Welcome, humans.

Check this out: AI is rewriting our world—literally. New research reveals the most rapid transformation of written communication EVER in history, with LLMs now assisting in writing across every sector of society. 

Apparently, by late 2024, ~18% of financial consumer complaints, ~24% of corporate press releases, and ~15% of job postings from small companies showed clear signs of AI assistance. And if you know what to look for, you can tell the real numbers are MUCH higher.

After ChatGPT's release, there was a 3-4 month lag before adoption exploded across all domains, then stabilized by late 2023. And as Ethan Mollick shared, these figures likely undercount actual usage, as increasingly sophisticated outputs can evade detection (and many detectors straight up don’t work). 

These numbers are just the tip of the iceberg. As commenter put it, “LinkedIn is ruined” with AI-generated corporate speak too, and many middle management memos probably are as well. Soon, we might need to archive “natural English” like it’s an endangered species. Isn’t 2025 fun?!

Here’s what you need to know about AI today:

• Sesame demonstrated a freaky new voice AI.

• Anthropic raised $3.5B while Murati's startup eyed $1B.

• Perplexity, Deutsche Telekom plan a sub-$1K AI phone for 2026

• AI chipmaker TSMC to spend $100B+ on US AI chip facilities.

Sesame's new AI voice tech finally sounds like a real person talking to you

There’s a new voice AI that’s freaking out the internet. Imagine a voice assistant that doesn't just respond to your questions but actually sounds like a human—pausing thoughtfully, expressing excitement, or offering warm reassurance. 

That's what Sesame (created by Oculus co-founder Brendan Iribe) is building with its new AI voice technology. The model, called the Conversational Speech Model (CSM), is available to try here via a demo of voice models Maya and Miles—who just might be the most human-sounding AI voice we've ever heard.

This isn’t your basic text-to-speech model. What makes Sesame's approach different is their focus on what the company calls “voice presence”—making spoken interactions feel genuinely real and understood. 

According to Sesame's technical team, their breakthrough comes from treating speech generation as a contextual problem rather than just a text-to-audio conversion.

Basically, the model understands:

• The emotional context of the conversation.

• Natural timing, pauses, and emphasis.

• When to adjust tone to match the situation.

• How to maintain a consistent personality.

The Verge's Sean Hollister described it as “the first voice assistant I've ever wanted to talk to more than once,” noting how it handles natural conversation in ways that leave other voice assistants feeling robotic by comparison.

Here’s a few of the wildest examples: 

1. Check out this convo that’s genuinely hard to tell who is AI and who is human. 

2. Or Miles offering advice to take a break from emails on the weekend. 

3. Or Maya becoming a D&D character on the fly.

4. Or someone already falling for Maya like Joaquin Phoenix in the movie Her

Sesame's model also performed impressively in benchmark tests:

• Near-human accuracy on traditional speech metrics.

• 90%+ accuracy on correctly pronouncing words like “lead” (as in metal) vs. “to guide.”

• Strong performance on consistent pronunciation (saying words like “route” the same way across a full conversation).

The most impressive stat: in blind listening tests without context, human evaluators often couldn't tell the difference between Sesame and actual human recordings.

What’s next: Sesame is planning to open-source key components of their research under an Apache 2.0 license, making it available for developers to build upon. The roadmap includes:

• Scaling to larger models.

• Adding support for 20+ languages.

• Integrating with pre-trained language models.

• Building models that can naturally manage conversation dynamics.

Our take: Voice interfaces like ChatGPT’s Advanced Voice Mode or Google Gemini’s Voice Assistant still feel a bit too robotic and awkward—they're useful, but not natural enough to become our primary way of interacting with technology (we still prefer the control of typing—for now).

Sesame (or more models like it) could change that, bringing us one step closer to talking to our devices just like we talk to each other.

FROM OUR PARTNERS

Here’s one AI tool we use every week at The Neuron.

A little well-kept secret for companies crushing it with AI is that only a handful of AI tools are actually worth using.

That’s why we—Noah & Grant—use ChatGPT and Claude, along with a killer product called Attention.

Here’s how it works:

1. We tell Attention exactly what to pay attention to in our meeting (goals, budget, etc).

2. Attention listens in to our call.

3. After, Attention outputs important insights, action items, and follow-up emails.

There’s plenty more features, too. So if you sell anything, literally drop what you’re doing and go book a demo with Attention ASAP. 

Agent Tip of the Day

In case you missed it, Tina Huang did a 1 hour livestream on how to use n8n to create AI agents (basically an hour-long hands on demo of her previous video here).

In it, Tina builds a simple calendar assistant that reads your schedule and blocks time for new activities. Here’s the TL;DR (you’ll still need to watch the full vid for this to make sense): 

1. Start with the simplest solution possible for your task (don't overcomplicate).

2. Learn prompt engineering fundamentals before attempting to build agents.

3. Set up your trigger point (e.g., Telegram message).

4. Create a system prompt that clearly explains the agent's purpose and available tools.

5. Define available tools (e.g., Google Calendar read/write permissions).

6. Configure your tools with proper parameters.

7. Add memory if needed to retain context.

8. Test frequently throughout the building process.

9. Refine your prompts iteratively based on your test results.

10. Implement the proper guardrails for fully autonomous agents.

She also credits David Ondrejs original video tutorial for teaching her these skills if you want another option to watch.

Treats To Try.

1. Data Science Agent is Google’s free new AI that automates your data analysis setup.

2. Teamble helps you give and receive better workplace feedback via a conversation coach that works inside Slack and Microsoft Teams.

3. Pika 2.2 lets you generate HD video clips (1080p resolution) up to 10 seconds long with “endless” transformations of content per clip (demo).

4. Chikka interviews your customers with AI voice agents, giving you deeper insights without doing the interviews yourself (free to try).

5. Currents analyzes social media discussions to deliver real-time insights about what your target audience is talking about.

6. Guse lets you automate any workflow using a familiar spreadsheet interface (demo)—free to try.

7. SmolVLM2 is a new small open source AI model you can run on your device that understands videos, images, and text—try it here (video demo).

See our top 51 AI Tools for Business here!

Around the Horn.

• Anthropic (maker of Claude) raised $3.5B at a $61.5B post-money valuation and ex OpenAI CTO Mira Murati’s Thinking Labs startup could raise $1B at a $9B valuation.

• TSMC (world’s largest contract chip producer) will spend $100B+ in the US on new AI chip facilities over the next four years.

• Perplexity and Deutsche Telekom announced a new AI phone that will sell for less than $1K when it’s released in 2026.

• Microsoft combined its Dragon Medical One dictation tool and DAX Copilot listening tool into a single tool called the Dragon Copilot, which aims to reduce healthcare documentation burden while improving patient care.

FROM OUR PARTNERS

Every day, data brokers profit from your sensitive info…

…Phone number, DOB, SSN—selling it to the highest bidder. 

The question is, who’s buying? 

• Best case: companies targeting you with ads. 

• Worst case: scammers and identity thieves. 

Incogni removes your personal data from the open internet so scammers and identity thieves can’t access it. 

Protect yourself online with Incogni—get 55% off with code NEURON.

A Cat's Commentary.