The Rundown AI / Articles / AI / 🎧 Gemini turns headphones into translators

🎧 Gemini turns headphones into translators

PLUS: Perform an instant attention audit on your webpage

Zach Mink

December 15, 2025

Good morning, AI enthusiasts. Google just quietly shipped the future of cross-language communication: any headphones, 70+ languages, instant translations that actually sound human.

With the new Gemini-powered Translate understanding context, slang, and cultural references in real time, the dream of seamless multilingual conversation may have just become a reality.

In today’s AI rundown:

Google adds real-time audio translation to any headphones
The Rundown Roundtable: Our AI use cases
Perform an instant attention audit on your webpage
Zoom claims new SOTA on AI reasoning benchmark
4 new AI tools, community workflows, and more

LATEST DEVELOPMENTS

GOOGLE

🎧 Google adds real-time audio translation to any headphones

Image source: Google

The Rundown: Google just rolled out a series of new Gemini-powered translation upgrades, including a new beta feature that streams live speech translations to any connected headphones — expanding on a capability previously restricted to its own Pixel Buds hardware.

The details:

The new Gemini 2.5 Flash Native Audio model upgrades conversational abilities, instruction following, and real-time info use for live voice agents.
The integration into Translate works with any earbuds on Android, supporting 70+ languages while preserving tone, pacing, and cadence of the speaker.
2.5 Flash Native Audio’s world knowledge also helps better interpret slang or culturally specific language contextually for more nuanced translations.
Google also expanded a Duolingo-style language practice mode to 20 new countries, with new features like streak tracking and pronunciation feedback.

Why it matters: The universal translators from sci-fi are getting closer, with Google’s update making cross-language conversation more seamless than ever. Between any headphones now facilitating real-time translation and the tech eventually coming to areas like YouTube, social media, etc., language barriers could disappear in the AI age.

TOGETHER WITH WARP

🤖 Spin up Warp Agents from anywhere

The Rundown: Small tasks shouldn’t slow your team down. Now you can summon a Warp agent right from Slack or Linear and let it handle the work for you.

Warp Agents allow you to:

Link Warp to Slack or Linear to enable cloud-run agents
Tag @warp in any Slack thread or Linear ticket to delegate tasks instantly
Let agents pull context and start executing in minutes
Get live updates and a clean PR ready to review

Start delegating work to Warp today.

THE RUNDOWN ROUNDTABLE

💡 The Rundown Roundtable: Our AI use cases

Image source: Reve / The Rundown

The Rundown: The Rundown Roundtable is a weekly feature in which we poll members of The Rundown staff about how we use AI in our work and daily lives.

Johannah, Finance: It’s year-end season & I'm doing a lot of modeling as we prep to pay taxes & pass off our financials to accountants. I use Excel in Google Drive the most, and recently I’ve been using Gemini to convert an accrual P&L estimate to a cash flow forecast. Gemini can pull the forecast into a separate spreadsheet, and I’ll ask it to list the assumptions it made to easily verify its accuracy. I would say it’s about 70% accurate; however, it still saves me a lot of time.

Joey, Head of Partnerships: To spice up my daily work check-ins, I share my completed work tasks into ChatGPT and prompt it to make my notes into pop-like song lyrics. From there, I pop it into Suno and create a fun “check-in” song for my team.

Zach, Lead AI Writer: I am not very handy by nature, but using ChatGPT and Gemini has given me the confidence to tackle home repairs I'd have never attempted before, from mounting fixtures to fixing our washing machine. Being able to send images for troubleshooting beats any generic guide and saves money without having to call a repair service.

AI TRAINING

🔥 Perform an instant attention audit on your webpage

The Rundown: In this guide, you will learn how to use Google Stitch to predict where people’s eyes go on your website, using these insights to help optimize your site’s design for conversions.

Step-by-step:

Sign in to Google Stitch.
Grab a screenshot of your site just above the fold (the hero section).
Go to Google’s AI Studio. Generate a heat map using your screenshot with the following prompt. Make sure you update the CTA text to match your site.
1. “Use Nano Banana to generate a predictive attention heat map on top of this landing page. The audit should highlight areas of high visual engagement and identify any elements that are distracting focus from the primary ‘Get Started’ CTA.”
Drop the heat map and your screenshot into a new Google Stitch session along with the following prompt. Make sure you toggle to start with a new Web design.
1. “Analyze the provided landing page screenshot alongside its predictive attention heat map. Generate a redesign of the hero section optimized specifically to maximize the click-through rate on the primary ‘Get Started’ button. The new design should use the heat map insights to eliminate distractions and focus user attention directly on the main conversion goal.”

Pro tip: Try comparing solutions across scenarios to understand the underlying reasoning patterns. This helps build better problem-solving skills for future challenges.

PRESENTED BY YOU.COM

🚀 Stop winging AI rollouts

The Rundown: Without a clear roadmap, AI adoption often fizzles, leaving ROI unproven and teams frustrated. The 90-Day AI Adoption Playbook from You.com gives you an effective, phased roadmap to move from pilot to real, measurable ROI — in only 90 days.

What you'll get:

A week-by-week rollout plan for secure, scalable AI adoption
Key actions to build user skills, certify competency, and drive consistent usage
Certification steps to ensure every user delivers impact

Make your first 90 days count. Download the playbook.

ZOOM

🧠 Zoom claims new SOTA on AI reasoning benchmark

Image source: Zoom

The Rundown: Zoom just announced that its ‘federated’ AI system scored 48.1% on Humanity’s Last Exam, surpassing Google Gemini on one of the industry’s most demanding expert-level reasoning tests.

The details:

Zoom’s federated approach orchestrates top models from OpenAI, Anthropic, and Google with its own small models through a “Z-scorer” selection system.
Zoom’s system edges out the previous leader Gemini 3 Pro’s 45.8%, which has since also been surpassed by the recently released GPT 5.2’s 50%.
The system will power the company’s AI Companion 3.0, with Zoom promising more accurate summaries, reasoning, and task automation across the platform.
The claim was also challenged by AI tool DeepWriter, which claims to have scored 50.91% on the benchmark in late November.

Why it matters: Zoom… The frontier AI research lab? Jokes aside, this is an impressive result, though one that warrants more proof — but a federated approach of combining multiple frontier models could be a compelling template for enterprises seeking new cutting-edge AI capabilities without building from scratch.

QUICK HITS

🛠️ Trending AI Tools

🪩 Disco - Google’s experimental browser that generates custom web applications on the fly
🤖 GPT-5.2 - OpenAI’s new most advanced frontier model family
👀 Cursor - New visual editor combining drag-and-drop with AI agents
⚙️ Tinker - TML’s API for fine-tuning language models, now generally available

📰 Everything else in AI today

Voxel51 introduced Sample-Level Evaluation, a new layer in the MLOps workflow that reveals hidden model failures (Like why Model A has the highest aggregate accuracy score but fails on critical scenarios) so teams can build better models.*

xAI is partnering with El Salvador to create a nationwide AI-powered education program for the country, bringing Grok across schools for both students and teachers.

Adobe integrated Photoshop, Acrobat, and Express directly into ChatGPT, enabling users to edit photos, create designs, and modify PDFs through conversational prompts.

OpenAI is ending its 6-month waiting period before new employees can access stock compensation, the latest move in an industry battle to attract and retain top talent.

Google is elevating engineer Amin Vahdat to a new chief technologist role to oversee AI infrastructure, putting him among a key group reporting to CEO Sundar Pichai.

*Sponsored Listing

COMMUNITY

🤝 Community AI workflows

Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.

Today’s workflow comes from reader Jeff N. in Atlanta, GA:

“I have two kids in middle school and was overwhelmed by emails from their teachers. Each kid has seven teachers, and each teacher emails us several times a week with various notices, assignments, simple reminders, etc. I set up a mailbox to collect these messages and created an n8n workflow that downloads them once a day. The workflow uses ChatGPT to read the emails, determine which items are actionable and urgent, and then compiles them into a daily summary that is sent to me and my wife so we can focus on what’s important.”

How do you use AI? Tell us here.