AI Daily 2025-10-31: Latest Updates from OpenAI, Google, Perplexity & Devin

The AI field had an explosive day on October 30, 2025! From OpenAI’s new security tool Aardvark, paid updates for Sora and Codex, to Google AI Studio’s developer tools, Perplexity’s AI patent search, and Devin’s public beta for computer control, this article summarizes all the key points for you.

The tech world is in constant flux, especially in the field of artificial intelligence, where disruptive progress can be made every day. Just yesterday, several top AI companies coincidentally released major updates, heralding that AI technology is penetrating into every aspect of software development, content creation, and even professional research at an unprecedented speed.

This article will give you a quick overview of the latest developments from OpenAI, Google, Perplexity, and Cognition, and understand how these new tools and features will change our work and lives.

OpenAI’s Triple Play: Major Updates for Aardvark, Sora, and Codex

OpenAI once again took center stage, launching three major updates at once, covering software security and the commercialization of creative tools.

1. Aardvark: An AI Security Researcher Powered by GPT-5

First up is a new product called Aardvark, an agentic security researcher powered by GPT-5. Software security has always been one of the most critical and challenging aspects of the tech industry, with thousands of new vulnerabilities discovered every year.

Aardvark aims to change the asymmetry of offense and defense, empowering developers and security teams with greater power. It can think like a human security expert, automatically discovering, evaluating, and fixing security vulnerabilities by continuously analyzing code repositories.

How does Aardvark work? Instead of relying on traditional methods like fuzzing, it utilizes the reasoning capabilities of large language models to understand code behavior. When a potential vulnerability is found, it verifies its exploitability in a sandbox environment and integrates OpenAI Codex to generate patches for developers to review and deploy with a single click.

In benchmark tests, Aardvark successfully identified 92% of known vulnerabilities, demonstrating its high performance in the real world. Aardvark is currently in private beta and plans to offer free scanning services for specific open-source projects.

Learn more: OpenAI’s official introduction to Aardvark

2. Sora Paid Plans Go Live, Opening a New Chapter in the Creator Economy

Since its launch, OpenAI’s video generation model Sora has attracted heavy users with its powerful capabilities. However, the free generation limit of 30 times per day is clearly no longer sufficient to meet the needs of professional creators, while also putting enormous computational cost pressure on OpenAI.

To address this, OpenAI has officially launched paid add-on quotas for Sora. Users can now purchase additional generation credits according to their needs, for example, by purchasing 10 generation credits for $4 on the App Store.

This is not just to solve the cost problem, but also a step towards a new Sora economy. OpenAI envisions that in the future, copyright owners (such as film companies or celebrities) can license their owned characters or images, allowing users to pay to have them cameo in Sora-generated videos, thereby creating new revenue streams.

3. Codex Introduces a Credit System for More Flexible Payments

Similar to Sora, OpenAI’s code generation tool Codex has also received more flexible payment options. In the past, if users exceeded the usage of the Plus plan, their only option was to upgrade to the expensive Pro plan.

Now, OpenAI has introduced a Credits system. When users reach their plan’s usage limit, they can purchase credits to continue using the service without having to upgrade the entire plan. These credits are universal and can be used not only for Codex but also for Sora, allowing users to “pay as they go” according to their actual needs.

Official explanation: Learn more about how the credit system works

Google AI Studio Launches Logs and Datasets to Simplify AI Development Workflow

Google also brings good news for AI developers. Google AI Studio has now launched new Logs and Datasets features, designed to solve one of the most troublesome problems in developing AI applications: how to ensure the stability and high quality of model output.

Key highlights:

• Track without modifying code: Developers simply click “Enable logging” in the dashboard, and all API calls, whether successful or not, will be automatically tracked.
• Simplified debugging process: Developers can easily view information such as the input, output, and response code of a request to quickly locate and resolve issues.
• Turn insights into product optimization: You can export these logs as datasets in CSV or JSONL format for model fine-tuning, performance tracking, or creating a reliable baseline of expected results.

This feature is available for free in all regions where the Gemini API is available, greatly lowering the barrier for developers to monitor and improve their AI applications.

Learn more: Google for Developers Blog

Perplexity Releases Patent Research Agent, Making IP Intelligence Accessible

Perplexity is known for its powerful conversational search engine, and now it is extending this technology to a highly specialized field: patent research. Perplexity has launched Perplexity Patents, claiming to be the world’s first AI patent research agent.

In the past, patent searches required precise keyword combinations and an understanding of obscure syntax, making it extremely difficult for non-professionals. Perplexity Patents completely changes this.

Core advantages:

• Natural language search: You can ask questions in plain language, such as “Are there any patents on AI language learning?”
• Beyond keyword matching: It can understand semantics and find important inventions that are conceptually related but do not use the exact words you queried.
• Integrated information sources: In addition to patent literature, it also explores academic papers, software repositories, etc., to provide a more comprehensive view of technology trends.

Perplexity Patents is currently in beta and is free to use for all users worldwide.

Try it now: Perplexity Official Blog

Devin Opens Public Beta for Computer Use, Taking AI Assistants a Step Further

Cognition, the developer of Devin, hailed as the “first AI software engineer,” has announced that Devin’s Computer Use feature has now entered public beta.

This means that Devin’s capabilities are no longer limited to code editors and browsers. It can now operate your desktop applications, create and test mobile apps, and even automate various tedious daily tasks just like a human.

In the demo, Devin shared a screen recording showing the complete process of controlling a computer. This advancement takes the practicality of AI agents to a new level and gives us more imagination for future human-computer collaboration.

Watch the demo: Cognition’s official X platform post