Welcome back to 4IR. Here’s today’s lineup:
Anthropic claims “best coding model in the world” with Claude Sonnet 4.5 - New model hits 77.2% on SWE-bench, runs autonomously for 30 hours, available immediately across major platforms
OpenAI launches shopping in ChatGPT, takes direct aim at Google and Amazon - Stripe-powered instant checkout turns 700M weekly users into massive commerce channel, Etsy stock surges 16%
California Governor signs SB 53, nation’s first frontier AI safety law - Landmark legislation mandates safety protocols and whistleblower protections despite OpenAI and Meta opposition
Hollywood erupts over “AI actress” Tilly Norwood as agencies circle - Digital performer sparks industry-wide backlash, boycott threats from A-list actors
AI startups raise $130M+ across infrastructure, agents, and enterprise automation - Eight major funding rounds signal investor confidence despite economic headwinds
TOP STORY: Anthropic claims “best coding model in the world” with Claude Sonnet 4.5
The story: Anthropic dropped what may be the most important AI model of 2025 today, releasing Claude Sonnet 4.5 with performance that beats every competitor on the industry’s toughest coding benchmark. The model achieved 77.2% on SWE-bench Verified—the gold standard for real-world software engineering—and can run autonomously for up to 30 hours on complex tasks. Available immediately through Claude API, Amazon Bedrock, Google Cloud Vertex AI, GitHub Copilot, and Microsoft Copilot Studio, the release includes major productivity features like file creation, expanded integrations, and a new Agent SDK.
What we know:
77.2% success rate on SWE-bench Verified, beating all competitors
Runs autonomously for up to 30 hours vs. 7-hour predecessor ceiling
61.4% on OSWorld benchmark (up from 42.2% four months ago)
Can autonomously purchase domains, stand up databases, perform SOC 2 audits
Priced at $3 per million input tokens, $15 per million output tokens
Simultaneously launched Claude Opus 4.1 for complex reasoning tasks
New features: file creation/editing, Claude Agent SDK, expanded integrations
Described as “most aligned frontier AI model yet” with improved safety
Available across all major platforms including GitHub and Microsoft Copilot
Why it matters: When a model can code for 30 hours straight and solve real engineering problems at 77% accuracy, we’ve crossed from “helpful assistant” to “actual colleague.” This could change how every software company builds products.
Anthropic just threw down against OpenAI. While everyone focused on o1’s reasoning, Anthropic built something that actually ships production code. The 30-hour autonomous run is the real story—that’s a full work sprint without humans. And the timing is perfect: launch on the same day OpenAI announces shopping features, reminding everyone the real AI race is about capability, not gimmicks. If these benchmarks hold up in practice, we’re looking at developers being 2-3x more productive within months. The pricing matching Claude Sonnet 4 means Anthropic is keeping prices low to win customers fast. This is how you steal market share.
OpenAI launches shopping in ChatGPT, takes direct aim at Google and Amazon
The story: OpenAI launched Instant Checkout in ChatGPT today, partnering with Stripe to let users buy products directly in conversations. Starting with over 1 million Etsy sellers, with Shopify merchants including Glossier, SKIMS, and Spanx coming soon, the feature turns ChatGPT’s 700 million weekly users into a massive commerce channel. The companies open-sourced the Agentic Commerce Protocol, a new standard for AI shopping that directly competes with Google’s search ads and Amazon’s marketplace.
What we know:
Instant Checkout available for 1M+ Etsy sellers immediately
Shopify integration coming with major brands (Glossier, SKIMS, Spanx, Vuori)
Powered by Stripe payment infrastructure
700 million weekly ChatGPT users become commerce channel
Etsy stock jumped 16% on announcement
Agentic Commerce Protocol (ACP) open-sourced
OpenAI takes small merchant fee per transaction
Available to ChatGPT Plus, Pro, and Free users
Direct challenge to Google Search ads and Amazon marketplace
Why it matters: Google built a $200B+ business by sitting between shoppers and merchants. OpenAI just declared war on that entire model by putting commerce directly into AI conversations, potentially capturing transaction fees while bypassing search ads completely.
This is OpenAI’s attack on Google and Amazon. When you can ask “find me sustainable running shoes under $150” and buy them without opening a browser, search advertising dies. The Etsy stock jump proves merchants get it—700 million weekly users is distribution most brands would kill for. But here’s the smart move: the Agentic Commerce Protocol isn’t just defense, it’s offense. OpenAI is setting the rules for AI shopping before Google or Amazon can respond. If this protocol becomes the standard, OpenAI controls the future of online shopping. The timing after Bain’s revenue warning isn’t random—this is OpenAI showing investors exactly how AI companies make money.
California Governor signs SB 53, nation’s first frontier AI safety law
The story: Governor Gavin Newsom signed SB 53 into law today, making California the first U.S. state to regulate frontier AI development with real transparency and safety requirements. The law forces large AI companies including OpenAI, Anthropic, Meta, and Google to publicly disclose safety protocols, creates whistleblower protections, and requires reporting of critical safety incidents. Anthropic supported the bill while Meta and OpenAI fought hard against it, with OpenAI even publishing an open letter begging Newsom to reject it.
What we know:
First comprehensive U.S. frontier AI safety legislation
Requires public disclosure of safety protocols
Whistleblower protections for AI company employees
Mandatory reporting of crimes committed by AI without human oversight
Covers cyberattacks, autonomous harmful actions, deceptive behaviors
Anthropic supported, Meta and OpenAI opposed
Tech firms argued state regulation creates “patchwork” hindering innovation
Will likely influence federal policy and other states
Comes as tech leaders pour millions into super PACs backing light regulation
Why it matters: California just created the template every other state will copy, forcing AI companies into complex compliance across dozens of states whether they like it or not. The tech industry lost the lobbying fight, which means they’ve lost control of regulation.
The split is everything: Anthropic supports it, OpenAI opposes it. Anthropic sees safety as competitive advantage. OpenAI sees it as a roadblock. But here’s what matters: Newsom signed this the same day Anthropic launches a model that runs autonomously for 30 hours. That timing proves why regulation was needed. When AI systems can act independently for days, “move fast and break things” becomes actually dangerous. OpenAI’s lobbying loss stings worse because they just launched shopping features. Regulators are basically saying: you want to handle money and make decisions alone? Fine, but we’re watching everything.
Hollywood erupts over “AI actress” Tilly Norwood as agencies circle
The story: The entertainment industry exploded today after the reveal of “Tilly Norwood,” the first AI-generated actress, with multiple Hollywood talent agencies interested in signing her. Created by Xicoia (spun off from Particle6) and founded by actor/technologist Eline Van der Velden, the digital character was announced at Zurich Film Festival with a full backstory and social media presence. Major actors including Melissa Barrera, Emily Blunt, and Toni Collette fired back hard, demanding agency boycotts.
What we know:
“Tilly Norwood” unveiled as first AI-generated actress
Created by Xicoia, founded by actor/technologist Eline Van der Velden
Multiple Hollywood agencies reportedly circling to sign
Goal: make her “the next Scarlett Johansson or Natalie Portman”
A-list actors demanding agency boycotts
Concerns: labor displacement, consent, composite likeness without permission
Van der Velden defends as “art,” not replacement for humans
Announced at Zurich Film Festival
Why it matters: The AI actress controversy proves every SAG-AFTRA worry about AI rights was correct. When agencies start signing digital performers, human actors face real competition that works 24/7, never ages, and costs nothing after creation.
The timing is brutal—actors just fought strikes over AI rights, and now agencies are signing up a computer program. Van der Velden calling it “art” is just PR speak for “please don’t sue us,” but the damage is done. Every actor watching this knows their face could be next, their roles replaceable, their careers potentially over. The backlash isn’t about one digital character—it’s about watching an entire profession get automated in real-time. And here’s the economics: if one AI actress can play every role, why would studios pay millions per film for humans? This isn’t a future problem. It’s happening right now.
AI startups raise $130M+ across infrastructure, agents, and enterprise automation
The story: Eight major AI funding rounds totaling over $130 million landed today, showing investors still have confidence despite economic concerns. Anything raised $11M at $100M valuation after hitting $2M ARR in just two weeks. Alvys secured $40M Series B for freight logistics automation. Paid raised $21M seed for AI agent billing infrastructure. Maximor emerged from stealth with $9M for AI-native finance platforms, while Gain raised $12M for autonomous procurement agents.
What we know:
Eight funding announcements totaling $130M+
Anything: $11M Series A, $100M valuation, $2M ARR in two weeks
Alvys: $40M Series B for transportation management, tripled revenue two years running
Paid: $21M seed for billing infrastructure for AI agents
Maximor: $9M seed for AI-native finance automation
Gain: $12M seed for autonomous procurement agents
European rounds: Polars (€18M), Lexroom.ai (€16M), Corintis ($24M)
Focus areas: enterprise automation, agent infrastructure, specialized verticals
Why it matters: While Bain warns about AI revenue shortfalls, investors are writing big checks for companies solving specific automation problems rather than building general AI. The market is getting smarter about where AI actually makes money.
Follow the money: investors aren’t funding another ChatGPT clone. They’re backing companies that automate procurement, billing, and logistics—the boring stuff that actually generates revenue. Anything hitting $2M ARR in two weeks shows you can find product-market fit instantly when you solve real problems. The $130M across eight deals isn’t one mega-round—it’s spread across different problems, which is what sustainable AI economics looks like. While OpenAI adds shopping and Anthropic flexes coding power, these startups are quietly building the infrastructure that turns AI capabilities into actual money. That’s exactly what Bain said the industry needs.
Note: Commentary sections are editorial interpretation, not factual claims