Good morning. It's Tuesday, April 21, and we're covering DeepMind’s coding push, Anthropic’s Pentagon standoff, the rise of AI-generated entertainment, and more.
Amazon Deepens Strategic AI Partnership with Anthropic
As part of the deal, Amazon is investing an immediate $5 billion into Anthropic, with plans to invest up to $20 billion more in the future tied to commercial milestones. This builds on an initial $8 billion investment made previously by Amazon. In return, Anthropic has committed to spending more than $100 billion over the next decade on Amazon Web Services (AWS) technologies. To train and power its advanced Claude AI models, Anthropic will utilize Amazon’s custom silicon—specifically its Trainium chips and Graviton processors—securing up to 5 gigawatts (GW) of capacity.
The "AI Dividend": Insurance for the Automation Age
A Democratic congressional candidate is proposing direct payments tied to AI-driven job loss, framing the policy as insurance rather than redistribution. Alex Bores unveiled an “AI dividend” plan that would trigger if artificial intelligence significantly displaces workers, as concerns grow over automation’s economic impact. The proposal would be funded through a tax on AI consumption, equity stakes in major AI firms, and tax code changes, with proceeds split between public payments, workforce training, and oversight efforts.
Most Streaming Content to be AI-Generated in 5 Years
iQiyi says AI could generate most of its films and shows within five years, a shift that would upend traditional production and redefine its platform. The Beijing-based streaming company is restructuring its business to focus on AI-generated content and more social, short-form video experiences. It launched the Nadou Pro toolkit to automate filmmaking—from scripts to final rendering—and plans to release a commercially viable AI-generated film as early as summer 2026.
POWERED BY SALESFORCE
Natural Language In, Production Code Out
Agentforce Vibes is redefining what it means to build software. Describe what you need — Vibes turns it into functional code and working applications, fully grounded in your project context. Enterprise-grade security via the Salesforce Trust Layer keeps every output production-ready.
Teams are already accepting over 1.2 million lines of agentic code per month, with AI generating up to 25% of their total output. Vibes 2.0 now enables Multi-Agent Swarms across models like Claude Sonnet and GPT-5 to build complex applications in parallel — following a 22x surge in monthly usage.
Your next app starts with a sentence. Explore Agentforce.
VIDEO
Opus 4.7 Just Dropped...
CODING
Google DeepMind Forms Strike Team To Close AI Coding Gap
Google DeepMind has assembled a “strike team” of engineers and researchers to improve its AI coding models, prompted in part by advances from Anthropic. Internal comparisons suggest Anthropic’s tools outperform Google’s Gemini in code generation, driving urgency among senior leaders including co-founder Sergey Brin.
Currently, AI generates about 50% of Google’s code, versus Anthropic’s reported near-total automation. The initiative focuses on long-horizon coding tasks and increasing internal adoption, including training models on Google’s proprietary codebase. → Read the full article here. (Paywall)
LABS
Anthropic Is Blacklisted by the Pentagon — Courted by Everyone Else
The same model that got Anthropic barred from Pentagon contracts is now being actively sought by Treasury, State, the Fed, and foreign central banks. CEO Dario Amodei is meeting White House chief of staff Susie Wiles today — a sign of a thaw in a standoff that began when the Defense Department declared Anthropic a national security supply chain risk after the company refused to allow its AI to power mass domestic surveillance or autonomous weapons systems.
The Office of Management and Budget is now setting up guardrails to let federal agencies begin using Mythos "in the coming weeks," according to an internal email from federal CIO Gregory Barbaccia. The model — rolled out so far to only a few dozen organizations under Project Glasswing — can identify zero-day vulnerabilities across major operating systems and browsers, often with minimal human input. Anthropic has itself warned that adversaries could reach similar capabilities within two years. The complication Andrew Ross Sorkin flags: every patch Mythos generates creates a "patch gap" — a window where hackers can study the fix and strike before it's implemented, turning the act of repair into a blueprint for attack. → Read the full article here. (Paywall)
AUTOMATION
Chinese Tech Workers Replicate Themselves With AI
A viral GitHub project called “Colleague Skill” has sparked debate among Chinese tech workers by demonstrating how AI agents can replicate coworkers’ tasks, workflows, and even communication styles. Though created as a spoof, the tool reflects a real trend: companies are asking employees to document their work so it can be automated using AI agents like OpenClaw or Claude Code.
Workers report mixed reactions—some see productivity gains, while others describe the process as reductive and unsettling, raising concerns about job security and identity. The trend has also triggered pushback, including “anti-distillation” tools designed to sabotage attempts to codify human work into AI systems. → Read the full article here.
POWERED BY SALESFORCE
Natural Language In, Production Code Out
Describe what you need — Agentforce Vibes turns it into functional, production-ready code. Teams are already accepting over 1.2 million lines of agentic code per month. Vibes 2.0 adds Multi-Agent Swarms across models like Claude Sonnet and GPT-5 to build complex applications in parallel.
Your next app starts with a sentence. Explore Agentforce.
NEWS
What Else is Happening
Meta Plans 8,000 Job Cuts: The company will lay off 10% of workforce starting May while ramping AI spending to $115–135 billion, reorganizing teams into AI-focused units to boost productivity and margins.
Starbucks Tests ChatGPT Ordering Tool: New AI feature suggests drinks based on mood, images, or context, aiming to learn customer preferences and friction points as “agentic commerce” experiments expand.
Google Targets NVIDIA With AI Chips (Paywall): Alphabet plans new inference-focused chips as demand surges, with customers like Meta and Anthropic adopting them to run AI models faster and cheaper.
AI Music Floods Streaming Platforms: Platform says 44% of new uploads are AI-generated, with most streams deemed fraudulent and demonetized despite minimal impact on listener recommendations.
Adobe Expands AI Agent Strategy: Company launches CX Enterprise with autonomous agents and 30+ partnerships to automate marketing and sales, as it battles investor concerns and rising AI-native competition.
