotary dial of the 1920s to the cordless phones of the 80s, the Motorola StarTAC in the 90s, and the iPhone in 2007, each leap brought us closer to the sleek - and sometimes foldable - smartphones we use today. The evolution of the telephone is a testament to how far technology has taken us in staying connected.

GOOGLE

Google's latest feature lets you turn your research into an AI-generated podcast, complete with two virtual hosts discussing your findings.

This feature, part of NotebookLM, Google’s AI note-taking app, uses the Gemini AI model to summarise your research and add a conversational twist.

While it makes research more engaging, though users are reporting that the results can be a bit… strange.

For instance, when discussing the invention of the lightbulb, the AI hosts used phrases like “messy as heck” and called platinum “bling bling metal.”

This informal tone may be entertaining but doesn’t suit more serious topics.

The takeaways:

  • AI hosts summarise your research in a podcast format.

  • The tone can be informal, and therefore may not fit all topics.

  • The feature has limitations, such as slower generation time and being English-only.

The future of study breaks

Google warns that this feature is “not a comprehensive or objective view of a topic” but simply reflects your notes.

It can take a few minutes to generate, and like many AI tools, it's not always accurate.

To try it, open NotebookLM, choose a notebook, and load “Audio Overview” from the bottom-right menu.

It’s a fun way to hear your research discussed.

Undeniably cool - but useful? We’re not convinced.

Mindstream Picks

Space: The FAA defended its launch licensing approach with SpaceX, stating that timelines depend on choices made by the company.

Crypto: U.K. authorities plan to classify cryptocurrencies as personal property under a new bill, granting legal status to digital assets like NFTs and tokenised assets.

Games: Microsoft Gaming will lay off around 650 employees, 3% of its workforce, primarily affecting corporate and support roles, following its Activision Blizzard acquisition.

Music: Jon Bon Jovi was praised by Nashville police for helping talk a distressed woman off the edge of a bridge, calmly engaging with her before helping her back to safety.

Don’t Miss - Valve has officially launched Steam Families, allowing parents to share games, manage parental controls, and approve purchases for their children. Unlike the old Family Sharing, multiple users can now play different games from a shared library at the same time. Steam Families supports up to six members, each with their own saved data and achievements.

AI Art

Our Image of the Day

Artwork submitted by Mindstream reader Slicer D: “A futuristic car driving in the 50s”

Daily Image Prompt

Titans from Greek Mythology

Yesterday’s Poll

“Is Gov-GPT the future of government services?”

Yes, can't wait. - 56%

I wouldn't trust it. - 44%

Your Views:

“The government is not equipped with secure services that are also reliable, at this point.” - pjay

“I’m not saying that it’s not an effective tool. It just is funny How everyone is just trying to repackage ChatGPT and present it as if they are some kind of heroes. Of course institutions are going to use AI. AI is just a tool, Hi chose I wouldn’t trust it Because it depends on who’s programming and employing that tool.” - coosbaydesign

Submit your opinions in our daily poll to be featured!

Collage of Authors

We've reached the end, weekend coming in hot!

❤️ We need your feedback to make our newsletter better.

📣 Refer our newsletter to your friends and make us feel all warm and fuzzy.

🚀 Advertise in our newsletter to reach 150,000+ founders, engineers, and content creators.

Hey You! Would You Recommend Us?

Login or Subscribe to participate in polls.