Welcome, humans.

It's Tuesday night, and you know what that means: A Star Wars themed cover of George Michael's Saxophone Masterpiece Careless Whisper!

What, that's not what Tuesday means to YOU? Weird.

In all seriousness, it's the holiday season, and we've been messing around with making holiday songs on Suno. IDK if you've done this lately, but they've gotten scary good.

Our new favorite thing is to write stream of consciousness prompts about our friends and family, listing all the weird things we know about them, and then instead of having the AI write the lyrics, have the AI use the prompt as the lyrics. Here's one for Corey!

They're actually HILARIOUS (and WAY better than they have any right to be). If you need a funny gift for your friends and family this year, you can't miss with this IMO.

To do this, go to Suno > Create > Custom > put your prompt in the "Lyrics" box.
Then, light up "Styles" with a bunch of musical vibes you want (e.g holiday, jazz, blues, but also: you can get REAL weird with it).
Pro Tip: the 1 min v5 demos you get for the first 2 prompts you submit are 🔥

Here's what happened in AI today:

OpenAI, Anthropic, and Block donate tools to new Agentic AI Foundation.
Companies spent $37B on generative AI in 2025, up from $11.5B in 2024.
ChatGPT is approaching 900M weekly users and a new CRO.
Mistral released Devstral 2 and Mistral Vibe, a terminal coding tool.

Advertise in The Neuron here!

P.S: 🔴 JOIN US LIVE tomorrow (Wednesday) @ 2pm PST | 4pm CST: Learn to speak a new language with AI (with Speak AI and co-founder Andrew Hsu).

Click the link above, then on YouTube, select "Notify Me" to get a notification when we go live.

We're diving into how AI is changing language learning: think real-time pronunciation feedback, personalized practice sessions, and conversational fluency without needing a human tutor. Watch out, Duo Lingo Bird. You're on notice.

P.P.S: Notice the time change in your inbox? We're expecting some big news announcements this week and next, so we'll be sending later in the afternoon to get the news to you in near real-time. Thanks for understanding!

AI's Most Unlikely Alliance Just Formed…

OpenAI, Anthropic, and Block just formed tech's most unexpected alliance… except instead of destroying a ring, or stopping a giant space portal from opening over New York, they're taking on an equally challenging task: building the plumbing for an AI agent era that enterprises are betting $37B won't turn into a proprietary mess.

Here's what happened: The Linux Foundation just launched the Agentic AI Foundation, and the dynamic trio mentioned above donated the core infrastructure:

What they donated:

Anthropic: The Model Context Protocol (MCP), a standard way to connect AI models to tools and data via "connectors."
Block: Goose, their open-source agent framework used by thousands of engineers.
OpenAI: AGENTS.md, a simple instruction file that tells AI coding tools how to behave in your repo.

The timing of this matters. Here's why:

Menlo Ventures just releases a new 2025 report that shows companies spent $37B on genAI this year.
That's up 3.2x from $11.5B in 2024, making it the fastest-scaling software category in history.
Today, 76% of AI use cases are purchased rather than built internally.
The most revealing stat: AI deals convert at 47% vs. traditional SaaS's 25% rate, because the productivity gains are immediate AND measurable.

Yeah, tell that to MIT (just playing)!

Okay so what does this actually mean? Enterprises pivoted from building to buying. Ready-made solutions reach production faster, but only if they work together. That's what AAIF solves.

Consider coding, which became AI's first killer use case at $4B in 2025 spending (up from $550M in 2024). Anthropic now commands 54% market share in enterprise coding spend with 18 months atop coding leaderboards. That makes open standards like AGENTS.md critical for preventing vendor lock-in.

More wild data from Menlo's report:

27% of AI application spend comes through product-led growth (where individual employees discover tools, prove they work, then convince their companies to buy).
- That's nearly 4x traditional software's 7% rate, btw! Also, Cursor apparently reached $200M in revenue before hiring a single enterprise sales rep.
The infrastructure layer (the behind-the-scenes tech that powers AI apps, like model APIs and data storage) captured $18B.
Caveat: only 16% of enterprise deployments actually qualify as true agents (most are just simple if-then workflows, not systems that actually think and adapt).

We're early. Which is exactly why AAIF matters. Anthropic's MCP co-creator David Soria Parra told TechCrunch: "We're all better off if we have an open integration center where you can build something once as a developer and use it across any client."

The skeptical take: Will AAIF become real infrastructure or just another industry logo alliance? That said, Linux Foundation director Jim Zemlin pointed to Kubernetes "winning" the container race as proof that "dominance emerges from merit and not vendor control."

If MCP, AGENTS.md, and Goose become standard infrastructure, the agent landscape shifts from closed platforms to an open, mix-and-match software world. We just need the same thing to happen for memory (so you can take your memories with you when you change providers) and preferences (how you like your tasks done, now handled by prompts and project folders, and does overlap a lot with memory, but not perfectly).

What's next: Menlo predicts AI will exceed human performance in daily practical programming tasks in 2026. OpenAI's own state of enterpise AI report found AI tools save knowledge workers an hour a day. How much time could superhuman coding agents save programmers a day…?

FROM OUR PARTNERS

Ideas move fast; typing slows them down.

Wispr Flow flips the script by turning your speech into clean, final-draft writing across email, Slack, and docs.

It matches your tone, handles punctuation and lists, and adapts to how you work on Mac, Windows, and iPhone.

No start-stop fixing, no reformatting, just thought-to-text that keeps pace with you. When writing stops being a bottleneck, work flows.

Give your hands a break ➜ start flowing for free today.

Prompt Tip of the Day

Following a random idea, we recently used Google's Nano Banana image model to enhance a poor image and were stunned with the results. Here's all we did...

Have or acquire a lousy image.
Panic over what to do about it.
Visit Gemini and toss it into Nano Banana with an uber-scientific prompt, like "Make my picture less fuzzy."
Wait.
Be stunned.

This is a trick crime drama's have been faking for decades, but now a quick trip to Gemini makes science fiction a reality.

Now, if you want to learn how to do this for bringing old photos to life via video, we put some advice together for you on the website (not a full tutorial, just more than we could fit in the newsletter today). Check it out here.

Treats to Try

*Asterisk = from our partners (only the first one!). Advertise to 600K readers here!

*Outskill teaches you AI in 2 days. It's almost 2026, and if you still haven't learned how to use AI to upgrade your career, this is your wake-up call. First 100 people get the 16-hour course free (normally $395). Register here.
Devstral 2 handles agentic coding tasks like exploring codebases, editing multiple files, and building software engineering agents, scoring 72.2% on SWE-bench Verified, and ships with Mistral Vibe, a CLI tool that lets you code in natural language from your terminal (free via API for now, on HuggingFace, or use in your IDE via this Zed extension) and use this system prompt.
Stitch by Google now generates working HTML code from its AI redesigns (previously you could only see visual mockups, now you can export the code.
Relace Search finds the exact files and code you need in your codebase 4x faster than Claude 4.5 Sonnet, combining parallel search with multi-step reasoning across 4-12 operations (OpenRouter).
Conductor runs multiple Claude Code agents in parallel on your Mac, handling the git worktree setup and giving you visual diffs to review and merge changes—no separate Git client needed (uses your existing Claude Code login, so API key, Pro, or Max plan).
Agno is an "agentOS" framework that builds agents who learn collectively and share knowledge system-wide; e.g. one agent learns to give actionable steps, then all your agents apply it (code, docs).
Essential released Rnj-1, an open-source language model focused on STEM and coding tasks (HuggingFace, run it on the cloud).
Cashew Research designs custom market surveys, sends them to real people, and uses AI to summarize the findings, turning weeks of traditional research into days at a fraction of the cost (raised C$1.5M).

Around the Horn

(LinkedIn Source)

Anthropic is bringing Claude Code into Slack as a research preview, letting developers tag @Claude in chat threads to spin up full coding sessions that read Slack context, pick repos, post progress and open PRs.
- Speaking of Slack: CEO Denise Dresser is leaving Salesforce-owned Slack to become OpenAI's chief revenue officer to steer its enterprise revenue push.
Microsoft plans to invest $17.5B in India by 2029—its largest bet in Asia—to build new AI data centers, expand cloud regions, and fund skilling programs that use Azure OpenAI to serve hundreds of millions of Indian workers.
EU regulators opened an antitrust probe into whether Google's AI Overviews and AI Mode illegally use publishers' and YouTube creators' content without fair compensation or opt-out options and shut out rivals in the AI search market.
The Information says ChatGPT is approaching 900M weekly users, but Gemini's rapid growth is narrowing the gap.
Anthropic and Accenture struck a three-year Claude partnership aimed at turning cautious enterprises into real AI spenders, with 30,000 consultants trained on Claude.
Wired says an OpenAI economist quit over concerns that the company's economic research is being steered away from frank analysis of AI-driven job loss and toward pro-AI messaging.
Amazon upgraded Alexa+ with delivery tracking, one-tap last-minute add-ons, and personalized gift ideas, turning Echo screens into hubs for watching orders, grabbing extra items before ship, and browsing holiday deals.
Google will launch its first AI-powered smart glasses in 2026, building on its Project Astra demos to compete directly with Meta's Ray-Ban line by bringing multimodal AI and on-the-go assistance into everyday eyewear.

FROM OUR PARTNERS

Taplio: meet your LinkedIn growth shortcut

Taplio is an all-in-one LinkedIn growth tool to help you turn content into pipeline and grow your personal brand.

With Taplio you can:

✅ Instantly write posts made for your niche
💬 Generate AI-powered Smart Replies that cut time from 3 hours → 30 minutes
🤝 Build trust, visibility, and pipeline daily

"Taplio is how I've grown by over 30,000 followers on LinkedIn. I highly recommend it."

- Amanda Goetz, LinkedIn Creator (80k newsletter subs, $400,000 generated on LinkedIn).

Try Taplio for free for 7 days

Tuesday Tool

If you've ever wondered what comes after the transformer architecture that powers everything from GPT to Claude, this interview breaks down the exact type of model it could be. Corey talks to Stefano Ermon, CEO of Inception Labs, about Mercury, a new kind of AI model that generates text completely differently than ChatGPT or Claude.

Here's the deal:

Most AI models you use today are "autoregressive," meaning they write one word at a time, left to right, and can never take anything back.
Diffusion models flip that approach. They start with pure noise (like random gibberish), then gradually refine it into coherent text through multiple passes… kinda like sculpting a block of marble into a statue.
Why does this matter? Because diffusion models can process many tokens simultaneously instead of one-at-a-time, making them 5-10x faster than traditional language models like GPT. Plus, they're trained to fix their own mistakes during generation, potentially leading to more accurate outputs.

Inception Labs just released Mercury (try it here) as the first commercial-scale diffusion language model, and the results are impressive: comparable quality to speed-optimized models like Gemini Flash and Haiku, but at a fraction of the cost and latency.

Speaking of frontier and diffusion, Google is cooking up its own diffusion model too…

P.S before you dip: We just noticed Claude web (and desktop, we presume) added the option to pay per message when you get cut off by its rate limits.

This is clutch! Are we gonna pay? IDK man… but this is a much needed option on tight deadlines (like today…)