Did you know that we have LinkedIn, Instagram and X accounts
that you can follow?
Hi {{first_name|everyone}},
Microsoft is building more of its own AI, and the price gaps it’s claiming against OpenAI and Anthropic are hard to look past. We’ve also got a look at running a capable model privately on your own laptop, a plain-English term worth knowing, a few new tools worth bookmarking, and a tiny submarine that doubles as a volume slider. Let’s jump into it.
🤿 DEEP DIVE
Microsoft Is Making Its Own AI
And the price comparisons are pretty hard to ignore.
Microsoft is in an interesting position. It works with OpenAI and other AI companies, but it's also trying to build more of the AI technology itself.
And judging by what it announced this week, it's getting pretty serious about it.
At its Ignite conference on November 17, Microsoft introduced a new lineup of AI models for coding, cybersecurity, image generation, and voice. But what's more interesting than the number of models is how much cheaper Microsoft says some of them are.
It was how much cheaper Microsoft says some of them are.
Those price comparisons are something
Take MAI Code 2, Microsoft's new model for everyday programming tasks.
According to Microsoft's own benchmarks, it delivers 98% of the performance of OpenAI's GPT-6.1 Sol while costing just 40% as much.
The comparison with Anthropic is even more striking. Microsoft claims MAI Code 2 reaches 96% of Claude Opus 5.5's performance at only 5% of the cost.
Yes, 5%.
Now, those are Microsoft's numbers, based on selected coding tests. They don't necessarily mean MAI Code 2 will perform almost identically to those models in every real-world situation.
Still, that's quite a price difference to be putting on the table.
And coding isn't the only area Microsoft is targeting.
It also introduced MAI Cyber 2 Flash, a model designed specifically for cybersecurity work. The idea is to help security teams find and address weaknesses in software.
Microsoft says this model can outperform a competing OpenAI model on a cybersecurity benchmark while costing significantly less per task.
So why build so many different models?
This is where Microsoft's approach gets interesting.
Rather than trying to make one AI model that's excellent at everything, Microsoft is building models for specific jobs.
One for coding. Another for cybersecurity. Others for images and voice.
The thinking is pretty straightforward. If a model is built around a particular task, Microsoft can focus on getting that job done well without necessarily paying the cost of running a more general model.
And that fits with the company's bigger goal of improving performance while keeping costs down.
There's another reason Microsoft wants this
For years, Microsoft has relied on OpenAI and other AI companies to help power its products.
But the more AI technology it develops itself, the more control it has over what goes into those products and how much that technology costs.
Microsoft isn't just trying to compete with other AI models on performance anymore. It's trying to compete on price, too.
For businesses using AI for coding, security, and everyday work, that could eventually mean more affordable tools.
Of course, Microsoft's benchmark claims are one thing. How these models perform across different tasks is another.
But the strategy is becoming pretty clear. Microsoft wants to build more of its own AI, make it cheaper to run, and rely less on the companies it currently works with.
The quick take
Microsoft is building more AI models in-house, targeting coding, cybersecurity, image generation, and voice.
Microsoft says it approaches the performance of leading rival coding models at a fraction of their price.
MAI Cyber 2 Flash focuses on security, with Microsoft claiming better benchmark performance than a competing OpenAI model at a lower cost.
Microsoft is building models for particular tasks rather than relying entirely on general-purpose AI.
Building its own models could reduce Microsoft's dependence on AI partners and potentially make AI tools more affordable.
🤝 FROM OUR FRIENDS
A quick share we thought you might find useful
Subscribe to The Flyover for FREE!
The news shouldn't eat your morning.
3.7 million readers start their day with The Flyover: the stories that actually matter in politics, business, sports, and tech, boiled down to the facts and delivered in a read you can finish before your coffee cools.
Our editors sift through hundreds of sources so you don't have to.
By clicking below, you are subscribing to The Flyover's FREE daily newsletter and will begin receiving future editions in your inbox.
🔥 WHAT’S THE BUZZ
Where the community stands on AI, tech, and productivity
Independent AI creator Alex Finn demonstrated Qwen 3.8 running locally on a laptop equipped with NVIDIA’s RTX Spark hardware. His pitch was straightforward: a capable model running privately on your own machine, without usage limits or sending every request to a cloud service.
Local models offer a different trade-off from ChatGPT, Claude or Gemini, more privacy and fewer usage restrictions, but more responsibility for hardware and setup. As capable models get smaller, running your own AI may become a realistic option for more workers and small teams.
⏱️ 1-MINUTE BREAKDOWN
Tech and AI terms explained in plain English
What is Grounding?
Giving AI relevant source material to base its answer on. At work, that might mean using your company’s actual policies or documents to answer a question. Grounding helps make answers more relevant and verifiable, though mistakes are still possible.

Example:
You want to know whether your company pays for home office equipment. You give the AI your company’s expense policy and ask it to explain the relevant section. It uses that document to answer, rather than giving general advice about what companies usually cover.
⚒ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
🎥 Creatify Max
Helps you make video ads and run Meta and TikTok campaigns.
Why it’s useful
It creates video ads, sets up the campaigns, and monitors results around the clock so you know what to fix next. It also keeps you in control since nothing goes live without your approval.🤖 Rakazo
Set up AI teammates that keep working after chat ends.
Why it’s useful
It gives each bot its own memory, routines, and computer so tasks can continue while you step away. You can watch the browser and take over when needed, and self host it with your own model and keys.🔎 EkLine
Find and fix what ChatGPT can’t implement from your docs.
Why it’s useful
It audits your site, docs, GitHub, and community posts to spot where AI gets confused or stuck building. It then prioritizes gaps, proposes code grounded fixes as pull requests, and retests AI answers to see what improved.
👀 ON OUR RADAR
One more thing we thought you'd like
Sales Owns Retention, Not Just Closed Deals
Pebl’s Chief Business Officer, Dutta Satadip, joins The GTM Garage to show how Pebl merged Operations and GTM under one leader, and how it aligned goals and incentives across the customer lifecycle.
You’ll also see how Pebl uses AI across that lifecycle and the stack behind prospecting, pipeline, and handoff to onboarding.
15 min conversation. 15 min open Q&A. For Seed to Series A teams.
🤖 AI MADE THIS
Interesting and inspiring creations made with AI
A Volume Slider That Takes You to the Bottom of the Ocean
Elizaveta Zabrodskaya used Claude Fable 5.2 to build a volume slider that looks like a tiny submarine descending through a pixel-art ocean. As it travels deeper, the scenery, depth readings, and sounds change. It's a clever little experiment in making ordinary software controls more fun.
AI used
Claude Fable 5.2
Mythos 5.2
ℹ️ ABOUT US
The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.
We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.
Our mission is to empower individuals, boost their productivity, and future-proof their careers.
We read all your comments - please provide your feedback!
Did you like today's email?
What more do you want to see in this newsletter?










