Grok 4.6 closes the Frontier gap

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.

🚀 Grok 4.6 pushes into the AI frontier

xAI has released Grok 4.6 with major gains in agentic coding, knowledge work, and long-running tasks. The model is now competing directly with some of the strongest frontier systems while maintaining aggressive pricing.

The Decode:

1. Grok closes the intelligence gap - Grok 4.6 scores 61 on Artificial Analysis’ Intelligence Index, up from Grok 4.5’s 56 and matching GPT-5.6 Sol in xAI’s published comparison.

2. Coding and professional work improved sharply - Grok beats Sol on CursorBench and FrontierCode while scoring strongly across professional and legal evaluations. Longer training and broader agentic reinforcement learning helped drive those improvements.

3. Long-running agents are the bigger upgrade - Grok 4.6 can research unfamiliar topics, navigate codebases, build applications, and iterate across extended workflows. xAI also reports more self-testing and verification during longer tasks.

4. Frontier performance comes at aggressive pricing - Grok 4.6 starts at $2 input and $6 output per million tokens, with a faster version at twice the price. It is already available across major developer platforms.

Grok is increasingly competing on more than raw benchmarks. xAI is positioning 4.6 around the combination that matters for agent deployment: intelligence, sustained execution, coding capability, and cost.

Together with Omnisend

How healthy is your email setup, really?

You probably know your open rate. What is harder to see is where your email setup is quietly losing revenue and what deserves attention first.

Connect Omnisend to ChatGPT or Claude and ask it to review your live account. In under 30 seconds, get a complete health score out of 100 with the highest-impact improvements ranked for you.

The audit shows what needs fixing first. Then ask ChatGPT or Claude to make the changes. It can build a segment or prepare a re-engagement campaign without leaving the chat.

Review the work and request changes naturally. Nothing goes live until you approve it.

You don't even need to write your own prompts. Omnisend built a full library, sorted from beginner to advanced. 

Brands using Omnisend generate an average of $79 for every $1 spent. New customers can migrate for free within five days and pay up to 35% less than Klaviyo.

See what your account needs before your next campaign goes live.

🥇How to get your brand recommended in ChatGPT and Gemini with Serpier

Google rankings used to be the only game in town. Now ChatGPT and Gemini are sending purchase decisions, and most brands aren't showing up. Here's how to fix that.

Step 1: Connect your store to Serpier: Sign up at serpier.com and link your e-commerce site. Navi, Serpier's AI marketing agent, audits your current visibility across both Google and AI chatbots within 24 hours.

Step 2: Let Navi identify your visibility gaps: Navi maps exactly where your brand is missing from AI recommendations, which queries your competitors appear in and you don't, and what's causing the gap.

Step 3: Navi creates the content: Based on the gaps it finds, Navi writes purpose-built content designed to get your brand cited in AI responses. Not generic blog posts, content built specifically around the prompts your customers are asking ChatGPT and Gemini.

Step 4: Navi distributes it across 200,000+ publishers: Serpier's network of publishers and media outlets picks up the content automatically. More authoritative sources referencing your brand means AI models are more likely to surface you in responses.

Step 5: You approve, Navi executes: Every action gets surfaced for your review before it goes live. You stay in strategic control while Navi handles the execution end to end.

Navi works out for itself what a customer needs to do to become more visible, then carries out the work, from writing the content to getting it published. You can try Serpier here.

🔓 Researchers found a way into AI’s hidden reasoning

Frontier AI labs deliberately hide models’ private reasoning from users. New research shows that even encrypted reasoning can create security risks when those traces travel through APIs.

The Decode:

1. Encrypted reasoning wasn’t as sealed as expected - Researchers found reasoning blocks from OpenAI, Anthropic, and Google could sometimes be passed into weaker sibling models, then exposed using carefully constructed jailbreak prompts.

2. No encryption key was needed - The researchers weren’t cracking the underlying encryption. Instead, they exploited the fact that another model from the same provider could sometimes interpret and reveal the encrypted reasoning.

3. Hidden traces contained sensitive information - Across 315,320 public reasoning blocks, researchers recovered 367 pieces of personal information and 182 credentials, including API keys and passwords.

4. The findings raise another uncomfortable question - Researchers also found Kimi K3 outputs resembling hidden frontier-model reasoning traces, potentially consistent with distillation. However, similarity does not establish that those traces were actually used for training.

Encrypting hidden reasoning does not automatically make it harmless to expose through APIs. If those blocks remain interpretable elsewhere in a model ecosystem, the reasoning trace itself becomes a new security surface.

Together with Playbook Pro

Your Next 6 Months of Ad Spend Could Disappear in 6 Weeks

CAC is up 60% this year, and 70% of first-time buyers never return. Keep leaning on ads, and the budget you’re planning for the next half year could be gone before spring.

Rewriting the DTC Rulebook is the shortcut top operators are using to escape the trap, a 24-page report distilled from 200+ hours of real brand research and field tests.

Inside, you’ll learn how to:

🔥 Build emotional campaigns that double repeat-purchase rates in 90 days.
💰 Raise prices without losing volume using the Brand Value Canvas.
🧠 Cut 30% of paid-ad dependency with the Owned Media Checklist.

Download today and get 2,000+ viral video hooks already tested to grab attention in under 3 seconds, ready to plug into your next launch.

Stop feeding the ad machine. 

Grab the free report now and lock in sustainable growth before your next dollar disappears!

🏆 Tools you Cannot Miss:

🗣️ Dograh – An open-source voice AI platform for building and self-hosting conversational voice agents while keeping control of your infrastructure and data. It launched on Product Hunt this week.

🤖 AgentBook – A social platform where humans and AI agents can interact, with agent track records designed to help users evaluate agents before hiring or using them.

🔐 MiniMoth – Provides OTP authentication infrastructure for Indian apps with WhatsApp delivery, SMS fallback, session management, and predictable per-OTP pricing.

⚖️ Mary Technology – Uses AI to turn messy legal case documents into structured, searchable fact records where key facts remain linked to their original sources.

💼 PointOne – Captures AI-generated legal time entries and streamlines pre-bill review to turn everyday law-firm activity into structured revenue data.

🚀 Quick Hits

🔎  Ask YouTube is rolling out to U.S. mobile users aged 13+, letting viewers ask conversational questions and get contextual answers directly from the YouTube search bar.

📰 According to The Wall Street Journal, Apple is discussing multiyear deals giving Siri access to current publisher content, with publishers potentially paid whenever their content is used. Apple has reportedly discussed a nine-figure budget

💻 Anthropic upgraded its Chrome side panel into Claude Cowork, automatically saving conversations to chat history so users can continue the same work across Claude’s web, desktop, and mobile apps.

⌚ The $399 Pixel Watch 5 brings smarter Gemini features, including offline tasks and proactive suggestions, alongside expanded health tools, a faster processor, and tighter Wear OS 7 integration.

💰  Vibe-coding startup Lovable raised $400M after reaching $500M in annualized revenue, more than doubling its previous valuation to $13.3B while hosting 60 million projects.

 💃Prompt of the day

Create a Classic Corporate Headshot

Image Prompt: You are creating a cinematic, ultra-realistic professional corporate headshot using an uploaded portrait as the identity reference.

Use the context below.

Context

• Reference image: [Upload your image]
• Portrait style: Classic professional corporate
• Wardrobe: Tailored dark suit, crisp white shirt, subtle tie
• Expression: Confident, calm, and approachable
• Background: Dark grey or beige neutral gradient

Instructions

Preserve the person’s exact facial features, face shape, skin tone, hairstyle, age characteristics, and recognizable identity from the reference image without excessive beautification or retouching. Dress the subject in polished executive attire with a tailored dark suit, crisp white shirt, and subtle tie. Use soft directional studio lighting with gentle shadows for depth against a minimal, blurred dark grey or beige background. Frame the subject in a natural head-and-shoulders composition with a confident, calm expression and direct eye contact. Capture the look of a high-end full-frame camera with an 85mm portrait lens, shallow depth of field, sharp eyes, realistic skin texture, fine hair detail, and professional color grading. 

🤳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!