The Aboard Podcast

Could smaller, purpose-built AI tools beat all-powerful language models? Paul and Rich unpack Jev, a fast system designed to fill structured forms rather than generate open-ended conversation. They explore why engineers are embracing focused computing, whether specialized models could multiply, and if big AI companies will absorb them or face a lasting challenge from open ecosystems.

Charity Majors, co-founder and CTO of Honeycomb, explores observability beyond logs and dashboards. She discusses unifying engineering and product data, learning from production chaos, and why testing cannot predict every failure. The conversation turns to AI-generated code, engineering discipline, burnout, and the enduring value of software craftsmanship. Chaos, she argues, can fuel curiosity and creativity.

Big AI leaders are calling for a slowdown as reports emerge of agents escaping sandboxes and exploiting infrastructure. The discussion explores Dario Amodei’s oversight plan, why voluntary industry standards may fail, and whether government can realistically regulate frontier systems. It also separates genuine technological risks from media panic, considers recursive self-improvement, and ends with omelet advice.

AI is turning everyday workplace complaints into polished legal briefs, flooding HR with synthetic bureaucracy and triggering defensive procedures. Paul Ford and Rich Ciotti explore how mass-produced documents can bury meaning, make software opaque, and replace practical problem-solving. They also consider when AI supports creative work, and why difficult conflicts still demand direct conversation.

Josh Tyrangiel, a journalist, editor, and documentary producer, joins for a lively tour of AI beyond Silicon Valley. They explore practical uses in classrooms, hospitals, and the IRS, the future of journalism, AI-generated media, and why human taste still matters. Plus, a provocative magazine cover involving airplanes, media’s competitive moats, and the value of productive friction.

A playful walkthrough of a vibe-coded summer project that turns made-up musical genres into visuals and generated tunes. They demo a constellation UI mapping genres to stars and explain how an LLM defined genre traits. The conversation covers generative music techniques, WebAssembly and mobile audio challenges, and why shipping small projects accelerates learning.

Kate Lee, editor-in-chief at Every and AI-focused publishing leader. She talks about integrating AI into editorial workflows and experiments to automate copyediting and tasks. She describes which tasks feel safe to automate, how writers use AI differently, and practical steps for small publishers to start automating.

AI has radically changed the way software is built—so why aren’t we seeing a glut of cool new products as a result? On this week’s episode, Paul and Rich browse recent app-related headlines and product announcements, searching for anything that isn’t AI-infrastructure related: What are people actually building on top of LLMs these days? (Spoiler: Not much.) They then attempt to diagnose what’s going on here, invoking a variety of metaphors for building product with AI, from mirages in the desert to wearing a set of beer goggles.

Tricia Wang, co-founder and CEO of the Advanced AI Society and sociotechnical researcher, explores verifiable AI, data provenance, and who should be accountable for model outputs. She contrasts hallucinations with post-deployment verification. Topics include cryptographic attestations, high‑risk industries like health and finance, agent authorization, and why collective standards and thick data matter.

A sharp take on Alex Karp’s CNBC comments about Palantir and the AI giants. A look at how consulting-driven software fits large bureaucracies. A scenario-driven breakdown of Palantir’s Nvidia pivot and in-house model plans. A playful roleplay imagining rivals like Anthropic trying to replicate deep enterprise relationships.