Get tomorrow's brief in your inbox
Today: Zapier raised its AI hiring bar, requiring every candidate to show repeatable AI systems embedded in core work. Baidu robotaxis stalled mid-traffic in Wuhan, trapping passengers and causing crashes. OpenAI raised $122 billion to expand frontier AI globally.
Zapier just published the playbook every business needs for hiring AI-ready talent, and it's a wake-up call for companies still treating AI as optional. One year after requiring AI fluency from every new hire, the company has hit 100% AI adoption across all teams and is now raising the bar even higher. Their V2 hiring rubric doesn't just ask if candidates have used ChatGPT. It requires proof of repeatable AI systems embedded into core work, measurable impact on quality or efficiency, and a visible learning trajectory. Someone who plateaued eight months ago using the same three tools is a different hire than someone actively experimenting.
The message is clear: if you're hiring people who aren't meaningfully improving their work with AI, you're hiring people who will be outpaced within months. Zapier's approach is built on four components: AI mindset, strategy, building, and accountability. That last one matters most as AI gets more capable. As they put it, "With AI, you can delegate the work, but not the accountability." Managers must show they've led teams to adopt AI, not just used it personally. This is how you build a workforce that stays competitive.
Zoho Solo launched for solopreneurs. If you're running a business alone and juggling 10+ tools (invoicing, payments, bookkeeping, CRM, project tracking), Zoho Solo consolidates everything into one mobile app. Free tier available, paid version is $96/year. It's mobile-only (no desktop version), so it works best if you're comfortable managing your business from your phone. The value proposition is simple: unified data means you can actually see which clients pay the most and which tasks take the longest, instead of copying data between Google Sheets and five other apps. Worth trying if tool sprawl is killing your productivity. Learn more
ChatGPT now works through Apple CarPlay. If you have iOS 26.4 or newer and the latest ChatGPT app, you can use voice-based ChatGPT through your car dashboard. No text display (per Apple's guidelines), just voice conversations. You can't use a wake word; you have to tap the app. This is useful if you're driving and need to draft emails, brainstorm ideas, or get quick answers without pulling over. Not a business game-changer, but handy for solo consultants and service providers who spend hours in the car. Details
Gradient Labs brings AI account managers to banks. Using GPT-4.1 and GPT-5.4 models, Gradient Labs automates banking support workflows with low latency and high reliability. Every bank customer gets an AI account manager. If you run a financial services business or a fintech startup, this signals where customer service automation is heading: personalized, always-on support that doesn't require hiring a contact center. OpenAI case study
Google launched Veo 3.1 Lite. This is Google's most cost-effective AI video generation model. If you're creating marketing videos, social media content, or product demos, Veo 3.1 Lite makes video generation cheaper and faster. No pricing details in the announcement, but the "Lite" branding suggests this is aimed at high-volume use cases where cost matters more than cutting-edge quality. Google announcement
Runway launched a $10M fund and free API credits for startups. Runway, the AI video generation company, is investing in early-stage startups building on top of their "video intelligence" platform. They're writing checks up to $500,000 for pre-seed and seed-stage companies, plus offering free API credits to startups from seed to Series C. If you're building anything involving AI video (marketing automation, training content, product demos), this is a way to get subsidized access to one of the best video generation tools. Their thesis covers three buckets: technical teams pushing AI frontiers, application-layer builders bringing AI to new use cases, and companies experimenting with new media creation. Apply here
AI benchmarks are broken. MIT Technology Review published a deep dive on why AI benchmark scores don't predict real-world performance. The problem: AI is evaluated on isolated tasks in a vacuum, but it's used in messy, complex environments with human teams and organizational workflows. Hospitals bought FDA-approved radiology AI that scored 98% accuracy in tests, then discovered it added delays in practice because it didn't fit hospital reporting standards or multidisciplinary decision-making processes. The lesson for SMBs: don't buy AI tools based solely on vendor benchmark claims. Run a pilot with real workflows, real users, and real data before committing budget. Full analysis
OpenAI raised $122 billion. This funding round will expand frontier AI globally, invest in next-generation compute, and meet growing demand for ChatGPT, Codex, and enterprise AI. For small businesses, this means OpenAI has the capital to keep improving models, expanding API access, and maintaining competitive pricing. The scale of this raise (largest in AI history) signals that OpenAI is betting big on enterprise adoption, which typically means better support, more reliable uptime, and potentially new pricing tiers for SMBs. OpenAI announcement
JPMorgan Chase committed $80 billion to reach 10 million small businesses. This is a multiyear effort to expand lending and banking services to SMBs nationwide. If you need capital for AI investments, equipment, or expansion, this signals that small business lending is opening up. Banks are competing for your business. Inc.com coverage
Baidu robotaxis stalled mid-traffic in Wuhan, causing crashes. Multiple Apollo Go robotaxis stopped in the middle of roads, trapping passengers and causing highway collisions. This is a reminder that even the most advanced AI systems (Baidu claims 300 million autonomous kilometers driven) can fail catastrophically. For businesses exploring autonomous delivery, logistics automation, or any AI system that operates in the physical world, this underscores the need for redundancy, failsafes, and incident response plans. Chinese regulators are finalizing insurance terms for driverless vehicles. If you're in logistics or delivery, watch this space: autonomous vehicles are coming, but operational risk is real. CNBC report
Microsoft faces UK competition probe over software business. The UK's Competition and Markets Authority is investigating Microsoft's enterprise software suite (Windows, Word, Excel, Teams, Copilot) for potential anti-competitive practices. If you rely heavily on Microsoft's ecosystem, this could eventually lead to forced interoperability changes, making it easier to switch to competitors or mix-and-match tools. For now, no immediate impact, but it signals that European regulators are pushing for tech stack diversification. CNBC coverage
Oracle cutting thousands of jobs to fund AI spending. Oracle is laying off 20,000 to 30,000 employees (potentially $8-10 billion in free cash flow) to fund massive AI infrastructure buildouts. The company raised $50 billion in debt and signed a $300 billion contract with OpenAI for cloud services. For SMBs, this means Oracle is betting its future on AI cloud infrastructure. If you use Oracle databases or cloud services, expect more AI features and potential price pressure as Oracle tries to compete with AWS and Azure. CNBC story
Supply chain attacks hit Axios and LiteLLM. Axios, the HTTP client with 101 million weekly npm downloads, was compromised via a leaked npm token. A malicious dependency stole credentials and installed remote access trojans. Mercor, an AI recruiting startup valued at $10 billion, confirmed a security incident linked to a LiteLLM supply chain attack. If you use open-source dependencies (and you do), adopt trusted publishing workflows and monitor for releases published without GitHub tags. This is the new normal for software supply chains. Details | Mercor incident
Audit your AI hiring process. Print Zapier's V2 AI Fluency Rubric and score your current team and next three candidates. If you don't have a formal AI fluency bar, you're hiring people who will need retraining in six months. Start with one question in every interview: "Show me a repeatable AI workflow you've built that measurably improved your work." If they can't answer, they don't meet the bar.
Run an AI benchmark reality check. Pick one AI tool you're considering (or already using). Ignore the vendor's benchmark scores. Run a one-week pilot with real users, real tasks, and real workflows. Measure time saved, errors introduced, and user frustration. If the pilot doesn't show clear value, don't buy it. Benchmarks lie; pilots don't.
Three patterns across today's stories: hiring for AI fluency is shifting from "nice to have" to "must have," real-world AI performance diverges sharply from lab benchmarks, and supply chain security is now an existential risk for any company using open-source code. This week, focus on one question: do your people have the AI skills to stay competitive, and do your tools actually work in production? Everything else is noise.