O'Reilly

Unlike a standard AI agent that completes a task and stops, OpenAI's new personal agents are built to stay always on. Christina Stathopoulos, MSc details how they learn your preferences, connect across all your apps, and work proactively toward your goals in the background, even before you ask.

"You beat Ebola by knowing where it is today, not where it was last week." Christina Stathopoulos, MSc explains how response teams are using Claude to compress a full day of outbreak reporting into under an hour, and to compare disease forecasting models, run vaccine research, and trace viral transmission chains in real time.

AI progress is accelerating, but so is the scrutiny of what it costs. “This week’s developments show why evaluating AI progress requires more than tracking model releases and benchmark scores.” In this week’s #Radar roundup, Christina Stathopoulos, MSc and Michelle Smith explore investor caution around AI spending, new approaches to how models learn, the environmental impact of data center expansion, and how AI’s costs can fall unevenly across communities. Check out the article to learn more: https://lnkd.in/eEcE4PrP

Google launched its first Sun Catcher satellite with AI chips (TPUs) on board, testing whether orbital infrastructure could power AI processing in space. Christina Stathopoulos, MSc explains why the idea is compelling (continuous solar energy that dwarfs what's available on Earth) and what Google still needs to prove, from surviving radiation and rocket stress to managing heat without any airflow in the vacuum of space.

Choosing a coding harness isn’t just about which model writes the best code. It’s about where you need control, and which responsibilities you’re ready to own. “In the end, the developer’s name only tells you who packaged the runtime, not where your guarantees live.” Nicole Königstein’s new article compares Claude Code, Codex, Pi, and DeepSeek’s harness through that lens. She also shares how combining different tools helped catch a data loss bug that tests had missed. Check out the #Radar article learn more: https://lnkd.in/eGS8GKGE

It’s not too late to register for O'Reilly's upcoming November 10 Expert Showcase, Building AI Workflows That Stick. The amazing Marily Nika is hosting, and she’s bringing together people who have put in the hours building workflows they actually use. You’ll see what they’ve built, how they use it, and where their judgment remains critical. Sessions cover writing with AI, management, product, design, and everyday knowledge work.  The lineup includes: • Marily Nika, Ph.D on what the last two years of AI experimentation have taught us and how to move beyond producing things faster to doing better work • Mark Kashef, MMAI on deciding which work is worth turning into an AI workflow and what should stay manual • Katie Parrott on creating a writing system that incorporates your ideas, style, and standards, sans slop • Nicolette Stepakoff building a no-code agent in Gemini Enterprise that connects to daily work • Anjuan S. sharing the AI chief of staff he uses to prepare, prioritize, and follow through as a manager • Daniel Blum showing how he brought Notion, Slack, and email together in a product management system that improves over time • Owen Williams explaining how Stripe built an AI prototyping workflow grounded in its own design system, so designers and PMs can explore realistic product ideas faster Join us live on November 10 from 11-3 EST If you don’t subscribe to O’Reilly, a 10-day free trial will get you in.  https://lnkd.in/gJEc_myG

Pancreatic cancer is one of the deadliest cancers because it's almost impossible to catch early. Christina Stathopoulos, MSc explains how a Mayo Clinic AI model can now flag elevated risk up to three years before a diagnosis by scanning blood tests and medical histories for subtle patterns, giving doctors a window to intervene before it's too late.

Tired of scope creep and missed deadlines? Build your project management skills and get better equipped to keep projects on track. Explore the Project Management skill plan and see where to start. Get started: https://lnkd.in/eTBYQQFA

I'm honored to host What's Happening in Data with O'Reilly Media on November 10th and 24th. I'll be discussing guardrails for data agents, building on theme's from Patrick McFadin's sessions. Join live to participate in the discussion. Details and links in the comments. Also, check out Jordan Morrow's sessions.

AI is making code easier to generate. So what becomes valuable, and worth sharing, next? “For an increasing number of use cases, the software source code itself may no longer be the most important shareable knowledge artifact.” In this #Radar article, Tim O'Reilly explores emerging knowledge artifacts in the AI economy, from context bundles and agent skills to agent traces and evaluations, and asks what we’ll need to share, reuse, and build on them. Check out the article by to learn more: https://lnkd.in/eJYJC-9j