Welcome to this week's edition of Overclocked!

This week, The biggest browser on the planet just folded a full assistant into the window you already use. Gemini now lives inside Chrome and it works across your tabs, apps, and tasks. Meanwhile, OpenAI shipped a new coding brain built on GPT-5 for Codex. Let’s dive in ⬇️

In today’s newsletter ↓
🧠 The daily window gets an AI copilot
🧑‍💻 A new coding brain arrives for teams
🌏 Huawei unveils data center super fabric
🛡️ France sharpens rules for AI threats
🎯 Weekly Challenge: See how well AI can mimic you

🧭 Google Fully Integrates Gemini Into Chrome

Google has embedded its Gemini assistant directly into Chrome on desktop in the United States, starting with English and expanding from there. The rollout introduces a visible Gemini button and a new AI Mode in the address bar that answers layered questions, summarizes pages, and works across your open tabs.

It also adds agent style abilities that will begin to handle multi step tasks. See Google’s Chrome AI features announcement and the Gemini in Chrome overview for specifics.

🔎 Smarter Research in Fewer Tabs

With AI Mode in the address bar you can ask complex questions, get on page answers, and follow up without leaving the site you are on.

Gemini can summarize a long page, compare information across several tabs, and resurface pages you saw earlier when you cannot remember where you read them. The goal is less tab juggling and faster decisions during research.

📆 Connected Apps and Agent Tasks

Because it sits inside the browser, Gemini now hooks into Google services. It can draft an email, add a Calendar event, open a YouTube video, pull directions in Maps, and soon perform step by step tasks like booking appointments or placing an order while you approve each step.

Google is also promoting an Agents to Payments standard so agents can check out with mainstream rails when merchants opt in.

🔒 Controls Rollout And Competitive Context

Chrome surfaces Gemini only when you click or use a shortcut, and you can manage or delete activity logs. The first wave targets Mac and Windows in the United States with Android already supported and iOS next.

The timing lands as regulators continue to pressure Google on competition, and as AI first browsers position themselves as alternatives. The integration makes the browser itself an assistant and raises the bar for everyone building AI centric browsing.

🧰 Guardrails And Safety Enhancements

Chrome is layering safety features alongside Gemini. The browser will warn on likely scams, help mute noisy notification prompts, and suggest changing compromised passwords. For business customers, Gemini in Chrome for Workspace adds controls so admins can govern what data the assistant may access and disable features for sensitive teams.

🧩 What This Means For Your Workflow

If you live in Google’s ecosystem this reduces context switching. Research can stay in one place with quick page gists, memory of what you opened last week, and an option to delegate simple chores to an agent. Even if you prefer other tools, the baseline for a mainstream browser just moved up.

🛠 Inside OpenAI GPT-5-Codex

OpenAI introduced GPT-5-Codex, a version tuned for agent style coding inside the Codex tools. It is available in the Codex CLI and IDE extension, in Codex on the web, and inside ChatGPT for paying tiers, with API access planned.

OpenAI says the model improves on real world tasks like code refactoring and long running jobs, and it adapts its reasoning time to the task at hand.

💡 What Changed for Developers

Early results point to better refactoring on large repositories and more useful code review comments from the assistant.

The model can take quick passes for simple edits or spend much longer on deep changes, which aims to reduce stop start loops when projects get complex. Because the tools run across terminal, IDE, web, and phone, teams can keep a single assistant attached to their code as they move between environments.

🧪 Benchmarks and Availability

OpenAI reports gains on SWE Bench Verified and on an internal refactoring evaluation. Coverage from the tech press and community reviews highlight the long task ability, with some tests running for several hours in one go.

For access, Codex is bundled with ChatGPT Plus, Pro, Business, Edu, and Enterprise, and the developer site says you can switch models in the extension or CLI once signed in.

🧰 Safety and Controls

The system card addendum outlines mitigations for prompt injection and harmful tasks, along with product level controls like sandboxed command execution, configurable network access, logs, and optional human approval gates.

For teams concerned about traceability, those logs and citations can be reviewed to see what the agent did and why. If your team uses an IDE extension today, the best next step is to try the CLI on a well scoped repo and compare cycle time on a weekly task before and after.

In short, GPT-5-Codex nudges coding agents toward practical daily use. It will not replace code review, but its steadier refactors and longer attention make it worth a trial on one real workflow this month.

The Weekly Scoop 🍦

🏗️ Huawei unveils SuperPoD interconnect as Nvidia faces new limits in China

🍽️ Top tech leaders attend UK state banquet for Trump’s visit

💳 Google introduces AP2 for agent payments across cards and real time transfers

🥇 OpenAI achieves perfect score at ICPC while rivals close the gap

🛡️ France outlines tougher rules in response to AI enabled threats

💽 Hard drives regain ground as AI storage demands surge

🎯 Weekly Challenge: The AI Doppelgänger Test

📃 How to do it:

1. Write a short paragraph in your natural style: maybe a text to a friend, a product review, or a social post.
   

2. Paste it into an AI tool and ask it to “rewrite this exactly as if I wrote it.”
   

3. Then, give the AI a fresh prompt (like describing your morning or recommending a restaurant) and see if it writes in the same voice.
   

4. Share both the real and AI-written samples with a friend (without telling them which is which) and ask them to guess.

💯 Score it
If your friend can’t tell the difference, your AI doppelgänger is ready for prime time. If they can spot the fake instantly, think about what quirks in your style the AI missed. Either way, you’ll learn something new about your own voice.

That’s it for this week! Do you think fully integrating Gemini into chrome will enhance your browsing experience, and will GPT-5-Codex tempt you to dive into the vibe coding world? Hit reply and let us know your thoughts.

Zoe from Overclocked