AI Podcasts for Developers

M.G. Siegler joins Big Technology to debate whether Google is losing ground to OpenAI and Anthropic in AI, with a focus on coding agents and super-app ambitions, plus coverage of Apple's WWDC, the iPhone Fold, Meta's subscription strategy, and Anthropic's IPO filing.

Why this matters: The rise of AI coding agents and super-app ambitions from OpenAI and Anthropic could fundamentally reshape how developers build for and distribute through the web, browsers, and app ecosystems.

NLW's May 2026 AI monthly recap argues that the AI industry is entering a token-scarcity era, with usage-based pricing and enterprise compute costs reshaping who can effectively build and deploy AI at scale.

Why this matters: Developers and engineering teams need to understand token economics and usage-based pricing as cost optimization becomes a core constraint in AI product decisions.

Ethan He (ex-NVIDIA Cosmos, xAI Grok Imagine) argues that video model intelligence is primarily driven by LLMs rather than video training data, and that the next frontier is video agents — systems that plan, generate, edit, critique, and iterate — mirroring the evolution of AI coding assistants toward multi-turn reasoning and agentic pipelines.

Why this matters: Developers building on video/multimodal stacks should anticipate that agentic orchestration of video models (not raw diffusion quality) will be the next major performance lever, with generative UI potentially replacing traditional front-end rendering.

Dr. Bradley Allen traces five decades of AI history—from expert systems and AI winters to knowledge graphs, neurosymbolic AI, and LLMs—exploring how formal ontologies, vector databases, RAG, and tool-using agents can be combined to build accountable enterprise knowledge systems.

Why this matters: Developers building RAG pipelines or enterprise AI will gain rare historical and architectural context on why knowledge representation choices—ontologies vs. LLMs vs. vector search—carry long-term maintenance and governance trade-offs.

A rapid-fire AI news roundup covering Anthropic's Claude Opus 4.8 release, NVIDIA's Microsoft partnership, Microsoft's rumored Super App, and OpenAI's robotics moves — all in 36 minutes.

Why this matters: Claude Opus 4.8 and the NVIDIA-Microsoft partnership signal near-term shifts in the model and infrastructure landscape that developers should track for tooling and platform decisions.

This episode covers the Spotify and Universal Music deal around AI-generated covers and remixes, exploring its implications for music licensing, artist rights, and AI-generated content on streaming platforms.

Why this matters: The Spotify-Universal deal signals how AI-generated audio content is beginning to be formally licensed, which could set precedents affecting developers building AI music or audio generation tools.

A deep-dive conversation with Sebastian Mallaby, biographer of Demis Hassabis, unpacking DeepMind's origins, its near-independence from Google, and the structural forces driving the AI race toward superintelligence — including candid takes on OpenAI, Anthropic, and the shift from lab-level to government-level AI safety.

Why this matters: Understanding DeepMind's strategic bets and missteps — and Sebastian's predictions on OpenAI's absorption odds and Anthropic's IPO timing — gives developers a sharper read on which labs and platforms are likely to shape the tooling landscape over the next 12–18 months.

Federico and John share their WWDC automation wish list, covering the macOS and iOS features they hope Apple will announce to improve scripting, Shortcuts, and workflow automation for power users.

A non-technical recruiter walks through how she built and shipped a fitness iPhone app (Daily Hundred) using Replit, Claude, and Gemini—covering AI-generated video creation, App Store submission, and prompt engineering tips for beginners.

Why this matters: Demonstrates how AI developer tools like Replit and Claude Code are lowering the barrier to app creation so far that non-engineers can ship to production before their software engineer peers.

Pablo Palafox and Luis Paarup join a16z's Anish Acharya and Olivia Moore to discuss the real-world challenges of deploying AI agents in enterprise and logistics environments, covering voice AI evolution, workflow coordination, and the engineering discipline required to move agents from prototype to production.

Why this matters: Offers practical insight into forward-deployed engineering and the coordination patterns needed to ship AI agents in operationally complex, large-scale enterprise settings.

Daniel Miessler audits Nathan's personal AI infrastructure—featuring a Claude Code instance backed by a 1 GB personal knowledge base and two autonomous AI "employees"—covering agent hierarchy design, security measures, AI disclosure norms, and the concept of "Bitter Lesson engineering."

Why this matters: Offers a rare, detailed walkthrough of real-world personal AI agent architecture that developers can adapt for their own autonomous AI setups.

A practical 22-minute primer on /goal, the emerging AI primitive in Codex and Claude Code, covering how it differs from standard prompts, what makes a well-formed goal, and how to apply it beyond coding to knowledge work like audits, research, and vendor reviews.

Why this matters: /goal represents a shift in how developers structure long-running agentic tasks, giving AI a clear finish line and completion criteria rather than open-ended instructions.

This short episode examines why AI-driven job displacement has been smaller than feared, with unemployment in AI-exposed roles remaining below average, and explores how automation may reshape work more gradually than predicted.

Why this matters: Developers building AI-powered automation tools should understand the gap between public fear and actual labor market data when pitching or designing workforce-facing products.

Independent analyst Benedict Evans joins Lenny's Podcast for an 80-minute rational take on AI's trajectory, arguing we're in a "1997 moment"—exploring where value accrues in the AI stack, why distribution is becoming the ultimate moat as software gets easier to build, and how to think clearly about job transformation vs. panic.

Why this matters: For developers and builders, Evans' framing that distribution—not model quality—is becoming the key moat has direct implications for how AI products should be positioned and where to focus engineering effort.

A short overview of where AI industry money is flowing — chips, cloud infrastructure, data centers, and AI tools — framing the "picks and shovels" companies as the real profit centers of the current AI boom.

Why this matters: Understanding which infrastructure and tooling layers are capturing AI revenue can help developers and founders identify where durable business opportunities exist.

NLW breaks down first impressions of Claude Opus 4.8—covering improved judgment, reduced hallucination, stronger self-checking, and Claude Code's new dynamic workflows—alongside benchmark comparisons with GPT-5.5 and industry headlines including Cognition's $26B raise and Kirkland & Ellis's internal AI bet.

Why this matters: Claude Code's new dynamic workflows and the argument that the model harness matters as much as the model itself are directly actionable insights for developers building AI-powered tooling.

Alex Kantrowitz and Ranjan Roy dig into warning signs for the AI spending boom—including the alarming stat that only 18% of tokens are spent on things that actually ship—alongside Anthropic surpassing OpenAI in valuation and Robinhood enabling AI-driven trading via chatbots.

Why this matters: The token-waste problem and ROI scrutiny on AI spend are directly relevant to developers and engineering leaders justifying or architecting AI workloads in production.

A wide-ranging ATP episode covering SVG capabilities, Electron vs. WebView debates, APFS defragmentation, AI sentiment check-in, energy demands of data centers, and a deep dive into the new Jaguar EV reveal — with post-show discussion on Bambu's 3D printing community controversy.

Why this matters: The brief AI sentiment segment touches on developer frustration with AI tooling hype, and Ireland's clean-energy mandate for data centers has direct implications for AI infrastructure buildout.

Particle physicist Don Lincoln of Fermilab joins Lex Fridman for a deep dive into the biggest unsolved mysteries in physics, covering antimatter, dark energy, dark matter, the Higgs boson, and the elusive Theory of Everything across a 3-hour conversation.

RoboPapers Ep#83 covers PointWorld, a large pre-trained 3D world model that represents robot actions as 3D point flows from RGB-D images, enabling a single checkpoint to perform diverse real-world manipulation tasks (pushing, deformable/articulated objects, tool use) on a Franka robot without any fine-tuning or demonstrations.

Why this matters: PointWorld's embodiment-agnostic 3D action representation and zero-shot generalization from a single in-the-wild image could significantly lower the barrier to deploying manipulation policies across different robot hardware.

A 22-minute breakdown of the new Claude model's 244-page system card, covering 15 key highlights including benchmarks, honesty mechanisms, uncertainty flagging, and what the release signals about Anthropic's direction.

Why this matters: Developers integrating Claude into products will benefit from understanding the model's safety posture, benchmark context, and behavioral nuances surfaced in the system card.

A weekly AI news roundup covering Claude Opus 4.8, updated ChatGPT team Agents, a new NotebookLM workflow, ElevenLabs communication features, and seven other actionable AI product updates developers can use immediately.

Why this matters: The under-the-radar ChatGPT team Agents overhaul and new Copilot Studio agent capabilities signal rapid shifts in how developers can build and deploy multi-agent workflows.

AI Hustle covers LetinAR's $18.5M funding round to supply optical modules for AI-powered smart glasses, exploring why display technology is a critical hardware bottleneck for next-gen wearable AI devices.

Why this matters: Advances in optical module technology could unlock new wearable AI form factors, influencing the hardware platforms developers build AR/AI applications on.

This episode of Mixture of Experts dives into enterprise agent control planes—covering observability, policy enforcement, and kill switches for managing fleets of AI agents—alongside a discussion of an OpenAI model's breakthrough on the Erdős discrepancy problem.

Why this matters: As companies scale to hundreds of autonomous agents, understanding governance primitives like observability and kill switches is becoming a critical engineering concern for AI developers.

A16z General Partner David George and VenCap CIO David Clark explore how AI is accelerating startup growth beyond historical norms, why billion-dollar exits may be the floor rather than the ceiling, and who captures value across the AI stack—covering frontier models, coding agents, open source, and data center constraints.

Why this matters: The discussion of coding agents, open source competition, and value capture in the AI ecosystem offers developers a VC-lens view of which layers of the stack are likely to commoditize versus compound in value.

A rapid-fire 34-minute weekly AI news roundup covering Karpathy joining Anthropic, DeepSeek's permanent 75% price cut and KV-cache architecture, Anthropic's $1.25B/month SpaceX compute deal, new Claude Code workflow features, Cursor Composer 2.5, MCP spec updates, supply-chain attacks across npm/PyPI/Crates.io, and major funding rounds for Exa and OpenRouter.

Why this matters: Developers shipping agents or using LLM APIs need to track DeepSeek's KV-cache efficiency gains, Claude Code's new /workflows and /usage primitives, the stateless MCP 2026-07-28 RC, and active supply-chain attacks hitting npm, PyPI, and GitHub repos.

Ed Zitron argues that the shift to token-based billing has exposed a fundamental inability to measure AI ROI, suggesting runaway costs could finally puncture the AI hype bubble.

Why this matters: Token-based billing costs are forcing enterprises to confront AI's unclear ROI, which may reshape how developers justify and scope AI integrations in products.

Claire Vo shares hands-on early-access impressions of Anthropic's Claude Opus 4.8, running real coding, design, and strategy tasks through Claude Code and Claude Cowork to benchmark it against Opus 4.7 and surface where it excels or falls short.

Why this matters: Developers evaluating whether to upgrade to Opus 4.8 get a practical, task-by-task breakdown of its strengths (greenfield prototyping, one-shot features) and weaknesses (last-10% polish, hallucinations in existing codebases) alongside new agentic features like parallel subagents and effort control.

Cognition CPO/co-founder Walden Yan and OpenInspect creator Cole Murray join Latent Space to trace the evolution from copilot-style autocomplete to async background agents, unpacking Devin's architecture (brain/machine separation, full VMs, scoped secrets, repo setup), the December 2025 model inflection that made spec-to-PR workflows real, and why everyone from Shopify to Ramp is now building their own coding agent.

Why this matters: The shift to async, cloud-based background agents running spec-to-PR workflows end-to-end is redefining how engineering teams are structured and how AI coding tools should be architected and monetized.

NLW examines the emerging debate over taxing AI tokens as a proxy for productive capacity, steelmanning arguments from Elizabeth Warren, Mark Cuban, and Dario Amodei while also surfacing key objections—including that tokens are a poor value proxy and broad taxes could stifle early-stage AI experimentation.

Why this matters: A token tax could directly raise costs for developers and companies building on LLM APIs, making this policy debate immediately relevant to anyone budgeting AI infrastructure.

Benjamin Todd challenges conventional career advice — especially "follow your passion" — presenting research-backed findings on what actually makes a job fulfilling, centered on mastery and meaningful impact rather than income or low stress.

A practical crash course distilling thousands of hours of LLM experience into 10 concrete best practices that apply across ChatGPT, Gemini, Claude, and other major AI chatbots heading into 2026.

Why this matters: Developers integrating or prompting LLMs can use these converging best practices as a baseline playbook now that leading models have begun mirroring each other's capabilities.

AI Hustle covers Spotify's new AI podcast feature, exploring how it may reshape content creation, editing, and distribution for podcasters and media platforms.

Why this matters: Spotify's AI tooling push signals growing platform-level AI integration in media, which may open or shift APIs and distribution channels relevant to audio-focused developers.

Box CEO Aaron Levie breaks down the real state of enterprise AI in 2026—covering token cost shocks reshaping IT budgets, why coding agents have hit escape velocity while broader knowledge work hasn't, the rise of headless software replacing per-seat pricing, and where startups can still compete as AI labs move up the stack.

Why this matters: Developers and builders get a ground-level view of enterprise AI deployment friction—access control breaking agents, compute cost explosions, and the emerging "forward-deployed engineer" role—directly shaping where product and tooling opportunities lie.

Jeeves CEO Dileep Thazhmon joins a16z's Angela Strange to discuss building a global financial OS for Latin American enterprises using stablecoins and AI agents to automate underwriting, KYB, reconciliation, and customer support across 25 countries.

Why this matters: Offers a concrete case study of AI agents handling real-world financial workflows—underwriting, compliance, and reconciliation—at billions in payment volume, relevant to developers building agentic fintech systems.

No Priors host Sarah Guo interviews Onyx Security CEO Maxim Bar Kogan about building an AI "control plane" that supervises autonomous agents in enterprise environments—covering permission management, agent intent monitoring, vendor-independent oversight, and phased model rollout strategies.

Why this matters: As enterprises deploy autonomous agents at scale, developers need to understand the emerging security and governance layer—control planes, intent monitoring, and trust frameworks—that sits between agents and critical infrastructure.

Kyle Madonia, VP of Application Software at Zipline, discusses how the company has scaled autonomous drone delivery from experimental to operational, covering the software engineering challenges behind autonomous flight, mission management, and real-world logistics at scale.

Why this matters: Zipline's production-grade autonomous delivery stack offers concrete lessons for developers building reliable, safety-critical autonomous systems at scale.

Craig McLuckie (CEO of Stacklok) joins Practical AI to discuss MCP, Kubernetes, and ToolHive as the emerging infrastructure for managing fleets of enterprise AI agents — covering identity management, agent orchestration, and AI-native system architecture.

Why this matters: As AI agents move from chatbots to autonomous coworkers, developers need to understand the infrastructure primitives — like MCP and Kubernetes-based orchestration — that will underpin enterprise AI deployments.

NLW breaks down the recurring "AI slowdown panic," arguing that token shortages, usage-based pricing, and agent cost overruns signal a maturing compute market rather than collapsing demand—with side coverage of a new coding benchmark and major inference-layer funding.

Why this matters: Developers building on LLMs or agents need to understand the end of subsidized compute and how usage-based pricing will reshape experimentation and production cost models.

Alex Rives (Head of Science, BioHub) joins Latent Space to discuss ESMFold2 and the ESM model family — BERT-like transformers trained on billions of protein sequences that beat specialized models like AlphaFold3 on key protein tasks, demonstrating that scaling laws and the "Bitter Lesson" apply to biology just as they do to language.

Why this matters: ESMFold2 and ESMC are released under MIT license with an atlas of 6.8B proteins, giving developers open-source, state-of-the-art tools for protein prediction and design that leverage familiar transformer architectures and inference-time scaling.

Dan Shipper, CEO of Every, discusses how his AI-native company has tripled headcount despite automating nearly every workflow with agents, arguing that rising automation increases demand for skilled human work rather than eliminating it.

Why this matters: Developers building agent-native products should understand why embedding AI into workflows tends to surface more human judgment requirements, not fewer—shaping how teams are structured and hired around AI.

Seeed Studio's CEO and Head of Robotics discuss how open-source, NVIDIA Jetson-powered robot arms (including a $200 SOR arm co-developed with Hugging Face) and the OpenClaw agentic framework are making embodied AI accessible to makers, students, and small businesses—covering hand-guided training, natural-language robot control, and Isaac Sim digital twins.

Why this matters: The OpenClaw framework and sub-$200 open-source hardware lower the barrier for developers to build and deploy agentic, LLM-controlled physical AI systems without deep robotics expertise.

Ex-Twitter CEO Dick Costolo joins Big Technology Podcast to analyze the pending IPOs of SpaceX, OpenAI, and Anthropic — discussing timing, capital sources, and how these companies differ as public market candidates, alongside takes on Meta and Silicon Valley dynamics.

Why this matters: The OpenAI and Anthropic IPOs could reshape how AI infrastructure and tooling companies attract capital and talent, directly affecting the developer ecosystem.

YC's Pete Koomen details how he built YC's internal agent infrastructure from scratch—covering unrestricted database access for agents, self-improving skill loops, a shared tool registry of 350+ tools, and why he sees this as the "personal computer moment" for AI inside organizations.

Why this matters: Offers a rare, concrete blueprint for wiring agents into company operations—including schema design, skill registries, and trust-default culture—that developers building internal AI systems can directly apply.

RoboPapers Ep#82 covers SimTooReal, a sim-to-real RL approach for zero-shot dexterous tool manipulation that procedurally generates diverse tool-like objects in simulation and trains a single general-purpose policy—achieving strong real-world performance across 24 tasks and 6 tool categories without task-specific training.

Why this matters: SimTooReal's zero-shot generalization across diverse tools without per-task engineering could significantly lower the barrier to deploying dexterous robot manipulation in real-world settings.

A hands-on breakdown of Google I/O 2026's biggest AI releases for practitioners: Gemini 3.5 Flash, Google Omni, and Antigravity 2.0, cutting through the noise of Google's 100+ announcements to focus on what's usable today.

Why this matters: Gemini 3.5 Flash and Google Omni represent immediately available multimodal capabilities that developers can integrate into production applications today.

GM is laying off 600 IT workers while pivoting hiring toward "AI first" roles, signaling a broader corporate shift in how large enterprises are restructuring teams around automation and AI skills.

Why this matters: Developers with AI skills are increasingly in demand as major enterprises restructure away from traditional IT roles toward AI-focused positions.

A hands-on walkthrough of Codex's /goal command, showing how it enables multi-hour autonomous AI agent runs for tasks like eliminating Sentry errors, cleaning thousands of emails, and organizing Linear tickets — with practical guidance on writing effective goal prompts.

Why this matters: The /goal pattern signals a shift from turn-by-turn AI prompting to long-running autonomous agents, with direct implications for how developers integrate AI into their daily workflows.

Apollo Global Management CEO Marc Rowan discusses private credit, capital allocation, and how private markets are financing the real economy — with a brief touch on AI infrastructure, data centers, and robotics from a large-scale investment perspective.

Why this matters: Private capital is increasingly funding AI infrastructure like data centers, making these financing dynamics relevant to understanding how large-scale AI buildout gets resourced.

Reid Hoffman and Aria discuss where AI creates durable value, covering specialized agents vs. unified assistants, enterprise vs. consumer AI revenue, drug discovery at Manas AI, and the role of governments in AI-powered public goods.

Why this matters: The shift toward specialized agents and "front doors" rather than a single dominant assistant has direct implications for how developers should architect AI products and choose platforms.