Gemini Is About to Get Wings on Your Phone With Agentic Skills

Google I/O is almost here, and all attention is turning to the company’s artificial intelligence ambitions—especially around Gemini. Although nothing has been officially confirmed, a fresh leak offers an early look at what Google might be preparing behind closed doors. The so-called Gemini Spark Model appears to be a new agentic AI layer designed to perform real-world tasks on your phone, from cleaning up your inbox to generating meeting briefs.

What Is the Gemini Spark Model?

According to a post shared by Waguri_Kaoruko8 on X, the Gemini Spark Model is currently in beta. It introduces an Agent or Chat Mode that enables more advanced, tool-based actions. The idea is simple: instead of just answering questions, Gemini would handle annoying digital chores automatically.

For example, one feature can summarize newsletters, archive clutter, and even unsubscribe you from mailing lists. Another tool generates meeting briefs by pulling together relevant information before an important call. There’s also a custom news digest that follows the stories you care about, rather than flooding you with random headlines.

This shift suggests Google is pushing Gemini toward becoming a background productivity layer—an AI that works for you, not just one you occasionally ask questions to. And honestly, that feels like the smarter direction.

Agentic AI: More Than Just a Chatbot

The leak was later quote-posted by the AI News testing catalog account, which noted that there currently isn’t support for importing SKILL MD files directly. Users may have to rely on the good old copy-paste method for now. The post also mentions no signs of browser control or full computer-use capabilities yet—two features many were hoping for as AI agents grow more capable.

However, the Gemini Spark Model still represents a major leap. It focuses on automation and personalization, turning Gemini into a true AI assistant that acts on your behalf. Building on this, the system may let users create custom “skills” for Gemini—giving it a title, explaining its purpose, and adding instructions for behavior. Think of it as building mini AI workflows without any coding.

How Gemini’s Agentic Skills Could Change Your Daily Routine

Imagine waking up to a phone that has already sorted your inbox, unsubscribed you from spam, and prepared a brief for your 10 a.m. meeting. That’s the promise of the Gemini Spark Model. Instead of manually sifting through emails or scanning news feeds, you let the AI handle the grunt work.

As a result, your phone becomes a proactive assistant. It learns your preferences, follows your interests, and acts without constant prompts. This is a significant departure from traditional chatbots, which require you to initiate every interaction.

Custom Skills: Build Your Own AI Workflows

The ability to create custom skills is particularly exciting. Users could design workflows for specific tasks—like summarizing daily reports, tracking project updates, or even managing social media posts. This DIY approach puts power in the hands of the user, making Gemini adaptable to individual needs.

Of course, it’s important to keep expectations in check. None of this is official yet, and leaks around Google I/O season tend to fly around fast. Still, the timing makes this especially interesting. With Google expected to go all-in on Gemini at I/O next week, there’s a very real chance we could see at least some of these features become official sooner rather than later.

What This Means for Google I/O 2025

Google I/O has always been a showcase for the company’s biggest innovations. This year, all signs point to Gemini taking center stage. The Gemini Spark Model could be just the beginning of a broader push toward agentic AI on Android.

For more insights, check out our guide on Google I/O 2025 preview: what to expect. And if you’re curious about how AI assistants are evolving, read our analysis on AI assistants vs. chatbots: the key differences.

Ultimately, the Gemini Spark Model represents a shift from passive AI to active assistance. If Google delivers on these promises, your phone might soon become your most productive tool yet.

ChatGPT Now Offers Finance Tips If You Connect Your Bank Account — But Should You Trust It?

OpenAI has rolled out a new personal finance feature for ChatGPT, allowing Pro subscribers in the U.S. to link their bank accounts and receive tailored financial advice. For $200 a month, users can now ask the AI chatbot about spending habits, upcoming bills, or even home-buying plans. However, the move raises serious questions about privacy and security. Should you really hand over your financial data to an AI? Let’s break down what this means for your money and your peace of mind.

How ChatGPT’s Finance Feature Works

OpenAI’s latest update, currently in preview for Pro users, integrates with Plaid, a platform that connects bank accounts to third-party apps. Plaid works with over 12,000 financial institutions, including Chase, Fidelity, Schwab, and American Express. Once you link your accounts, ChatGPT can access your balances, spending history, active subscriptions, upcoming payments, stock portfolio, and liabilities like credit card debt or mortgages.

It cannot view full account numbers or make changes to your accounts. OpenAI says the feature uses its newer GPT-5.5 model, which handles financial reasoning better than previous versions. A dashboard provides a snapshot of your finances, and you can ask questions like, “Has my spending changed recently?” or “How can I plan for buying a home?” Support for Intuit is also coming soon, which would let ChatGPT analyze tax impacts of stock sales or loan approval odds.

Privacy Risks of Sharing Financial Data with ChatGPT

OpenAI claims users stay in control. You can disconnect accounts at any time, and synced data is removed within 30 days. You can also delete saved financial memories and choose whether your data trains OpenAI’s models. However, the company does not clearly explain what happens to your financial data beyond AI training or what protections exist in case of a security breach.

Your bank balance, spending patterns, and credit card debt are highly sensitive. Trusting an AI chatbot with this information is a personal decision, but it’s worth noting that similar concerns arose with ChatGPT Health, introduced in January. That feature also raised trust issues around handling sensitive personal data. As AI privacy experts often warn, once data is shared, it can be difficult to fully control its use.

What Happens to Your Data After Disconnecting?

OpenAI says data is removed within 30 days of disconnecting, but this doesn’t account for potential backups or third-party sharing. The company has not disclosed whether Plaid or other partners retain copies of your information. This ambiguity is concerning, especially given the rise in data breaches across financial platforms.

Is AI Financial Advice Reliable?

While GPT-5.5 offers improved financial reasoning, AI models are not infallible. They can misinterpret data or provide advice that doesn’t account for your unique circumstances. For example, asking about home-buying plans may yield generic suggestions rather than personalized strategies. The feature is still in preview, meaning bugs and inaccuracies are possible.

OpenAI has not published independent audits of its financial advice accuracy. This lack of transparency makes it risky to rely on ChatGPT for critical financial decisions. As financial planning tools evolve, it’s wise to cross-check AI advice with a human advisor or trusted source.

Should You Connect Your Bank Account to ChatGPT?

The decision ultimately depends on your risk tolerance. If you value convenience and are comfortable with OpenAI’s privacy policies, the feature might help you track spending or plan goals. However, given the lack of clear data protection guarantees and the sensitive nature of financial information, many users may choose to wait.

Consider alternatives like dedicated budgeting apps that offer similar insights without sharing data with an AI chatbot. These apps often have stronger privacy protections and are designed specifically for financial management. Until OpenAI provides more concrete security measures, connecting your bank account to ChatGPT carries significant risks.

In summary, ChatGPT’s finance tips are a bold step toward AI-driven personal finance, but they come with serious privacy and reliability concerns. Proceed with caution, and always prioritize your data security over convenience.

OpenAI brings Codex coding agent to ChatGPT mobile app: What you can do from your phone

Imagine being able to keep an eye on your code while you are away from your desk. That is now a reality. Codex, the powerful AI coding agent developed by OpenAI, has officially arrived in the ChatGPT mobile app. Available on both iOS and Android, this update lets developers monitor and control coding tasks directly from their phones.

The feature is rolling out in preview across all subscription tiers, including the Free and Go plans, in every supported region. Mobile support currently works with the macOS version of Codex. Windows users will not have to wait long, as support is expected soon.

What can Codex do from your phone?

Do not expect to write a full application on a tiny screen. Instead, think of this as a remote control for an AI-powered coding session that is already running on your computer. When you link your phone to any machine where Codex is active — whether that is a laptop, a Mac mini, or a remote development environment — the app pulls in the live state from that setup.

From there, you can review active threads, approve commands, switch models, check terminal output, inspect test results, and look at diffs. All of this happens without you being anywhere near your desk. Your files, credentials, and local configuration stay on the machine doing the work. Real-time updates flow back to your phone as Codex makes progress.

OpenAI uses a secure relay layer to keep your trusted machines reachable across devices without exposing them to the public internet. This means your code remains safe while you step away.

OpenAI sees Codex as a major priority

The company has been reshaping its products around Codex. There are confirmed plans to combine ChatGPT, Codex, and the Atlas browser into one larger AI superapp. Recently, OpenAI brought Codex into Chrome, putting its coding agent directly inside the browser. It even introduced quirky “Codex Pets” that show live progress updates while AI coding tasks run in the background.

Building on this momentum, the mobile expansion shows that OpenAI wants Codex to be accessible everywhere. For developers who spend long hours writing and debugging code, being able to check in from a phone is a significant convenience.

Competition is heating up

OpenAI is not alone in this space. Rivals like Anthropic’s Claude Code gained a remote monitoring feature back in February. As a result, OpenAI is making its own push to stay ahead. The mobile release is a clear signal that the company does not want to fall behind in the race for AI-powered development tools.

For developers looking to try it out, the feature is already live in the ChatGPT app. Simply update the app and connect to a machine running Codex. For more tips on using AI in your workflow, check out our guide on AI coding assistants. If you are interested in other mobile productivity tools, read our best productivity apps roundup.

In conclusion, the arrival of Codex in the ChatGPT mobile app marks a practical step forward for remote development. It allows you to stay productive even when you are not at your desk. This means that grabbing coffee or commuting no longer means losing touch with your code.