You can’t access Gemini 4 because it’s so powerful
Plus, 💪How to turn Claude into a personal fitness coach using your real health data, America’s biggest AI companies just agreed to police themselves, and more!
Hola Decoder😎
If someone forwarded this to you and you want to Decode the power of AI and be limitless, then subscribe now and Join Decode alongside 30k+ code-breakers untangling AI.
🧠 Google’s new Gemini 4 Argon is aiming straight at the frontier
Google has unveiled Gemini 4 Argon, a new frontier model built for long, complex workflows across coding, professional work and cybersecurity. For now, access is limited to vetted cyber defenders and trusted testers.
The Shift:
Argon Takes the Top Spot - Google says Argon beat GPT-6 Astra and Claude Opus 5.5 on 13 of 19 internal benchmarks, debuted at No. 1 on Arena’s text leaderboard and scored 53 on Artificial Analysis’ Intelligence Index.
Coding Gets a Big Jump - Argon scored 77.9% on DeepSWE, while Google says it is already helping with large code migrations, debugging and infrastructure optimization inside the company.
Built for Very Long Tasks - Google expanded Argon’s output limit from 64K to 1 million tokens, giving it room to work through long-horizon tasks involving code, documents, charts and video in a single run.
Cyber Comes First - Argon is initially rolling out through Google’s Fairwind Program for trusted cyber defenders. Pricing starts at $2/$10 per million tokens, with broader access planned later for developers, enterprises and consumers.
Argon is Google’s push to make Gemini competitive at the very top of the model market, especially for coding, enterprise work and cybersecurity. The catch is access: Google is deliberately starting small while it tests safeguards, even as it claims benchmark performance that puts Argon alongside the strongest models available.
Together with Billo
You Need To Check This Before You Scale Another Q4 Ad
Your best-performing ad may be underperforming against the ads competing for your customer.
Q4 benchmarks change by category and month. Scale against the wrong one, and every extra dollar can widen your performance gap.
Billo analyzed 80,069 sales-objective Meta video ads across 14 categories and turned the findings into a free Q4 Performance Calculator.
- See how your results compare with the right category benchmark.
- Find the month when your category offers the biggest opportunity.
- Identify the performance gap to fix before increasing spend.
In under a minute, you get a sharper target and a clear priority for your next creative batch. That means faster decisions and more confidence behind the ads you choose to scale during the most competitive quarter.
Run the calculation before your next Q4 budget increase locks in an avoidable disadvantage.
💪How to turn Claude into a personal fitness coach using your real health data
Generic workout apps give you the same plan regardless of how you slept or recovered. Claude reads your actual Apple Health data and builds every session around how your body is doing today. Here's the five-minute setup.
Step 1: Create a Fitness Coach project in Claude - Go to claude.ai, click Projects in the left sidebar, and create a new project. Name it Fitness Coach. Everything, your preferences, workout history, and the master prompt, lives here permanently.
Step 2: Connect Apple Health - Open the Claude app on your iPhone. Go to Settings > Permissions and toggle on Health. Approve steps, sleep, heart rate, and workouts. If you wear an Apple Watch, Garmin, or Whoop, that data flows into Apple Health automatically and Claude reads all of it.
Step 3: Connect Google Calendar - In claude.ai, go to Settings > Connectors and connect Google Calendar. Claude now drops your planned workouts directly into your schedule at your preferred time.
Step 4: Paste the master prompt into project instructions - Open your Fitness Coach project, click Set Custom Instructions, and paste the full coaching prompt from the guide. This is the brain of the system, it tells Claude to check your sleep, heart rate, and recovery before prescribing anything.
Step 5: Tell Claude your goal - Open a conversation inside the project and say something like:
"I want to build muscle. I have dumbbells at home and can train 4 days a week for 45 minutes."
Claude interviews you once, saves your profile permanently, and starts building your plan from your real data.
From that point, just ask "What should I do today?" every morning. Claude checks last night's sleep, your resting heart rate, and what you did yesterday, then tells you exactly what to do, or tells you to rest.
Requires Claude Pro or Max. Apple Health integration currently available in the US.
🤝 America’s biggest AI companies just agreed to police themselves
Six major AI companies, Google, Anthropic, Meta, OpenAI, xAI and Nvidia, have signed the White House’s new Joint Commitment on Frontier Responsibilities, a voluntary safety pact built around internal controls, outside audits and board oversight.
The Shift:
Models Get Watched Closely - Companies agreed to monitor frontier models during training and deployment for risks including cyberattacks, unintended system access, biosecurity and chemical threats.
Dedicated Safety Teams Get Power - Each company must empower an internal team to make sure those controls actually work, investigate failures and push for fixes when problems are found.
Outside Auditors Come In - Companies also agreed to bring in independent external evaluators to test whether their safety controls, monitoring and detection systems are working as intended.
Boards Get Direct Oversight - An independent board committee will receive reports from safety teams and auditors and oversee whether identified problems are actually addressed.
The deal creates a common safety framework, but it is still voluntary and carries no penalties or fixed implementation deadline. For now, Washington is betting that frontier AI companies can build stronger checks around themselves, with the agreement leaving open the possibility that some of these practices could later become law.
Together with Playbook Pro
Your Growth Strategy Is Costing You. Here’s the Math
Ad costs are climbing. Most first-time buyers disappear within months, and margins are shrinking fast when discounts drive sales.
We spent 200+ hours on research, distilling what top brands do into a field guide you can read and implement in under 30 minutes.
Inside Rewriting the DTC Rulebook, you’ll learn how to:
📈 Lift retention by 5% to unlock 25–95% profit growth
🔥 Launch emotional campaigns proven to double repeat sales in 8 weeks
💰 Use the Brand Value Canvas to raise prices 15–40% without losing volume
🧠 Slash decision cycles from 3 weeks to 3 hours using the Assumption Buster Toolkit
📊 Cut paid-ad dependence by 30% in 60 days with the Owned Media Checklist
🏆 Tools you Cannot Miss:
🎬 Pexo – Turns a product, website, or existing assets into polished launch videos, handling the story, scenes, voiceover, music, captions, motion graphics, and conversational edits.
🎥 RxFilmStudio – Creates and edits marketing and product videos from ideas, websites, screenshots, recordings, or footage, with natural-language control over scenes, animations, voiceovers, music, and subtitles.
📊 Datastory – Turns uploaded or public datasets into interactive charts and visual stories, with AI helping select visualizations and prepare them for publishing or embedding.
📝 Ferndesk – Creates and maintains customer-facing help centers by checking documentation against product changes and automatically drafting updates when information becomes outdated.
🔎 CrawlRaven MCP – Connects Search Console and GA4 to AI assistants so marketers can find declining pages, content opportunities, keyword gaps, and what content to refresh or create next.
🚀 Quick Hits
📚 SpaceXAI refreshed Grokipedia with a v0.3 redesign, adding a new logo, featured and most-read articles, live edit tracking, and updated article layouts as the AI-powered encyclopedia resumes making edits.
🧩 Google is replacing Gemini Gems with reusable Skills, letting users save custom instructions for repeated tasks. Skills begin rolling out to Workspace on October 5 and the Gemini app on October 13.
🔐 OpenAI says it disrupted a coordinated effort to extract protected reasoning from its models, tracing a core cluster of the activity to Moonshot, the Chinese AI company behind Kimi.
🔍 The FTC has reportedly opened an investigation into OpenAI, Anthropic and other AI companies, examining potential risks from their models after recent AI-related hacking incidents intensified existing safety concerns.
🎬 Instagram is adding an AI creative assistant to Edits that analyzes account performance and trends to suggest content ideas, hooks, captions, scripts and audio while explaining why certain posts performed better.
🥳 Prompt of the day
Create a Cohesive 9-Post Instagram Feed for One Product
Image Prompt: Create a 9-image Instagram feed for [Insert product] using [the same aesthetic / a newly defined cohesive aesthetic] across every image.
Context
- Product: [Insert product]
- Visual direction: [Same aesthetic as reference / describe desired aesthetic]
- Feed layout: 3x3 Instagram grid
- Goal: Make every post feel different while the full feed still looks visually connected
Instructions
Generate nine distinct product-focused images using different locations, camera angles, compositions, distances, and storytelling moments. Mix clean product shots with lifestyle scenes and naturally incorporate people, animals, nature, architecture, interiors, streets, travel settings, or unexpected environments where appropriate. Vary the framing across close-ups, wide shots, overhead views, candid moments, detail shots, environmental compositions, and interaction-based scenes. Keep the product recognizable and visually important in every image. Maintain a cohesive feed through consistent color palette, lighting character, styling, texture, mood, and overall art direction. If “same aesthetic” is selected, preserve the reference aesthetic closely without repeating the exact same composition.
🤳AI Nugget of the Day
Thanks for Decoding with us🥳
Your feedback is the key to our code! Help us elevate your Decode experience by hitting reply and sharing your input on our content and style.
Keep deciphering the AI enigma, and we'll be back with more coded mysteries unraveled just for you!