OpenAI Answers Anthropic in 90 Minutes

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 launches Claude Opus 5.5 with better performance at a lower cost

Anthropic has released Claude Opus 5.5, the first model in its new 5.5 family, improving coding, professional work, writing, and safety while making its flagship model significantly cheaper and faster to use.

The Shift:

Built for Bigger Jobs - Opus 5.5 improves at coding, research, financial analysis and other complex professional tasks, including long projects such as full codebase migrations and software audits. You can get started with Opus 5.5 here. 

40% Cheaper to Run - Anthropic says typical workloads cost 40% less than Opus 5, with prices at $4/M input and $20/M output tokens. The model also generates output 30%+ faster.

Less “Claudish” Writing - Anthropic addressed criticism of Claude’s writing style, saying Opus 5.5 uses less jargon, puts important information first and follows users’ own writing and style instructions more closely.

Best Safety Score Yet - Opus 5.5 achieved Anthropic’s “best score to date” on its internal alignment testing, with fewer unwanted actions and stronger prompt-injection defenses.

Opus 5.5 is Anthropic’s first model release since its call for “pacing the frontier,” combining continued capability gains with stronger safety measures. It is available now, with Sonnet 5.5 and Haiku 5.5 coming in the weeks ahead.

Together with AirOps

The Buying Decision Starts on Pages You Don’t Own 

Your team may appear strongly in Google AI Mode and disappear when the same buyer asks Claude. That blind spot can cost consideration while your reports still look healthy.

On September 29, Foundation CEO Ross Simmonds and AirOps search marketing lead Josh Spilker will reveal how six AI platforms build software recommendations, using evidence from nearly 380,000 answers and 3 million citations.

By the end of the session, you’ll be able to:

• Decide whether to strengthen your vendor pages or earn more review-site coverage, based on which source type captures the most citation share.
• See how Google AI Overview and AI Mode build recommendations differently from ChatGPT, Claude, Gemini, and Perplexity.
• Rank G2, Reddit, YouTube, LinkedIn, Capterra, and Gartner by their influence before committing your next quarter’s content budget.

You’ll leave with a repeatable way to understand visibility shifts across AI-generated shortlists and decide what your team should fix first.

Can't attend the session live? Register anyway, and you’ll get the recordings in 24 hours.

🥹How to stop AI from confidently making things up using the truth prompt

AI models sound equally confident whether they are right or wrong. This one setup forces them to tell you the difference before you act on anything. Here’s how.

Step 1: Add the Truth Prompt to your custom instructions - Open ChatGPT Settings > Personalization > Custom Instructions. For Claude, create a new project and paste this into the system prompt:

“For every factual claim you make, do the following: label it as either VERIFIED (you are highly confident and can cite a source), LIKELY (you believe this is true but are not certain), or UNCERTAIN (you are guessing or extrapolating). If you do not know something, say so directly. Never fill a gap with a confident-sounding guess.”

Step 2: Add a confidence score to research tasks - For any research question, add this to your prompt:

“Before answering, rate your overall confidence from 1 to 10. Tell me which parts of your answer you are most confident in and which parts I should independently verify.”

Step 3: Force source separation - “Separate your answer into two sections: what you know from reliable training data, and what you are inferring or estimating. Label each section clearly.”

Step 4: Use it on any high-stakes output - Before acting on any AI-generated legal, financial, medical, or strategic advice, paste this:

“Review your last response. Flag every claim that a fact-checker could dispute. Tell me what you would need to verify each one and where I should look.”

Step 5: Test it immediately - Ask any AI a question you already know the answer to. See whether the confidence labels match reality. Calibrate from there.

One setup. Every AI response becomes auditable.

⚡ OpenAI launches GPT-6 Sol and Luna at half the price

OpenAI has introduced GPT-6 Sol and GPT-6 Luna, bringing many of GPT-6 Astra’s advances in coding, factuality, computer use and professional work to faster, significantly cheaper models.

The Shift:

☀️ GPT-6 Sol

Sol targets complex professional and coding work. At its highest effort, it beats Claude Opus 5 on AutomationBench at just 9% of the cost per task, while competing closely with more expensive Claude models on coding benchmarks.

API prices are 50% lower at $2/M input and $10/M output tokens. Sol also makes about half as many factual mistakes as GPT-5.6 Sol in OpenAI’s difficult internal factuality test.

🌙 GPT-6 Luna

Luna is built for fast, high-volume work at much lower cost, while improving over GPT-5.6 Luna across coding, professional tasks and computer use. Its Artificial Analysis Intelligence Index score also rises slightly over its predecessor.

Luna costs just $0.10/M input and $0.50/M output tokens, 50% cheaper than GPT-5.6 Luna. OpenAI says it can deliver coding performance comparable to significantly more expensive models on some tests.

Astra remains OpenAI’s most capable model, while Sol and Luna bring GPT-6 improvements to more everyday workloads at much lower costs. Sol and Luna are available in ChatGPT Work, Codex and the API, while Free and Go users can access Luna on desktop. They are not yet available in regular Chat. 

Together with Playbook Pro

The Playbook for Profit That Funds Itself

Great operators don’t just count cash, they make it multiply. They use timing and leverage to turn everyday revenue into rapid expansion.

Financial Alchemy is your cheatsheet packed with real operator insights and ready-to-use templates. Inside, you’ll learn how to:

💰 Master cash-conversion cycles to grow while suppliers fund your inventory, just as Wayfair scaled with a –34-day cash conversion cycle.

📊 Use the Cash Flow Simulator to forecast scenarios and uncover up to $100K in hidden operating cash within a single quarter.

🧮 Pinpoint profit levers with the Unit Economics Analyzer to cut CAC payback windows from 6 months to as little as 3 months.

🤝 Negotiate better terms using the Supplier Toolkit, a framework inspired by Skinny Pasta’s strategy that extended supplier terms to 120 days and fueled 100%+ Y-o-Y growth.

Start compounding capital instead of chasing outside funding.

Get your ebook now and turn your cash flow into the most powerful growth engine!

🏆 Tools you Cannot Miss:

📣 Naise AI – Turns a marketing brief into social content, campaign visuals, influencer outreach, PR activity, and market research while managing execution across multiple channels. 

🎬 VideoFlow Studio – Reads your product website and uses a coding agent to plan, render, review, and deliver an editable product trailer based on the actual product and positioning. 

📊 Anomalo – Continuously monitors company data for quality problems and unusual changes, then uses AI agents to investigate issues and surface analyst-style insights without manual querying. 

🪄 WZRD – Turns documents, presentations, spreadsheets, and forms into interactive AI experiences that users can speak to, question, and explore instead of simply viewing static files. 

💬 ToneBird – Drafts contextual replies inside email and messaging apps by using past conversations and connected files to understand who you are talking to and what was previously discussed.

🚀 Quick Hits

🤖 Tesla is getting Grok Bot, an AI agent that can manage emails, calendars, files and tasks hands-free. It reportedly reached 400,000 weekly users, but requires the $300/month SuperGrok Heavy plan.

⚡ Texas Gov. Greg Abbott ordered a halt on state permits sought by data centers until ongoing audits assess their impact on the electric grid and water resources, alongside tougher water-use reporting enforcement. 

🇺🇸 President Trump said U.S. government documents will now use “super intelligence” instead of “artificial intelligence,” arguing the word “artificial” makes the technology sound fake and hoping other countries follow. 

📱 Qualcomm unveiled Snapdragon 8 Elite Gen 6 and Elite Extreme Gen 6, emphasizing on-device AI agents, with the Extreme chip capable of running 30-billion-parameter MoE models locally on smartphones.

🤖 Meta says Muse was heavily inspired by OpenClaw but built from scratch, with the team intentionally adopting similar workspace concepts while aiming to make personal AI agents accessible to billions of users.

🦾 What's trending on socials

🌍 One website, 600+ live cameras - Someone built a website that lets you spin the Earth and jump between 600+ live cameras worldwide, watching cities and locations broadcasting in real time from one browser tab.

🎬 Someone built a free netflix alternative - A developer built Kino, a website claiming to let anyone stream movies, TV series and anime in HD for free, with no ads.

📱 Apple’s $250M Siri settlement opens for claims - Eligible U.S. iPhone owners can now file claims in Apple’s $250 million Siri AI settlement, with payouts reaching up to $95 per eligible device.

🎓 A new school built for the AI era - The Horowitz Andreessen Academy is launching in San Francisco with $42M in funding, offering a tuition-free founding fellowship focused on building, backed by partners including OpenAI, Anthropic, Google, Meta and NVIDIA.

🌀 Turning a 2D video into 4D - A researcher reconstructed space and time from ordinary 2D video, converting each frame into 3D Gaussian representations and preserving exactly where each moment occurred along a time dimension.

🤳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!