Pentagon's Modern Models; Anthropic & Your Memories; Big Brother King

Today's AI Outlook: 🌥️

Pentagon May Have Opus 5 Running Before You

According to AI Secret, the next generation of Anthropic’s Claude, internally dubbed Opus 5, is already running in classified environments.

After President Trump ordered a federal ban on Claude, U.S. forces allegedly used a custom Claude model inside military intelligence systems during strikes on Iran. Anthropic confirmed it has a classified deployment for the Pentagon. Insiders suggest it operates at an Opus 5 or even 5.5 level, well beyond the consumer facing Opus 4.6 tier.

The system is said to run on isolated, classified cloud infrastructure with full compute allocation, handling strategic reasoning, target identification, and live scenario simulations. Compute devoted to these systems reportedly doubles every four months.

Why it matters

If accurate, the frontier is no longer defined by ChatGPT leaderboards or prompt engineering tricks. It is defined by who gets the most compute and where it gets pointed. The gap between public AI and classified AI may now be measured in orders of magnitude.

When frontier systems are optimized for battlefield acceleration rather than medicine or infrastructure, it reframes the entire alignment debate.

The Deets

• Classified Claude deployment confirmed by Anthropic
• Alleged Opus 5 or 5.5 capability level
• Compute scaling reportedly doubling every four months
• Embedded directly into command and intelligence pipelines

Key takeaway

Opus 5 looks less like an upgrade and more like a threshold moment. Frontier AI is no longer theoretical. It is operational.

🧩 Jargon Buster - Classified cloud infrastructure: Isolated, government secured computing environments that are physically and digitally separated from public cloud systems.

Supreme Court Sidesteps AI Copyright Bomb

The U.S. Supreme Court declined to hear the biggest AI authorship case yet, leaving intact lower court rulings that say only humans can hold copyright.

The case centered on computer scientist Stephen Thaler, who sought copyright protection for artwork generated by his AI system DABUS. The Copyright Office denied the claim. Federal courts agreed, calling human authorship a “bedrock requirement.”

The Justice Department backed the Copyright Office, arguing copyright law was written for human creators. The appeals court also noted that Thaler could have claimed authorship himself rather than listing the AI as the creator.

Why it matters

AI content is flooding creative industries. The legal framework was built for paintbrushes, not neural nets. By ducking the case, the Supreme Court keeps the humans only standard in place but avoids addressing the deeper structural tension.

Studios, publishers and tech companies with billions at stake will not let this rest.

The Deets

• Case centered on AI system DABUS
• Lower courts ruled only humans qualify as authors
• DOJ backed the Copyright Office
• Supreme Court declined review

Key takeaway

For now, AI cannot own what it creates. The human in the loop remains legally indispensable.

🧩 Jargon Buster - Authorship standard: The legal requirement that a work must originate from a human creator to qualify for copyright protection.

⚔️ Power Plays

Anthropic Opens The Memory Floodgate

Anthropic rolled out a tool that lets users port saved preferences and context from other chatbots into Claude with a single copy paste workflow. Users generate a memory export prompt inside ChatGPT, Gemini or Copilot, paste it into Claude, and the transfer activates within 24 hours.

The company also opened Claude’s memory feature to free users and upgraded Claude Code with auto memory that saves debugging patterns and workflow habits across sessions.

The timing coincides with backlash over OpenAI’s Pentagon deal, with some users reportedly switching platforms.

Why it matters

Switching costs are the moat. Well, they were. By making it painless to bring over saved instructions and preferences, Anthropic turns viral outrage into retention strategy.

Memory is no longer a premium feature. It is a competitive weapon.

The Deets

• One step memory migration tool
• Supports ChatGPT, Gemini and Copilot exports
• Memory now available to free users
• Claude Code auto saves project context

Key takeaway

In the AI wars, context portability may matter more than raw model benchmarks.

🧩 Jargon Buster - Persistent memory: An AI feature that retains user preferences and context across conversations instead of resetting each session.

Cursor Boss Bows To The Agents

Cursor CEO Michael Truell declared that AI coding has entered its third phase, dominated by autonomous cloud agents that run for hours with minimal supervision. Inside Cursor, 35% of merged pull requests are already generated by autonomous cloud agents. Agent usage has grown 15x in a year and now surpasses traditional tab completion.

Local machines struggle with long running multi agent loops, so Cursor is shifting compute to cloud VMs. Developers move from writing code to auditing outcomes.

Cursor reportedly surpassed a $2B annualized revenue run rate, doubling in three months, driven largely by enterprise customers.

Why it matters

This is not a UI tweak. It is a workflow inversion. The leverage shifts from typing speed to review judgment.

The IDE era is giving way to agent infrastructure.

The Deets

• 35% of merged PRs AI generated
• Agent usage flipped from 2.5 to 1 minority to majority
• 15x growth in one year
• $2B annualized revenue run rate

Key takeaway

Developers are becoming supervisors of autonomous systems. The keyboard is no longer the center of gravity.

🧩 Jargon Buster - Cloud agent: An AI system running on remote servers that can execute multi step tasks autonomously over long periods.

🛠 Tools & Products

Burger King Installs AI Manners Monitor

Burger King is piloting an OpenAI powered chatbot called Patty inside employee headsets across roughly 500 locations. The system listens for phrases like “please” and “thank you,” scoring stores on friendliness.

Managers can access performance reports. The system also connects to kitchen equipment, inventory and drive through audio via a centralized cloud POS that updates menus and stock in under 15 minutes.

Courtesy is now a measurable KPI.

Why it matters

This is workplace AI moving from logistics into behavioral scoring. Employees are fulfilling orders while also producing data points.

Surveillance disguised as optimization tends to generate backlash.

The Deets

• 500 pilot locations
• Real time speech monitoring
• Friendliness scoring dashboards
• Integrated with kitchen and POS systems

Key takeaway

When labor feels mic’d and graded, resentment scales faster than efficiency gains.

🧩 Jargon Buster - KPI: Key Performance Indicator, a measurable value used to evaluate performance against objectives.

🔬 Research & Models

Alibaba’s Tiny Model Packs A Heavyweight Punch

Alibaba released Qwen3.5 Small, a family of four open source AI models ranging from 0.8B to 9B parameters. The 9B model reportedly outscored an OpenAI model 13 times its size on graduate level reasoning and multilingual benchmarks.

All models support text, images and video natively, and are free for commercial use.

Why it matters

Frontier models grab headlines. Small models drive adoption. Running powerful multimodal AI on laptops and phones without massive cloud spend changes the economics for startups and enterprises alike.

The Deets

• Four models from 0.8B to 9B
• 9B outscored 120B class model
• Multimodal support
• Open source commercial license

Key takeaway

Efficiency is the new arms race. Intelligence density is becoming a competitive metric.

🧩 Jargon Buster - Parameter count: The number of adjustable weights inside a model, often used as a rough proxy for size and capacity.

⚡ Quick Hits

• Anthropic expanded Claude memory to free users and added cross platform import tools.
• OpenAI’s Pentagon deal triggered reputational backlash and user switching.
• Apple is reportedly considering Google servers to power a Gemini based Siri upgrade.
• Google updated Gemini for Google Home with improved automation and live camera queries.

🧰 Tools Of The Day

• Viktor: An AI coworker that lives inside Slack, connects to 3,000 plus tools, builds reports, runs ad campaigns and automates recurring workflows. Comes with $100 in free credits.
• openai whisper: Free local transcription model that converts video to text and subtitles in minutes.
• Qwen3.5 Small: Alibaba’s open source multimodal models optimized for laptops and mobile deployment.

Today’s Sources: AI Secret, The Rundown AI