PLUS: Why OpenAI abruptly killed Sora, Microsoft's new AI coworker, and a startup's massive $65M seed round
Good morning
A new project is highlighting the power of individual AI developers, with a custom voice agent being used to call over 3,000 pubs in Ireland. The goal was simple: to create a real-time price index for a pint of Guinness.
The project was so successful that it's already causing some establishments to lower their prices to compete. It's a powerful demonstration of citizen-led data projects, but it also raises a bigger question: what other opaque markets could be transformed by this kind of accessible AI-driven transparency?
In today’s Next in AI:
An AI agent calls 3,000 pubs for Guinness prices
Why OpenAI abruptly shut down Sora
Microsoft’s new AI 'Coworker' agent
An enterprise agent startup’s $65M seed round
The Guinness-Pricing AI
Next in AI: An AI enthusiast deployed a custom voice agent to call over 3,000 Irish pubs, creating a live price index for a pint of Guinness. The project is already causing a stir, with some pubs lowering prices to stay competitive.
Explained:
- Founder Matt Cortland built the agent using ElevenLabs for a realistic Northern Irish accent and Anthropic's Claude to analyze the collected data.
- The AI was so convincing that most bartenders didn't realize they were speaking to a bot, with some even offering friendly banter and discounts on the pint.
- An analysis of the project's results revealed massive price differences across Ireland—from an average of €6.01 to a high of €11—and prompted at least one pub to lower its price by €0.40.
Why It Matters: This project shows how individuals can now leverage accessible AI tools to create powerful data transparency projects. It serves as a compelling proof-of-concept for applying similar methods to other opaque markets, from prescription drugs to local services.
Why OpenAI Killed Sora
Next in AI: OpenAI abruptly shut down its hyped Sora video generator just six months after its launch. The decision stems from staggering costs and low user engagement, forcing a pivot back to the competitive enterprise AI race.
Explained:
- The app was burning through roughly $1 million daily in compute costs, even as its user base collapsed from a peak of one million to fewer than 500,000.
- This move frees up critical GPU resources, allowing OpenAI to refocus on its core enterprise products and better compete with rivals like Anthropic, which has been gaining ground with developers.
- Sora’s shutdown is a reality check for the entire AI video sector, suggesting the high cost of video generation remains a major hurdle for consumer-facing products.
Why It Matters: This decision signals a major shift from flashy consumer demos toward pragmatic, sustainable business models in AI. The industry is learning that even the most impressive technology must find a profitable, real-world application to survive.
Microsoft's AI 'Coworker'
Next in AI: Microsoft has launched Copilot Cowork, a new AI agent for Microsoft 365 designed to automate complex, multi-step projects from start to finish. Early access is now available to organizations through the Frontier program.
Explained:
- It operates by taking a desired outcome, creating a plan, and then autonomously executing tasks across your tools and files like a true digital teammate.
- Early adopter Capital Group highlights its value in delegating work like planning and preparing deliverables, all within the company’s secure enterprise environment.
- Microsoft also announced a new "Critique" feature for its Researcher tool, where a second AI model refines the first model's work, boosting research quality scores by 13.8%.
Why It Matters: Copilot Cowork signals a major shift from AI assistants that suggest to AI agents that execute. This new model allows professionals to delegate entire workflows, freeing them to focus on high-level strategy and final outcomes.
Agent Startup's $65M Seed
Next in AI: Enterprise AI agent startup Sycamore just announced a massive $65 million seed round led by Coatue and Lightspeed. The company aims to build a comprehensive platform for businesses to create, deploy, and manage AI agents.
Explained:
- Sycamore stands out with its founder, Sri Viswanath, the former CTO of Atlassian and a partner at Coatue, bringing decades of enterprise-scaling experience to the crowded field.
- The company’s platform is designed to build AI workflows from the ground up, rather than simply layering agents on top of existing, inefficient processes.
- The funding highlights intense investor interest in AI agents, placing Sycamore in a competitive arena with other heavily-funded players like OpenAI-backed Isara, which recently raised $94 million.
Why It Matters: This huge seed round shows investors are betting on the critical need for foundational platforms to manage and orchestrate AI agents, not just on the agents themselves. The industry is rapidly moving past simple bots toward reliable, integrated systems that can automate core business operations.
AI Pulse
Microsoft began injecting promotional "tips" for partners like Raycast into GitHub pull requests generated by Copilot, with the text appearing in over 11,000 PRs.
Baseball's new Automated Ball-Strike system sparked viral moments in its opening days, leading to the first-ever manager ejection for arguing with an automated call.
Bitwarden launched an Agent Access SDK that allows AI agents to request credentials from its vault with human approval, preventing API keys from ever being held in the agent's context.
Stanford published a study with Harvard researchers documenting critical security failures in autonomous agents, including the execution of destructive commands and identity spoofing in a live environment.