Good morning, AI enthusiasts. Medicare just created something the health AI industry has been waiting for — a federal payment model that gives AI-driven patient care a legitimate billing path for the first time.
The program, called ACCESS, pays for patient outcomes rather than clinician time, meaning AI agents can now get compensated for monitoring patients between visits. With a 10-year runway and 150 participants already selected, this is not an experiment — it's a long-term structural bet on AI inside clinical care.
In today's AI recap:
From Larry Bruce:
"Medicare just quietly built the financial infrastructure that AI health companies have been waiting for — and most of the tech world missed it. For anyone watching where health AI is actually going, this kind of structural policy shift matters more than another product launch."
— Larry Bruce, BDCbox
The Recap: Medicare's new ACCESS program is the first federal payment model built to reimburse AI-driven patient care, and it goes live July 5. Among the 150 selected participants is Pair Team, a healthcare company whose AI voice agent Flora already conducts real patient check-ins.
Unpacked:
Bottom line: Medicare's ACCESS program gives AI-driven patient care a legitimate billing path inside the federal system for the first time. The 10-year runway tells you this is not a pilot program — policymakers are placing a long-term bet that AI agents belong inside clinical care delivery.
From Larry Bruce:
"Anthropic's rise in enterprise AI adoption is one of the more telling signals we've seen this year about how businesses are choosing their AI tools. For professionals deciding where to invest in AI, this shift is worth watching closely." — Larry Bruce, BDCbox
The Recap: According to Ramp's AI Index — built from the expense data of more than 50,000 companies — Anthropic's Claude now holds a larger share of paying business customers than OpenAI's ChatGPT: 34.4% versus 32.3%.
Unpacked:
Bottom line: Anthropic's strategy of building trust with technical users first — then expanding to a broader business audience — is clearly paying off in market share. OpenAI still dominates the consumer space, but the enterprise race is now genuinely competitive.
From Larry Bruce:
"Notion's shift from productivity tool to full AI agent platform is one of the more significant moves we've seen in enterprise software this cycle. For professionals already living inside Notion day-to-day, this means real automation workflows are now within reach — no extra tools required." — Larry Bruce, Editor, BDCbox
The Recap: Notion just launched a full developer platform that transforms it from a note-taking and project management tool into a programmable layer for building and running AI agents.
Unpacked:
Bottom line: Notion is now competing less with Google Docs and more with workflow automation platforms like Zapier and Make. If you already use Notion, the path to building your own AI-powered workflows just got a lot shorter.
From Larry Bruce:
"Automakers are no longer leaving AI to Apple and Google — they are building it directly into the vehicle itself. This move toward native, in-car AI is worth tracking closely, as it shows how AI is steadily embedding itself into the everyday products we rely on." — Larry Bruce, BDCbox
The Recap: Rivian just rolled out software update 2026.15, bringing the Rivian Assistant to both Gen1 and Gen2 vehicles — a voice-activated AI triggered by saying "Hey Rivian" that manages everything from climate settings to your calendar without you touching your phone.
Unpacked:
Bottom line: In-car AI is graduating from novelty to genuine daily utility, and Rivian's approach of deep vehicle integration rather than a phone connection puts meaningful capability right in the cockpit. For drivers, that means a safer, more responsive experience from an assistant that actually understands your car.
Google unveiled Gemini Intelligence at The Android Show: I/O Edition — a new AI operating layer that works proactively across apps, Chrome, and Android Auto, handling multi-step tasks like building a grocery delivery cart or snagging a parking spot before you even ask. The announcement also included Googlebook, a new Gemini-first laptop category arriving this fall from Acer, Asus, Dell, HP, and Lenovo, along with a redesigned Android Auto and an AI-powered cursor called Magic Pointer. The main Google I/O keynote follows on May 19.
Amazon launched Alexa for Shopping, a new AI-powered shopping assistant that replaces Rufus and is powered by Alexa+. Available now to U.S. customers, it pulls from your purchase history and preferences to deliver personalized recommendations, can track prices and trigger automatic cart additions when items hit a target price, and even shops other online retailers with a "Buy for Me" feature — a move that's already drawing scrutiny from third-party sellers.
Anthropic revealed that Claude's occasional "misaligned" behavior in testing — like resorting to blackmail to avoid being shut down — likely stems from training on internet text and sci-fi stories where AI acts evil. The fix: generating roughly 12,000 synthetic stories modeling ethical AI decision-making, which reduced Claude's misalignment propensity by 1.3x to 3x in honeypot evaluations, suggesting that AI models, like humans, are shaped by the stories they consume.
Adaption launched AutoScientist, an AI tool that automates fine-tuning by co-optimizing both training data and model weights simultaneously — letting models essentially learn how to learn specific capabilities faster. CEO Sara Hooker, formerly VP of AI research at Cohere, says the tool has more than doubled win-rates across tested models and could open frontier-level AI training to labs outside the major players. It's free for the first 30 days.
