Welcome, humans.

You could look at the state of the software job market as an indicator of what’s to come for the rest of us as AI use spreads.

So far, that means: 1. leaner teams, and 2. management being more selective with new hires, leading to 3. a slightly higher unemployment rate of 5.7% (versus the overall 4% rate from January).

And if OpenAI’s planned $10K/month software agents are any indication, that trend could likely accelerate very soon.

As for the actual productivity? AI is boosting output by just 10-30% for most developers rather than the hyped 5-10x gains, with benefits concentrated in specific tasks like writing test cases and small scripts.

Meanwhile, overall development remains constrained by meetings, requirements gathering, and business decisions that AI can't solve. Bureaucracy strikes again…

Real quick: Want to reach 500,000+ AI-obsessed professionals? All you gotta do is click the button below and fill out a quick survey (takes less than 1 min!).

We’re booking ad spots for Q2 right now (~4-6 weeks out), and they’re filling up fast—so don’t miss your chance to get your product or service in front of 500K+ people who care about AI as much as you do!

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

• Microsoft developed and tested new AI to ween itself off OpenAI.

• Apple delayed personalized Siri for another year.

• A new AI agent tool called Manus broke the internet.

• US proposed more AI bills in early 2025 than all of 2024.

Microsoft is about to release its own AI models to rival OpenAI, Anthropic, and Elon…

It’s been a minute since we dived into any Microsoft vs OpenAI drama, largely because OpenAI CEO Sam Altman’s been busy with a new baby (congrats!) and a potential eye-popping fundraise from SoftBank.

How’s this for drama: The Information just published a fascinating exclusive about how Microsoft’s “AI CEO”, Mustafa Suleyman, has made significant progress with internal AI models called MAI that supposedly match OpenAI’s models on benchmarks.

Here’s the TL;DR:

• Microsoft is swapping out OpenAI models for MAI models inside Copilot.

• Microsoft will offer these MAI models directly to developers later this year.

• Suleyman said “we are now using, under the hood, essentially all of the models from major labs.”

That last point means Microsoft has tested xAI’s Grok, Anthropic’s Claude, Meta’s Llama, and DeepSeek’s open source models in Copilot, too.

The article also spilled some tea on where Suleyman and Altman’s relationship soured—allegedly, when OpenAI wouldn’t reveal how o1’s “thinking” works on a tense phone call. So really, OpenAI started it yeah?

To us, this is all just more evidence of AI “models” becoming like software stacks—a layer or two deeper than most people care to know about. If that’s not true now, GPT-5 will make sure of that.

Right now, OpenAI has so many models, it just had to make a new page to make it easier to compare them all.

But to regular people chatting with ChatGPT or Copilot (and their bosses), it doesn’t really matter what model is running under the hood—as long as it gets the job done.

Which AI is best for the job, then? As far as the LM Arena is concerned, the answer is Grok 3, now free for anyone to try with Think mode (Grok's version of o1’s reasoning) and DeepSearch (Grok’s version of Deep Research). You get ~15 normal messages, ~10 thinking, and ~10 DeepSearch questions every 2 hours.

However, many people won't try Grok (even IF it's free) because they don't want to support Elon. They’ll stick with OpenAI, Claude, Copilot or Gemini instead.

This is why what really matters isn't what's under the hood, but what's on the surface—the brand, user experience, and most important of all, the distribution.

People who do use Grok like it because it's “elegantly integrated” inside of X.com. Google is about to embed a Perplexity-style “AI Mode” directly into search results (demo), and Meta can instantly distribute any new AI tools to a billion users on launch.

While ChatGPT recently doubled from 200M to 400M weekly users in just six months (175M on mobile), Copilot has its own powerful distribution channel—it's built right into the workflow of 2M companies who require their employees to use Microsoft 365.

But what if your Copilot (which you have to use for work) was actually running on Gemini, or… *~Le Gasp~* …Grok?! Would you even notice? Would you even care?  

FROM OUR PARTNERS

Live Demo: Automate compliance to fuel your startup's growth

Want to learn why compliance is critical to growing your business, and how automation can simplify the time-consuming process? 

On March 11, join this 45-minute live product demo of Vanta and get insights into: 

• Why compliance is important for building trust and moving upmarket.

• How Vanta centralizes program requirements and automates evidence collection for frameworks like SOC 2, ISO 27001, HIPAA, and more.

• Best practices for managing risk and proving security in real-time.

Plus, you can ask all your questions to Vanta team members during the session. 

Discover how automation can transform your compliance efforts into a streamlined, efficient process—secure your spot today.

Prompt Tip of the Day

QwQ-32B is Alibaba’s newest AI model, and it’s available to run both on your computer via Ollama or LM Studio, as well as on the Qwen Chat Cloud—just make sure you select “QwQ-32B” from the picker as the model to test it out.

Julian’s video explains it well. QwQ-32B is a big deal—it’s better than DeepSeek’s R1 (y’know, the AI that broke NVIDIA’s hot streak?), and even outperforms OpenAI’s o1-mini on many tasks, AND you can run it on your computer for free(!!!).

The catch? You need a computer that can run it. To see if yours will, use this.

P.S: If you wanna get real nerdy, follow this video for how to run Bolt.DIY (open source version of Bolt.new) locally on your computer to “vibe-code” offline with QwQ.

Treats To Try.

1. *Join Intercom for their biggest AI announcements yet, including Fin Voice—AI phone support for real conversations—and exclusive first look at the latest upgrades to the #1 AI agent in the market. Register here.

2. Turing helps you implement real-world AI solutions by connecting you with elite talent and tech (raised $111M).

3. Instrumental connects to your manufacturing data and uses AI to detect issues before they become costly problems.

4. DeepDub creates authentic, emotion-rich voice dubbing for your content in 130+ languages. 

5. Windsurf Wave 4 lets you build apps faster with live previews, automated imports, lint checking, suggested next steps, and simple file sharing.

6. Gum.new helps you make customizable landing pages for your services sold via Gumroad.

7. Pig automates Windows apps with AI by controlling your mouse and keyboard through chat or code.

See our top 51 AI Tools for Business here!

*This is sponsored content. Advertise in The Neuron here.

Around the Horn.

• Manus is a new agent that took the internet by storm this weekend—it aims to complete tasks from planning trips to analyzing Tesla stocks while you sleep (video). Some called it a second DeepSeek moment, while others said it’s probably not, and someone’s already reverse-engineered it (details).

• Apple delayed the “more personalized Siri” part of its Apple Intelligence push for at least another year.

• US Lawmakers proposed more laws on AI (781 bills) in the first 66 days of 2025 than all of 2024 (743 bills).

FROM OUR PARTNERS

Sales teams using Attention are crushing their targets

Attention's platform is transforming how companies sell. HorizonIQ boosted post-call summaries from 50% to 99%, saving 25 minutes per call and doubling productivity. Unify increased win rates by 40% and reduced sales ramp time by 3x.

We’re big Attention fans at The Neuron and can confirm: if you sell literally anything, you need this tool. Book a free demo.

Monday Meme

A Cat's Commentary.