Welcome back! We are watching the definition of "AI" splinter into two very different realities. On the software side, the technology is becoming intensely militarized. The biggest labs are now being forced to choose between strict corporate ethics and massive defense contracts, and the fallout is rewriting the geopolitical map of Silicon Valley.
In today’s Generative AI Newsletter:
OpenAI signs a Pentagon deal after Anthropic gets blacklisted.
Honor unveils a robot smartphone that moves and nods.
Figma partners with OpenAI to bridge design and code.
OpenArt offers a sandbox for precision AI images.
Latest Developments
Anthropic Refused The Pentagon. OpenAI Just Took The Deal.
The AI industry split along geopolitical lines as Anthropic was blacklisted by the U.S. government for refusing to lift safety filters on military operations, while OpenAI immediately signed a new Pentagon contract. Simultaneously, OpenAI secured a conditional $110 billion funding round that values the company at $840 billion, provided it hits specific AGI milestones tied to human productivity.
The Details:
Anthropic Blacklisted: Dario Amodei banned Claude's use for autonomous weapons or mass surveillance of Americans. The Pentagon labeled the company a "supply chain risk" equivalent to a foreign adversary.
OpenAI’s New Deal: Sam Altman signed a major Pentagon contract shortly after Anthropic's exile. Government officials stated the deal permits all "lawful purposes" without exceptions.
Active Deployment: CENTCOM reportedly utilized Claude’s intelligence for target identification and combat simulations during recent strikes in Iran.
Claude reaches #1: Claude has been rising to the top of the free app rankings in Apple’s US App Store. It overtook OpenAI’s ChatGPT to claim the number one spot.
While Sam Altman claimed the contract includes strict safety red lines, government officials clarified that the deal allows for "all lawful purposes," effectively bypassing the human-in-the-loop safeguards Anthropic refused to compromise. This shift occurred despite an open letter signed by over 360 AI employees urging companies to reject military contracts.
Special highlight from our network
AI gives you better results when it’s connected to real, up-to-date data. But access alone isn’t enough.
MongoDB’s document model brings structured and unstructured data together, making it easier to power accurate, context-aware RAG applications with vector search.
The result: smarter apps you can trust.
Learn how to build it step by step.
Special highlight from our network
Few understand what actually changes, when everyone is talking about AI.
Rebecca Hinds shares key findings from the AI Transformation 100 report and what separates high-performing teams.
You’ll discover:
• Where teams reclaim time
• Why adoption often stalls
• What professionals should do now
AI will change how everyone works.
Those who adapt early win.
Honor is Building the First Phone With an AI Robot
Honor officially unveiled its Robot Phone at MWC Barcelona. This device is the first embodied AI smartphone to feature a 200MP camera on a motorized three-axis gimbal arm. Instead of a static lens, the phone uses a custom micro-motor system that is 70 percent smaller than standard hardware to give the AI a physical presence. This allows the AI buddy to nod in agreement, shake its head, or even dance to music.
The specs:
Multimodal Perception: The AI uses its 4DoF gimbal arm as a sensory neck to track motion and maintain real-time visual awareness of its environment.
Agentic Shooting: The 200MP sensor is paired with ARRI Image Science and an AI video editor to automate professional-grade cinematography.
AI Object Tracking: The hardware leverages on-device models trained on millions of simulations to lock onto subjects with zero latency.
Embodied EQ: The AI displays "emotional body language" through physical gestures such as nods and tilts to make human-to-machine interaction feel natural.
The launch coincided with the debut of Honor’s first humanoid assistant, which famously executed a backflip on the MWC stage. Together, these devices represent Honor’s $10 billion Alpha investment plan to transition from a handset maker into an "ecosystem company." Honor is attempting to solve a decade-long stagnation in smartphone design.
Figma Partners With OpenAI To Bake in Support for Codex
The partnership centers on Figma’s MCP (Model Context Protocol) server, which creates a bi-directional bridge between the two platforms. Engineers can now iterate on visual components without leaving their coding flow, while designers can push their work closer to real-world implementation without mastering complex syntax.
Key Features:
Bilateral Workflow: Users can initiate a design in Figma or a script in Codex and sync changes across both platforms in real-time.
Asset Portability: The integration pulls data directly from Figma Design files, Figma Make, and FigJam into Codex for immediate implementation.
Platform Dominance: The move follows the massive success of the Codex MacOS app, which saw one million downloads in its first week of release earlier this month.
Broad Adoption: OpenAI reports that over one million users are now using Codex weekly, making it a critical rival to Anthropic’s Claude Code.
By acting as a neutral "infinite canvas," Figma is positioning itself as the essential interface for the AI-augmented workforce. The irony of this "broadening of builders" is that as the barrier between designer and engineer collapses, the specialized roles of both are being redefined. As Codex product lead Alexander Embiricos noted, the tool "doesn't assume you're a 'designer' or an 'engineer' first user.”
OpenArt: A Creative Sandbox for Precision AI
OpenArt is an AI creative suite that allows access to 100+ models, including Stable Diffusion variants, Flux, and DALL·E, all within a single workspace. It is used by creators and marketing teams who need tighter control over output rather than relying on purely stylistic prompting.
Core functions (and how to use them):
Multi model image generation: Switch between specialized models for photorealism, illustration, anime, or painterly styles without leaving the interface.
Magic Brush editing: Select a specific region of an image and describe changes such as color swaps, object removal, or background replacement.
Sketch to image: Upload a rough layout or drawing to guide composition before rendering a polished result.
Character locking: Use Character 2.0 to maintain a consistent face or identity across multiple scenes.
Custom LoRA training: Train a style or product model using 20 to 30 reference images and trigger it with a keyword.
Try this yourself:
Start in the Canvas workspace, select a fine-tuned base model, upload a reference image, set the aspect ratio, and refine with Magic Brush or Chat to Edit until the output matches a specific campaign need.
