Skip to main content

Microsoft's CEO reveals that AI writes up to 30% of its code — some projects may have all of its code written by AI

 


Microsoft CEO Satya Nadella revealed that the company now uses Artificial Intelligence to write between 20% and 30% of the code powering its software. Satya joined Meta CEO Mark Zuckerberg this Tuesday at the LlamaCon conference to discuss developments in AI and their contributions to the open-source ecosystem. The fact that a major company like Microsoft relies so much on AI underlines how this technology is revolutionizing software development. Still, it also draws attention to the growing unease and uncertainty experienced by fresh software developers.

Almost every company uses Artificial Intelligence to a certain extent. At its Q3 financials last year, Sundar Pichai revealed that 25% of new code at Google is AI-generated. As it stands, AI is mostly there to power repetitive, data-heavy, and predictable tasks, which would yield a noticeable gain to corporate efficiency by cutting down on entry-level jobs. While it is true that AI-generated code has improved significantly over the recent years, it still requires senior developer oversight to ensure the production environment doesn't go haywire.

Mark and Satya spent part of their discussion detailing each company's AI adoption scale. AI generates one-third of the code in Microsoft's repositories and projects. The Meta CEO didn't provide an exact figure, but did outline plans for a future AI model to build and create future iterations of their Llama AI models. Although using AI to design AI might sound counterintuitive, there's an entire dedicated field that focuses on automating this design process, called AutoML.

“I’d say maybe 20%, 30% of the code that is inside of our repos today and some of our projects are probably all written by software,”

Satya Nadella, CEO of Microsoft

Microsoft is seeing better results with AI-generated Python code than C++, remarked Satya Nadella. There are several good reasons, including Python's simpler syntax, dynamic typing style, error handling, and memory management, since C/C++ don't natively feature a garbage collector. Likewise, C/C++ are mainly related to low-level code, which can be challenging to automate. You probably wouldn't want AI to be writing the next major Windows update.


Such developments naturally lead to job displacement concerns for new programmers, particularly in today's competitive job market. That said, it would be unwise for a software developer to ignore AI or become overly dependent on it. A balanced approach, with a grasp on fundamentals, knowing how to leverage AI as a tool, and strong critical thinking skills, appears to be the best way forward. Though the future is hard to predict, this percentage will probably change.



Comments

Popular posts from this blog

The RADIO framework provides a strong foundation for designing APIs and system integrations with consistency and maintainability in mind

  System Integrations & API Design: The RADIO Framework The RADIO framework provides a consistent, maintainable, and scalable approach to designing APIs and system integrations. It stands for Resource-oriented, Addressable, Documentable, Idempotent, and Observable. Resource-Oriented (R) Principle Focus on nouns (resources) over verbs (actions). 1. Aspect Implementation Detail Maintainability/Consistency Impact API Endpoints Use nouns in the URI (e.g., /users, /products/{id}). Employ standard HTTP methods (GET, POST, PUT, DELETE, PATCH) for CRUD operations. Predictability: Developers easily infer endpoint purpose. Clarity: Leverages standard REST principles, separating the what (resource) from the how (action). Data Models Define stable, versioned schemas (JSON/XML) for resource representations that reflect the resource's state. Decoupling: Protects consumers from internal system changes by maintaining a stable external API contract. 2. Addressable (A) Principle Every resource ...

Whooping cough is an illness that can spread easily. It's also called pertussis

  Whooping cough is an illness that can spread easily. It's also called pertussis. An infection with bacteria causes it. Many people with the illness get a serious hacking cough. Breathing in after coughing often causes a high-pitched noise that sounds like a "whoop." A case of Whooping Cough (pertussis) has been reported. Due to the nature of the illness we want to provide you with the necessary information about Whooping Cough and what steps you can take to protect your child and family. Extra advice may need to be sort if you have a newborn baby or are currently pregnant. Most students have been vaccinated against Pertussis when they were and infant. What is Whooping Cough (Pertussis) and how is it spread? Whooping Cough is a highly contagious respiratory infection caused by the Bordetella pertussis bacteria. It primarily affects the lungs and airways and can lead to severe coughing fits, especially in young children. It is spread through droplets when an infected pers...

Developing User Interfaces with GitHub Copilot

  Developing User Interfaces with GitHub Copilot, Part 3 by  John Miller  | April 30, 2025 This post is the third installment in the series on AI assisted UI development. While this post is largely stand-alone, consider reading parts  1  and  2  before reading this post. Have AI Add a Data Visualization We've looked at using AI to create and add UI components in prior posts. In this post I'll add a data visualization to a page. This figure shows a rendered Sales Funnel Summary page before making any changes: The goal in this post is to have AI add a data visualization of the sales funnel to the page. Prompt: Using css, add a Sales Funnel graphic from the data in the detail-table. Include the total value and the average age. Don't include ‘unknown’ or ‘closed lost’. Below is the response from the Claude 3.7 Sonnet Thinking model. Begin Response I'll add a CSS-based Sales Funnel visualization between your summary table and the details table. Here's how t...