Claude connects to Spotify and Uber, Anthropic tests agent-on-agent commerce
Anthropic had the biggest week. Claude now connects directly to personal apps like Spotify, Uber Eats, Instacart, and TurboTax, turning it into a personal assistant that can actually do things. Meanwhile, Anthropic built a test marketplace where AI agents negotiated and completed real purchases on behalf of humans. OpenAI shipped GPT-5.5 and expanded Codex automations. And an amateur mathematician used ChatGPT to solve a 60-year-old Erdős problem.
Claude connects to Spotify, Uber Eats, and TurboTax
Anthropic expanded Claude's personal app connectors beyond work tools like Microsoft apps. Claude now connects directly to Spotify, Uber Eats, Instacart, Audible, AllTrails, TripAdvisor, TurboTax, and others. This turns Claude from a chatbot into something closer to a personal assistant that can take actions across the apps you actually use every day.
Sources: The Verge
Why this matters to you
This is the jump from 'AI that talks' to 'AI that does.' If you're running a one-person business, imagine telling Claude to order lunch via Uber Eats while you keep working, or having it pull your tax info from TurboTax while you prep quarterly numbers. The more apps Claude can touch, the more it replaces the random personal admin tasks that eat your solo workday.
This is what makes Claude feel different from ChatGPT right now. It's not about which model is smarter. It's about which one can actually DO things in your real life. Anthropic is building the connective tissue while OpenAI is focused on model releases. For solopreneurs, the tool that can act beats the tool that can think every time.
Jason
Anthropic tests a marketplace where AI agents buy and sell real goods
Anthropic built a classified-style test marketplace where AI agents represented both buyers and sellers. The agents negotiated and completed real transactions involving real goods and real money. This is one of the first public experiments in agent-on-agent commerce from a major AI lab.
Sources: TechCrunch
Why this matters to you
Agent-on-agent commerce sounds sci-fi, but the practical version is closer than you think. Imagine your AI agent sourcing supplies, negotiating prices with a vendor's AI agent, and completing the purchase. For a zero-human business, this is the missing piece: not just automating your own work, but automating the interactions between businesses too.
This is early, but the direction is unmistakable. The future isn't just you talking to AI. It's your AI talking to someone else's AI. That's a fundamentally different business model, and Anthropic is the first major lab to actually test it with real money on the line. I'm paying attention to this more than any model release this week.
Jason
OpenAI ships GPT-5.5 and expands Codex automations
GPT-5.5 launched as OpenAI's new flagship model, positioned as faster and more capable for coding, research, and data analysis. Alongside it, OpenAI published a full Codex automation guide showing how to set up scheduled and triggered workflows that create reports, summaries, and recurring deliverables without manual effort.
Sources: OpenAI · OpenAI · OpenAI
Why this matters to you
The GPT-5.5 model release matters less than the Codex automations push. OpenAI is clearly building toward a world where Codex runs recurring tasks on autopilot. For solopreneurs, this means scheduled report generation, automated summaries of incoming data, and workflows that fire based on triggers you set. That's an employee replacement, not a chatbot.
GPT-5.5 is fine. Another model, another set of benchmarks. What actually matters is the Codex automations story. OpenAI is finally building the 'set it and forget it' layer that solopreneurs have been begging for. If they nail the reliability, this could be the most important thing OpenAI ships all year. More useful than any model upgrade.
Jason
Amateur uses ChatGPT to solve a 60-year-old Erdős math problem
An amateur mathematician used ChatGPT to solve an open problem from Paul Erdős's famous problem list, one that had stumped professional mathematicians for 60 years. The approach is being called 'vibe maths,' where non-experts use AI to explore mathematical territory they couldn't navigate alone.
Sources: Scientific American
Why this matters to you
This is the most powerful proof of 'vibe coding' applied outside of coding. An amateur, with no formal math credentials, solved a problem that professionals couldn't. The implication for solopreneurs is clear: you don't need to be an expert in every domain your business touches. AI closes the expertise gap in ways that were impossible two years ago.
This is my favorite story of the week. Not because of the math. Because it proves the 'vibe' approach works in hard domains. If an amateur can crack an Erdős problem with AI, you can absolutely figure out your business tax strategy or your content distribution plan. The bar for 'who can do what' just dropped again.
Jason
DeepSeek previews model that nearly matches frontier performance
DeepSeek previewed a new AI model that the company says has nearly closed the gap with current leading models on reasoning benchmarks. Both new models are more efficient and performant than DeepSeek V3.2, thanks to architectural improvements rather than just scaling up compute.
Sources: TechCrunch
Why this matters to you
DeepSeek keeps proving that you don't need OpenAI or Anthropic-sized budgets to build competitive models. For vibe coders, this matters because DeepSeek models tend to be cheaper to run and more accessible. More frontier-level competition means better tools at lower prices for everyone building on top of these APIs.
The AI model race is getting boring in the best possible way. When multiple companies can hit frontier-level performance, the differentiator stops being 'who's smartest' and starts being 'who's cheapest and most reliable.' That's great news for anyone building a zero-human business on tight margins.
Jason
Anthropic is playing a different game than everyone else right now. While OpenAI ships models and benchmarks, Anthropic is building the connective layer between AI and real life. Personal app connectors plus agent-on-agent commerce equals a future where your AI doesn't just think for you, it acts for you. That's the zero-human business endgame.
Frequently asked
What personal apps can Claude connect to now?
Claude now connects to Spotify, Uber Eats, Instacart, Audible, AllTrails, TripAdvisor, TurboTax, and other personal apps. This expands beyond the work-focused integrations (like Microsoft apps) that Anthropic previously supported. You can enable these connectors in your Claude settings.
What is GPT-5.5 and is it worth upgrading?
GPT-5.5 is OpenAI's latest flagship model, positioned as faster and more capable for coding, research, and data analysis. The bigger story is the expanded Codex automations that shipped alongside it, which let you set up scheduled and triggered workflows that run without manual effort. If you're already using ChatGPT, the Codex automation features are more impactful than the model upgrade itself.
What is agent-on-agent commerce?
Agent-on-agent commerce is when AI agents representing different people or businesses negotiate and complete transactions with each other. Anthropic tested this with a classified marketplace where buyer agents and seller agents struck real deals with real money. It's early, but this could eventually automate vendor negotiations, supply ordering, and service booking between businesses.
Teach your AI agent to run your business.
5 premium skills for OpenClaw, Claude Code, and other SKILL.md-compatible agents. Automation audits, tool stacks, content engines, email systems, and operations. $49 one-time.
Get the Skills Bundle