The reasoning models coming for o1

Read time: under 4 minutes

The era of blindly throwing compute at AI might be coming to an end, replaced by a smarter, more surgical approach. Newer startups and legacy tech giants alike are starting to find that with reasoning-focused models, less can indeed be more.

Today’s Insights

• Today in AI: Experimental audio, Luma’s triple whammy, and Gen Z trends

• Tutorial: How to master public speaking using ChatGPT voice

• OpenAI’s o1 faces fierce competition

• Everything else you should know today

• 5 new AI tools to boost your productivity

• AI-Generated Images: Meowtini time

TODAY IN AI

1. Nvidia unveils ‘Swiss Army knife for sound’: Fugatto, a first-of-its-kind model from Nvidia, can turn text prompts into sound effects, voices, music — and even audio it was never trained on.

• You can ask for multiple sounds within the same prompt, too — say, birds singing during a thunderstorm

• Or, get weird and “make a trumpet bark or a saxophone meow”

• Since the experimental platform “generates sounds like humans do,” it can pick up on nuances like accent and emotion, or even isolate different instruments within a song

 2. Introducing the Model Context Protocol: Anthropic introduced a groundbreaking standardization system that’ll make it easier for models to interact with third-party data and software, like letting Claude instantly connect to GitHub. This could fix many current headaches about LLM apps - allowing users to bypass having to build custom integrations for every data source they want to include.

3. Fan favorite video generator gets an overhaul: Luma Labs’ popular Dream Machine just got a new UI update, an image generator, and a mobile app, all on the same day. You can now ask the model to incorporate specific colors and textures from reference images, or use the same characters across multiple generations. Plus, a “brainstorm” feature lets you pick from multiple style presets like “Art Deco meets sci-fi.”

4. Gen Z embraces the AI revolution: Just when it looked like the AI boom had reached a fever pitch, Gen Z stepped into the mix. A recent poll by Alphabet's Workspace app shows that 93% of Gen Z professionals — including 82% of young adults in leadership positions — are using AI at work. Not everyone’s thrilled though — some companies are pushing back, banning or restricting AI to address its potential side effects.

PRESENTED BY TAPLIO

Automate Your Entire LinkedIn Content Strategy with Taplio's AI

Taplio's AI, trained on over 500 million LinkedIn posts, does the heavy lifting to help you grow your influence. In just a few clicks, you can:

• Generate optimized post ideas and copy tailored to your audience

• Select the perfect format and hook for maximum engagement

• Edit the AI-generated content to match your brand voice

• Schedule posts at peak times for optimal reach

All in one platform. No more juggling multiple tools.

"Taplio is the most complete solution in the market for anyone looking to grow on LinkedIn.” — Justin Welsh, 658k LinkedIn followers and $2.39M in platform revenue

Start your free Taplio trial today and take the effort out of LinkedIn content creation. 

Black Friday Offer: 50% off any of our monthly plans for the first 6 months, with the code SUPER6

THE AI ACADEMY

How to master public speaking using ChatGPT voice

• Download the ChatGPT app on your phone and log in.

• Now go to Settings > Personalization > Customize ChatGPT > Enable

• In the textbox below, add how you want ChatGPT to respond.

Sample prompt: In voice chats, just reply with 'hmm' to everything I say until I finish the sentence with the word 'Done'

• Click on the save button and tap on the mic icon near the textbox to activate voice mode and start speaking.

Sample prompt: You are my live audience, and I’m practicing my speech on [enter topic]. After each section, please engage as an attentive audience member by asking insightful follow-up questions to clarify or deepen your understanding of the content. Also, provide feedback on my delivery, focusing on aspects like pacing, tone, clarity, and audience engagement. Share practical suggestions to help me refine and enhance my presentation for maximum impact.

• Now deliver each section followed by ‘Done’ and wait for ChatGPT’s feedback.

• You can customize the prompt and add specific details to make it more accurate.

FROM THE FRONTIER

Why reasoning models could be the answer to the scaling wall

OpenAI’s next foundation model will reportedly cost not millions, but billions to train. Not everyone has that kind of money — but that limitation might be the very thing that helps startups find a better way forward.

It costs far less to build a reasoning-focused model like o1, compared to more general-purpose LLMs. That’s because you don’t need vast amounts of data — just a small batch that’s been fine-tuned to boost a model’s coding and logic capabilities.

With that in mind, everyone from Chinese e-commerce giant Alibaba to California-based newcomer Fireworks AI have released their own chain-of-thought LLMs over the past few weeks, the Information reports. They’ve quickly closed the gap too, beating o1 across multiple benchmarks.

Alphabet, meanwhile, has a team of 200 working on its own o1 equivalent. The next big hurdle: While reasoning models are cheaper to build, they can cost up to six times more to run.

PRESENTED BY INNOVATING WITH AI

Want to become an AI Consultant?

Innovating with AI just welcomed 200 new students into The AI Consultancy Project, their new program that trains you to build a business as an AI consultant:

• Tools, frameworks, and a 6-month plan to build a 6-figure AI consulting business

Click here to request early access to The AI Consultancy Project

AI & TECH NEWS

Everything else you need to know today

👥 Clone Mode: Coding platform Replit will now let you take a screenshot of an app, then create your own version with AI.

👀 Coding Help: Eagle-eyed users noticed Gemini added a new feature that lets you upload a folder containing a code base and ask the model questions about it.

📚 Book Boom: A new startup called Spines says it’ll use AI to publish at least 8,000 books in 2025. With AI proofreading and distribution, authors can expect to see their manuscripts transformed into published titles in as little as three weeks.

📊 Insight Engine: Pyramid Analytics, which uses AI to help companies make sense of their data, raised $50M in funds from investing giant BlackRock.

🐿️ Squirrel Saviors: British conservationists are using AI to save endangered red squirrels. Their model lures red squirrels to food sources while cutting off access to their invasive gray archenemies, with 97% accuracy.

PRODUCTIVITY

5 AI Tools to Supercharge Your Productivity

✅ Aithenticate: Achieve greater compliance regarding AI regulations and ensure transparency for your users and readers, all with an easy-to-use plugin for your website.

✅ ElevenLabs: Structure, edit, and generate long-form audio with precision. Turn books into audiobooks and scripts into voiceovers.

✅ Micro1*: Access 5,000+ pre-vetted software engineers and make a hire in less than 48 hours with payroll, benefits, and compliance handled on autopilot.

✅ Layer: Visualize your goals and simplify tasks and projects with AI-powered mind map trees.

✅ Futudo: Reduce decision anxiety with an AI platform that learns about you over time and suggests actions relevant to your personal situation.

^{* indicates a promoted tool, if any}

PROMPT OF THE DAY

Thanksgiving Prep

Prompt: Act as a professional chef and help me plan a unique Thanksgiving menu with a mix of traditional dishes and creative twists. I'm hosting Thanksgiving dinner for [number] people. My kitchen is [small with limited equipment / big and well-equipped]. I want to cook [simple / extravagant] dishes, and I'm comfortable spending [number of hours or days] preparing. The ingredients I currently have are [insert list of ingredients]. Include tips for prepping ahead, wine pairings, and vegetarian options.

_{Source: Tahnee Perry / Medium}

AI-GENERATED IMAGES

Meowtini time

Prompt: A long-haired white cat lies on an antique blue and gold velvet sofa. Next to it is a table with a martini with an olive in it. The walls are pink. There is the edge of a gold frame on the wall. Impressionist, brush strokes, impasto, american tonalism, low detail
--ar 7:5 --v 6.1

Acquire new customers and drive revenue by partnering with us

Superhuman is the world’s biggest AI newsletter for businesses and professionals with 800,000+ readers and 1.5 Million followers on socials working at the world’s leading startups and enterprises. Companies like Amazon, Hubspot, and Salesforce feature their products in Superhuman. You can learn more about partnering with us here.

🧞Your wish is my command

Got more feedback or just want to get in touch? Reply to this email and we’ll get back to you.

Thanks for reading.

Until next time!

Zain & the Superhuman AI team