Anthropic finally fixed Claude’s biggest problem

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.

🧠 Anthropic’s new Claude model pushes further on research and coding

Anthropic released Claude Fable 5.1, upgrading its previous flagship with stronger performance on long coding tasks, scientific research, and knowledge work while reducing unwanted safety refusals.

The Shift:

Fable 5.1 sets a new benchmark high: The model scored a record 66 on Artificial Analysis’ Intelligence Index, with particularly large improvements across scientific research and knowledge-work evaluations.

Long tasks get more capable but potentially expensive: Anthropic estimates typical workloads cost 25% less, although Artificial Analysis found maximum-effort tasks cost 20% more because Fable 5.1 produces substantially more text.

Safety filters intervene less often: Compared with Fable 5, refusals reportedly dropped 60% for cybersecurity tasks and 85% for basic medical and biology questions following complaints about excessive filtering.

Anthropic also released a less-restricted version - Mythos 5.1 uses the same underlying model with fewer guardrails, but access is limited to screened U.S. cybersecurity and biology researchers.

Fable 5.1 focuses on improving both capability and usability, particularly for longer, technical work. The reduced refusals also show Anthropic adjusting how aggressively its safety systems intervene without removing protections entirely.

Together with Omnisend

They switched before Black Friday and made $113K

Weeks before BFCM, one broken flow can wipe out months of planned revenue. An overdue migration is harder to ignore when platform costs are already cutting into the season’s upside.

Ecom2Win moved a client from Klaviyo to Omnisend just before Black Friday. The migration took 24 hours, and its automations were rebuilt and checked within four days.

The switch gave the team:

• Up to 35% lower platform costs compared with Klaviyo.
• Free migration of automations, segments, forms, and templates within five business days.
• More room to grow, with Omnisend customers averaging $79 in revenue for every $1 spent.

The client generated more than $113K in attributed email revenue during its first Black Friday on Omnisend. Email produced 53% of total store revenue for the period.

The team entered BFCM with working flows and confidence that every send could earn.

Read Jocelyn’s full story to see how Ecom2Win switched without sacrificing peak-season revenue. Waiting could mean paying more during your most valuable sales period.

🤝How to make ChatGPT write like a human every time

One setup. Every response passes AI detection permanently.

Step 1: Pull AI detection signals - Paste this into ChatGPT:

"Search Wikipedia for 'AI writing detection.' List every pattern that flags content as AI-generated, sentence structure, word choice, phrasing, formatting, everything."

Step 2: Convert it into a standing instruction file - Paste this next:

"Turn that list into strict writing rules, what to never do, what to always do, how to vary tone and structure naturally. Make it dense and specific."

Copy the full output.

Step 3: Add it to permanent memory - Go to Settings > Personalization > Custom Instructions. Paste the framework into the response box. Save. Done. Every future response follows these rules automatically.

Step 4: Test and refine - Run output through GPTZero or Originality.ai. If anything flags, paste it back with:

"Rewrite this to pass AI detection without changing the meaning."

Add any new flagged patterns back into your Custom Instructions. The framework compounds over time.

💻 Perplexity brings hybrid AI computing to Mac

Perplexity introduced hybrid compute for Mac, allowing its Computer agent to combine powerful cloud models with local AI for files and information users want to keep on their device.

The Shift: 

Local and cloud AI work together - Cloud models handle web search, planning, and complex reasoning, while local models process private files, sensitive information, and actions directly on the Mac.

A privacy gate controls what leaves the Mac - An on-device classifier detects sensitive details and can mask information, keep it local, block requests, or ask permission before sending anything to the cloud.

Local AI can be accessed remotely - Users can send prompts from an iPhone while Perplexity Computer runs on their Mac, enabling an always-on setup using devices such as a Mac mini.

Three local models launch with support - Hybrid compute includes Gemma 4 E4B, Qwen3.6 35B-A3B, and Perplexity’s model, requiring an Apple silicon Mac with macOS 15+ and at least 24GB memory.

Perplexity is trying to combine frontier cloud intelligence with the privacy of local AI. Hybrid compute lets sensitive information remain on-device while still using cloud models for tasks requiring stronger reasoning, research, and planning. You can try hybrid compute on Mac here.

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:

🎯 Creatium Coach – Creates personalized learning sessions around a specific goal using interactive lessons, simulations, roleplays, and assessments to practice the skill rather than only explain it.

🌐 Sider Code – Lets you modify how websites look and behave with natural-language instructions, including restructuring pages, hiding elements, adding actions, and saving changes for future visits.

🔐 TrustedRouter – Routes AI requests across multiple models through one interface with privacy-focused controls for teams that need to manage model access and data handling.

📞 ThunderPhone – Provides infrastructure for building AI phone agents that handle inbound and outbound calls, use business tools, transfer to humans, and operate across 40+ languages.

🎙️ Tovel AI – Records calls and meetings, generates transcripts and summaries, identifies follow-up actions, and prepares approved tasks for calendars, CRMs, and other workflows.

🚀 Quick Hits

🎬 Google is reportedly seeking licensing deals with major Hollywood studios to train AI models on copyrighted characters and films, offering potentially huge payouts. The strategy could give Google valuable IP while helping make generative AI feel more mainstream and culturally acceptable.

🛡️ OpenAI delayed development of its unreleased Astra model suite after a separate unreleased model caused a major security incident, giving the company additional time to strengthen its safety work.

⚖️ Apple is accusing OpenAI of risking the destruction of evidence in its trade secrets lawsuit, after a former Apple employee’s MacBook allegedly revealed discussions about wiping forensic data and use of a confidential Apple circuit schematic at OpenAI. 

🚜 John Deere is testing “JD,” an AI assistant that uses farmers’ own field, machine and operational data to answer questions about equipment settings, fuel usage and harvest timing, with the goal of improving farm productivity and profitability.

📱 Google is rolling out five Android updates, including Motion Assist to reduce motion sickness, Gemini-powered Guided Vision for blind and low-vision users, item-location memory in Find Hub, and new Google Messages personalization and Keep integration.

🤳What’s trending on socials 

🍫 Snickers feeds hungry AI - Snickers created a digital “HungrAI” bar users can feed to chatbots when they give bad answers, prompting the AI to reconsider and improve its response. 

🪥 Dyson puts AI in toothbrushes - Dyson launched a $499 AI-powered toothbrush with a built-in 100,000-pixel macro camera that analyzes your mouth while you brush.

🔎 Claude takes on people search - An open-source people-search tool built with Claude claims the top spot on People Search Bench, offering searches for $0.0089 per lead with no markup.

📹 App maps nearby security cameras - A new app shows security cameras around a user on an interactive map, including their locations and approximate fields of view.

🫧 The AI Bubble hits your portfolio - A comedian jokes that even people who don’t directly invest in AI are financially exposed through index funds packed with AI-linked companies. 

🤳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!