Welcome back to Daily Zaps, your regularly-scheduled dose of AI news ⚡️ 

Here’s what we got for ya today:

• 👩🏻‍💻 OpenAI Releases “Codex” AI agent for coding

• ❌ Microsoft lays off engineers as AI writes 30% of code

• 🏅 New Google Cloud AI certification

• 😎 Vibe-coding startup Windsurf launches AI models

Let’s get right into it!

Sign Up | Advertise | Tools

STARTUPS

OpenAI Releases “Codex” AI agent for coding

OpenAI has launched Codex, an agentic coding tool in research preview designed for experienced developers to delegate routine programming tasks to an AI that generates production-ready code while transparently showing its process. Accessible via the ChatGPT sidebar, users prompt Codex to either write code or provide advice within a container reflecting their development environment.

Developers can enhance Codex’s performance by adding an “AGENTS.md” file with custom instructions about the codebase and project standards. Built on codex-1, a fine-tuned version of OpenAI’s reasoning model trained with reinforcement learning on diverse coding tasks, Codex aims to address previous concerns about AI-generated code quality, security, and debuggability by showing its reasoning during code generation.

Although users must still review all AI-produced code, Codex is now available to ChatGPT Pro, Enterprise, and Team users at no extra cost during this initial rollout, with plans for rate limits and pricing in the future, and expanded access for Plus and Edu users coming later.

BIG TECH

Microsoft lays off engineers as AI writes 30% of code

Microsoft’s recent layoffs in Washington hit coders hardest, with over 40% of the 2,000 affected employees coming from software engineering, according to Bloomberg’s analysis of state filings. While coders make up a large part of Microsoft’s workforce, the company hasn’t shared exact figures. These cuts are part of a broader layoff affecting about 6,000 employees.

Notably, this follows CEO Satya Nadella’s statement that AI now writes up to 30% of the company’s code. When asked if AI-assisted coding influenced the layoffs, Microsoft declined to comment, emphasizing instead that the reductions aim to streamline management layers.

BIG TECH

New Google Cloud AI certification

As AI use grows across industries, employees are increasingly expected to have AI knowledge. To support business leaders and non-technical professionals, Google Cloud has launched a "first-of-its-kind" Generative AI Leader certification, focusing on strategic AI adoption beyond coding. The 90-minute exam costs $99 and is available starting May 14, positioned as a career growth investment.

Google also offers a free, 7–8 hour learning path with five courses covering foundational AI concepts, applications, and organizational impact, accessible without taking the exam. Registration for the exam and courses is available online with no prerequisites, and the certification lasts three years. For those seeking alternatives, many free AI courses and certificates are available from other providers, including LinkedIn’s popular AI courses temporarily offered for free.

STARTUPS

Vibe-coding startup Windsurf launches AI models

Windsurf, a startup known for AI tools that help software engineers “vibe code” via chatbots, announced its first in-house AI models called SWE-1, SWE-1-lite, and SWE-1-mini, designed to optimize the entire software engineering process—not just coding. Despite OpenAI’s reported $3 billion acquisition deal with Windsurf, this launch signals the startup’s move from building AI-powered apps to creating its own foundational models.

SWE-1 competes well with models like Claude 3.5 Sonnet and GPT-4.1 but lags behind top-tier models like Claude 3.7 Sonnet. SWE-1-lite and SWE-1-mini will be free for all users, while SWE-1 is for paid users only, with pricing yet to be announced but claimed to be cheaper than Claude 3.5 Sonnet. Windsurf emphasizes that coding alone isn’t software engineering, highlighting their models’ ability to handle multiple development environments and complex tasks through new training methods that reflect real-world engineering workflows.