Breaktime Tech Talks

This week, hear a chat with Michael Notte about systems thinking, AI, and the human edge in insurance. Lessons in other industries can translate to software development. Designing benefits and compensation packages using systems-thinking principles Treating upfront communication like requirements gathering AI chatbots for FAQs and enrollment: what works and where they hit a wall The human edge: reading between the lines when an AI bot can't Balancing AI tools and people when scaling Governance and accountability: who owns the decisions an AI agent makes? AI as reactive, not proactive and why humans still need to surface the indicators Technical debt in insurance vs. software: shortcuts that come back to bite you Links and Resources: LinkedIn: Mike Notte Website: MJN Insurance Services Instagram: @smallbizbenefitspro

In this episode, Jennifer recaps her Northeast US Java User Group tour and dives into the week's biggest tech release of Jev (a new System 1 model) along with Neo4j NODES 2026, upcoming projects, and two must-read blog posts. Java User Group tour: shout-out to Northeast US JUG tour groups! TriJUG (Raleigh), Charlotte JUG, and a combined Richmond JUG / GraphRVA provided experiences that smashed it out of the park. NODES 2026 agenda now live: Neo4j's annual online developer conference returns November 12th, featuring all in-depth technical sessions (no lightning talks this year). Pre-event "Road to NODES" workshops begin around October 1st (seven sessions available). Java book milestone: All chapters submitted; editing and publishing prep remain. Jev — a new System 1 Model: Released early this week. Rather than outputting text, Jev focuses on decision probabilities and traditional ML calculations. Potential use cases include database query planning, deterministic agent tool selection, and much more! Jev operations use three modes — choice (pick best from a list), noul (score 0–1 for truthfulness), and score (numeric scoring). Spring AI support: Spring AI already has integration for Jev; blog post linked in notes. Upcoming explorations: Jev, RAG pipeline benchmarking, multi-agent applications, MCP security & access restrictions, and guardrails for agentic systems. Blog post — The startup’s guide to getting started building on Neo4j Aura: Pradeep Ponduri's walkthrough on provisioning Aura instances via CLI, plus training and GraphRAG library resources for startups. Blog post — Navigating a Neo4j Knowledge Graph with Jev: Michael Hunger's early exploration of Jev with Neo4j, including demo code and pros/cons. Thanks for listening, and happy coding!

Jennifer Reif talks with Dr. Robert Kerstein, a prosthodontist with 42 years of experience researching the T-Scan bite computer. T-Scan uses pressure-mapping sensors to measure bite force, timing, and friction. It's the same Tekscan sensor tech used in aerospace, automotive, and robotics. Software records bite dynamics like a movie, enabling data-driven analysis instead of subjective guesswork with traditional paper and ink. Recurring tracking lets clinicians compare recordings over time to catch changes early. Effective use requires hands-on, chair-side training, just like developers need real keyboard time with a new tool. Emerging research shows the bite communicates directly with the brain and cranial nerves, linking dental mechanics to neurology. The T-Scan incorporates AI to flag problem areas in bite data, but medical device regulations prohibit it from making diagnoses. This is a key constraint for developers building AI into medical tools. Resources: Website: Dr. Robert Kerstein Tekscan: Digital Occlusal Analysis Resources: DTR Providers | Digital Occlusion Seminars Thanks for listening and happy coding!

Jennifer shares updates on her upcoming AI First Java book, fall event prep, and spring 2027 conference abstracts. Key topics this episode: GraphAcademy makeover. Neo4j's learning platform now features browser-based code environments, auto-generated API keys, AI-assisted data modeling, and public datasets for self-paced learning. GraphRAG workshop. Jennifer recently delivered a virtual workshop on generative AI and GraphRAG using the updated GraphAcademy platform. Graph Summit NYC. Neo4j's flagship event featuring a brand-new hands-on workshop format where attendees build a working project. Upcoming events. Northeast JUG tour in late September (Raleigh, NC; Charlotte, NC; Richmond, VA) and a RAG benchmarking webinar in the works. "Building Software is Learning". Thorsten Ball's article on shortening the feedback loop between building and validating requirements, and why iterative development matters even more in the age of AI coding tools. Real-world example. Jennifer shares how writing her book's capstone project with Claude reinforced the need for upfront context, boundaries, and tight iteration when working with AI. Thanks for listening, and happy coding!

This week, I took a detour with a different kind of productivity and share highlights from two must-read articles on AI-era software development. My "brain sabbatical": a strategy day for deep focus, ideation, and conference abstract brainstorming. Upcoming events: Intro to AI Workshop (Sep 3) GraphSummit NYC (Sep 17) Northeast US meetup tour (Sep 21–23, Raleigh, Charlotte, Richmond) "Domain Knowledge Is the Leverage" by Beshr Kayali Reinholdsson: practices like TDD and DDD are more relevant than ever in the AI coding era Spec-driven development: specs as living decision documents, not page-count exercises Tests matter more than ever. AI-generated code still needs independent verification. "Coding Is Solved, Software Is Not" by Gao: writing code is no longer the bottleneck; context, specs, verification, and human checkpoints are Confirmation bias risk: AI-generated test suites may just validate the implementation the agent already chose Key takeaway: developers are spending more time on design decisions, domain clarity, and proving the right thing was built Thanks for listening, and happy coding!