Podcast
Agents Hour
Hosted by Mastra · EN
The AI Agents show that discusses hot topics in the world of AI, talks with guests building AI agents and applications, and shows the actual code of how AI applications are being built today. Hosted by Shane Thomas and Abhi Aiyer from Mastra.
Watch the livestream on Youtube and X on Monday at 12PM pacific time. Watch the video versions on Spotify or YouTube.
37episodes
Episodes
Newest firstAll episodes
Freestyle's CEO Runs His Cloud From a House — Ben Swerdlow on Agent VMs
5d ago00:18:56Tap to summarizeBen Swerdlow is CEO and co-founder of Freestyle, which builds Linux VMs for AI agents. He joins Abhi Aiyer to explain why he refuses to call them sandboxes. Freestyle runs on bare metal in a house. There is duplicated internet, a battery, a backup kerosene generator, a fingerprint-protected door, and a hole cut in a wall. The move followed a $45,000 cloud bill and the discovery that an old computer outperformed the cloud version of an early prototype. Ben's advice to most of the audience is not to use his product. But he draws the line at execution time and names the case where a sandbox is simply the better tool. His path to VMs ran through two abandoned products. A database engine built at Apple in evenings and carried through an entire YC batch was killed in November 2024 after an early o3 demo. A serverless JavaScript runtime with 10ms cold starts came next, and nobody wanted it. You'll also learn why Freestyle moved off Firecracker and what rust-vmm makes possible, the destructive-snapshot problem that caused Ben real pain, and why stripped-down kernels break on Supabase. 📎 READ MORE Freestyle: https://www.freestyle.sh Freestyle VMs: https://www.freestyle.sh/products/vms Ben on LinkedIn: https://www.linkedin.com/in/benswerdlow rust-vmm: https://github.com/rust-vmm/community 📚 MASTRA RESOURCES https://mastra.ai https://x.com/mastra_ai Discord: https://mastra.ai/community/discord GitHub: https://github.com/mastra-ai Learn Mastra: https://mastra.ai/learn Principles of Building AI Agents: https://mastra.ai/books/principles-of-building-ai-agents Patterns of Building AI Agents: https://mastra.ai/books/patterns-of-building-ai-agents WHAT IS MASTRA Mastra is an open-source TypeScript framework designed for building and shipping AI-powered applications and agents with minimal friction. It supports the full lifecycle of agent development — from prototype to production. AI Agents Hour is a weekly livestream by Mastra CPO Shane Thomas and CTO Abhi Aiyer. Mondays 12PM Pacific. ⏱️ CHAPTERS 00:00 Cold open 00:32 Ben Swerdlow, Freestyle 00:45 Why "sandbox" is the wrong word 00:58 The serverless JavaScript runtime nobody wanted 01:35 Apple, YC, and the database they killed 03:11 Running server racks in a house 05:43 What it takes to operate that 07:20 Agent VM vs. sandbox 08:07 Background agents and the 8GB ceiling 10:13 Firecracker, rust-vmm, and destructive snapshots 12:01 Demo 14:02 Networks and VPN APIs 16:22 Who should not use Freestyle 17:09 Virtualizing themselves on every PR 7:28 What Ben would tell you to build for
Transcribe →Opus 5 vs Fable, Rogue Model Hacks Hugging Face & the Open Weights Letter | This Week In AI
1w ago00:28:53Tap to summarizeBig week. OpenAI had a security incident, where a model reportedly broke out of its evaluation parameters and hacked Hugging Face — and the guardrails on OpenAI's and Anthropic's own models blocked responders from investigating, so they had to fall back to an open model to get to the bottom of it. Opus 5 lands close to Fable at 26% lower cost and tops the front-end arena, but the reviews are mixed, and the token usage is brutal — one chart clocks it at 97 million tokens on a task GPT 5.6 did in 4.3 million. Then the fight everyone's watching: NVIDIA, OpenAI, and a hundred others sign an open-models letter, Anthropic declines and calls for mandatory safety testing, and the room asks the obvious question — who gets to write the test? Plus MCP goes fully stateless (it's basically a REST API now), Kimi K3 weights drop with a K3.1 leak, SSI partners with NVIDIA, Stripe eyes OpenRouter for $10B, Cognition buys Poke, and DoorDash announces drone delivery — so your agent really can order you pizza by helicopter. AI Agents Hour is a weekly livestream by Mastra CPO Shane Thomas and CTO Abhi Aiyer. Mondays 12PM Pacific. 📚 READ MORE Anthropic destroyed the books (Project Panama): https://x.com/HedgieMarkets/status/2081534588485296565 OpenAI security incident: https://x.com/sama/status/2079661132302995790 Claude Opus 5 launch: https://x.com/claudeai/status/2080699495453528290 Opus 5 tops the front-end arena: https://x.com/arena/status/2081831019377004727 Opus 5 intelligence & cost: https://x.com/ArtificialAnlys/status/2080734447717298483 Siqi Chen walks it back: https://x.com/blader/status/2082275046937292859 Theo on Opus 5: https://x.com/theo/status/2081880182936502474 The token-usage chart: https://x.com/jpschroeder/status/2081751862768717956 Jensen's open-models letter: https://x.com/JensenHuang/status/2080643682408321103 Denny's stays open: https://x.com/dennysdiner/status/2081069889931112816 Anthropic declines to sign: https://x.com/anthropicai/status/2081864750296658008 MCP goes stateless: https://x.com/claudedevs/status/2082164248697069935 Kimi K3 weights are out: https://x.com/Kimi_Moonshot/status/2081760186235289764 Kimi K3.1 leak: https://x.com/Mr_Salio/status/2081345681734713635 SSI x NVIDIA: https://x.com/ssi/status/2081732119194394763 Stripe eyes OpenRouter: https://x.com/andrewcurran_/status/2080397684134007189 Notion as Code: https://x.com/notionhq/status/2080331924732850687 Cognition acquires Poke: https://x.com/cognition/status/2080311229256540194 ChatGPT Voice on desktop: https://x.com/openai/status/2080378182469857576 DoorDash Air: https://x.com/DoorDash/status/2082451101840740753 📚 MASTRA RESOURCES Mastra: https://mastra.ai Mastra on X: https://x.com/mastra_ai Mastra Discord: https://mastra.ai/community/discord Mastra GitHub: https://github.com/mastra-ai Learn Mastra: https://mastra.ai/learn Principles of Building AI Agents (Book): https://mastra.ai/books/principles-of-building-ai-agents Patterns for Building AI Agents (New Book): https://mastra.ai/books/patterns-of-building-ai-agents WHAT IS MASTRA? Mastra is an open-source TypeScript framework designed for building and shipping AI-powered applications and agents with minimal friction. It supports the full lifecycle of agent development—from prototype to production. You can integrate it with frontend and backend stacks (e.g., React, Next.js, Node) or run agents as standalone services. If you're a JavaScript or TypeScript developer looking to build an agentic or AI-powered product without starting from first principles, Mastra provides the scaffolding, tools, and integrations to accelerate that process. ⏱️ CHAPTERS 00:00 Cold open 00:27 Welcome 00:34 Anthropic destroyed the books ("Operation Panama") 02:32 The OpenAI security incident: a model hacked Hugging Face 04:35 Opus 5 launches — is it better than Fable? 05:46 The mixed reviews 07:05 Opus 5 is token-hungry 08:17 The open-models letter: who writes the safety test? 10:03 Anthropic declines to sign 13:35 The 1,200-employee "pacing" statement 15:28 Trump's China AI advisors 15:48 Be a guest on the show 16:03 MCP is not dead, it's just stateless now 22:24 Quick hits: Kimi K3 weights & the K3.1 leak 23:11 SSI + NVIDIA 23:43 Stripe eyes OpenRouter for $10B 24:24 Notion as Code 25:28 Cognition acquires Poke 25:59 ChatGPT voice & the beach test 27:14 DoorDash Air: pizza by drone 28:10 Outro
Transcribe →Is Gemini Actually a Coding Model? Google DeepMind's Ivan Leo Answers
2w ago00:17:42Tap to summarizeIvan Leo works on Developer Experience at Google DeepMind. He joins Shane Thomas and Abhi Aiyer to demo what he has been building on Gemini and to answer the question Google keeps getting asked: is Gemini actually a coding model? Every demo in this episode runs on Gemini 3.5 Flash. Gemini Managed Agents was the first product Ivan took from first commit to launch. One API call opens the same Antigravity harness Google runs internally and provisions a persistent Linux sandbox. On stream, it builds a NotebookLM-style podcast generator from a one-line prompt, with native text-to-speech. You'll also see a one-shot Windows XP clone, with an animated Clippy and a working Tamagotchi, as well as demos of Gemini Omni, the conversational video model Ivan's team describes as Nano Banana for video. You'll also hear which everyday errand finally made Ivan feel the AGI. Google's internal token consumption has been doubling every few weeks. Ivan was previously a core contributor to Instructor and built knowledge agents at Manus before its acquisition by Meta. 📎 READ MORE Ivan Leo: https://ivanleo.com Ivan on X: https://x.com/ivanleomk Gemini Managed Agents: https://blog.google/innovation-and-ai/technology/developers-tools/managed-agents-gemini-api/ Gemini Omni Flash: https://ai.google.dev/gemini-api/docs/models/gemini-omni-flash fofr on X: https://x.com/fofrAI 📚 MASTRA RESOURCES Website: https://mastra.ai X: https://x.com/mastra_ai Discord: https://mastra.ai/community/discord GitHub: https://github.com/mastra-ai Learn Mastra: https://mastra.ai/learn Principles of Building AI Agents: https://mastra.ai/books/principles-of-building-ai-agents Patterns of Building AI Agents: https://mastra.ai/books/patterns-of-building-ai-agents WHAT IS MASTRA Mastra is an open-source TypeScript framework designed for building and shipping AI-powered applications and agents with minimal friction. It supports the full lifecycle of agent development — from prototype to production. ⏱️ CHAPTERS 00:00 Cold open 00:19 Welcome, Ivan Leo 00:48 Instructor, Manus, and the road to DeepMind 02:54 What actually shipped at I/O 03:32 Launching Gemini Managed Agents 04:03 3.2 quadrillion tokens a month 04:44 Answering the "DeepMind is cooked" post 05:57 Why Google Search came back 06:45 Antigravity: CLI, app, and IDE 07:31 Gemini Omni, or Nano Banana for video 09:38 Demo: one-shot apps on Gemini 3.5 Flash 10:51 The errand that felt like AGI 11:08 Demo: Managed Agents building a NotebookLM clone 13:13 Is Gemini a coding model? 14:54 Learning the temperament of a new model 15:47 Prediction: agents in everyday life 17:01 Where to find Ivan
Transcribe →Is Ramp an AI Company Now? Plus Kimi K3 Beats Fable on Front-End | This Week In AI
2w ago00:27:32Tap to summarizeShane and Abhi are live from London this week, and the big one is Kimi K3 — a 2.8 trillion parameter open frontier model with a million-token context window that lands right next to Fable 5 on the benchmarks, takes the number one spot on the front-end Code Arena, and reportedly doubles Fable's performance on legal tasks. Then demand got so heavy Moonshot had to pause new subscriptions. Also this week: Anthropic scrambles to keep Fable in Max and Team plans, an Opus 5 leak, and the distillation debate — if you paid for the output, whose data is it? Plus Anthropic's $1.5B copyright settlement, Grok Build going open source, Thinking Machines' Inkling, three new Gemini models that land with a thud, and Ramp shipping Ramp Router — which prompts Shane's bold call that Ramp is about to become an AI company with a fintech arm, not the other way around. Then DoorDash ships a CLI for your agent, and everyone argues about what a "factory engineer" actually is. AI Agents Hour is a weekly livestream by Mastra CPO Shane Thomas and CTO Abhi Aiyer. Mondays 12PM Pacific. 📚 READ MORE Kimi K3 launch: https://x.com/Kimi_Moonshot/status/2077830229968683203 Kimi K3 on Artificial Analysis: https://x.com/ArtificialAnlys/status/2077832874183860404 Number one on the front-end arena: https://x.com/arena/status/2077824029126504525 Kimi K3 on legal tasks: https://x.com/nrehiew_/status/2077782070785634767 Open weights coming (Yulun Du): https://x.com/yulun_du/status/2077831915999228192 Pausing new subscriptions: https://x.com/kimi_moonshot/status/2078855608565207130 Fable stays in Max & Team plans: https://x.com/claudeai/status/2078302415804379218 Opus 5 leak: https://x.com/vral/status/2079267940021477864 The distillation debate: https://x.com/pankajkumar_dev/status/2079288624928158160 Grok Build goes open source: https://x.com/spacexai/status/2077494535387828644 Thinking Machines' Inkling: https://x.com/thinkymachines/status/2077454609551921208 Three new Gemini models: https://x.com/Google/status/2079589747366724030 Ramp Router: https://x.com/andyfang/status/2077516962515599799 DoorDash CLI: https://x.com/denisewu/status/2077984660211269870 Build your own CLI from network requests (Dax): https://x.com/thdxr/status/2078727284865827140 OpenShip: https://x.com/openshipio/status/2078143909125923137 AI SDK for Python: https://x.com/1st1/status/2078393737407848683 What the devil is a factory engineer? https://www.youtube.com/watch?v=qd6ltKUcp9U Yardarm, an Electron app for Mastra Code: https://github.com/JJJ-Mo3/yardarm 📚 MASTRA RESOURCES Mastra: https://mastra.ai Mastra on X: https://x.com/mastra_ai Mastra Discord: https://mastra.ai/community/discord Mastra GitHub: https://github.com/mastra-ai Learn Mastra: https://mastra.ai/learn Principles of Building AI Agents (Book): https://mastra.ai/books/principles-of-building-ai-agents Patterns for Building AI Agents (New Book): https://mastra.ai/books/patterns-of-building-ai-agents WHAT IS MASTRA? Mastra is an open-source TypeScript framework designed for building and shipping AI-powered applications and agents with minimal friction. It supports the full lifecycle of agent development—from prototype to production. You can integrate it with frontend and backend stacks (e.g., React, Next.js, Node) or run agents as standalone services. If you're a JavaScript or TypeScript developer looking to build an agentic or AI-powered product without starting from first principles, Mastra provides the scaffolding, tools, and integrations to accelerate that process. ⏱️ CHAPTERS 00:00 Cold open 00:27 Welcome — live from London 01:08 Kimi K3: 2.8T parameters, open frontier intelligence 02:57 Number one on the front-end arena 04:25 Kimi K3 doubles Fable on legal tasks 05:07 Demand breaks capacity: subscriptions paused 05:53 What is Alibaba Cloud, and why speed matters 07:13 Fable stays in Max & Team plans 07:56 The Opus 5 leak 08:32 The distillation debate: whose data is it? 12:54 Anthropic's $1.5B copyright settlement 13:34 Be a guest on the show 13:48 Grok Build goes open source 14:37 Thinking Machines' Inkling & the fine-tuning future 16:09 Three new Gemini models (and a sick dark horse) 18:44 Is Ramp an AI company? Ramp Router & the pivot call 21:53 Will our agents order us pizza? DoorDash CLI 23:07 Build your own CLI from network requests 23:54 Quick hits: OpenShip & AI SDK for Python 24:21 What the devil is a factory engineer? 26:31 GitHub star party: Yardarm 27:02 Outro
Transcribe →Three Models Drop: Grok 4.5, Meta Muse & GPT-5.6 Sol (Which Nuked a Business) | This Week In AI
4w ago00:35:29Tap to summarizeThree model drops, a lawsuit, and a couple of genuine horror stories. Shane and Abhi cover the busiest week in a while — Grok 4.5, Meta's Muse Spark 1.1, and the one everyone actually tried, GPT-5.6 Sol. Sol lands right around Fable on the benchmarks at a third the price, and the receipts are wild: it beats Fable on cost in real usage, tops the Code Arena front-end board — and, in a few now-infamous posts, deleted someone's Mac files and canceled every active Stripe subscription a business had while its owner slept. Meanwhile Anthropic extends Fable again (compute crunch and all), Apple sues OpenAI over trade secrets in the same 24 hours OpenAI's head of safety quit, China floats restricting its open models, and the "own your stack" argument gets a Satya Nadella cosign. Plus software factories at Uber and Sierra, Cognition's SWE-1.7, Bun rewritten in Rust, TypeScript 7, 1X's eerily good robot hands, and distilling your own small model with Inference Autotune. AI Agents Hour is a weekly livestream by Mastra CPO Shane Thomas and CTO Abhi Aiyer. Mondays 12PM Pacific. 📚 READ MORE Grok 4.5 (Elon): https://x.com/elonmusk/status/2074740539874775163 Grok dumping repo data: https://x.com/XBToshi/status/2076338252051841512 Muse Spark 1.1 (Zuckerberg): https://x.com/finkd/status/2075218444056707458 GPT-5.6 launch: https://x.com/openai/status/2074704958419792299 Mitchell Hashimoto's verdict: https://x.com/mitchellh/status/2074862990214787301 Sol vs Fable cost (Dax): https://x.com/thdxr/status/2075963734745342361 Code Arena front-end #1: https://x.com/arena/status/2075672492312768683 Sol deleted my Mac files: https://x.com/mattshumer_/status/2075657271401390161 Sol canceled every Stripe sub: https://x.com/bridgemindai/status/2074604410727288974 Apple sues OpenAI: https://x.com/SawyerMerritt/status/2075680682844569993 OpenAI's brutal 24 hours: https://x.com/katiemiller/status/2075996304883265935 Fable extended to July 19: https://x.com/claudeai/status/2076351399999557669 China may restrict its models: https://x.com/jukan05/status/2074443936865960205 Uber's agentic pods: https://x.com/praveenTweets/status/2074605343439810922 Harness benchmarks (Ali Ghodsi): https://x.com/alighodsi/status/2074996561306955958 Cognition SWE-1.7: https://x.com/cognition/status/2074882968770728416 Bun rewritten in Rust (Jarred): https://x.com/jarredsumner/status/2074973674332123157 TypeScript 7: https://x.com/typescript/status/2074892178745327926 1X's Neo hands: https://x.com/1x_tech/status/2075252899442204952 Satya's reverse information paradox: https://x.com/satyanadella/status/2076323181154230284 Notion ShipOS: https://x.com/NotionHQ/status/2075249002208227743 Inference Autotune (Sam Hogan): https://x.com/samhogan/status/2076044602554159240 📚 MASTRA RESOURCES Mastra: https://mastra.ai Mastra on X: https://x.com/mastra_ai Mastra Discord: https://mastra.ai/community/discord Mastra GitHub: https://github.com/mastra-ai Learn Mastra: https://mastra.ai/learn Principles of Building AI Agents (Book): https://mastra.ai/books/principles-of-building-ai-agents Patterns for Building AI Agents (New Book): https://mastra.ai/books/patterns-of-building-ai-agents WHAT IS MASTRA? Mastra is an open-source TypeScript framework designed for building and shipping AI-powered applications and agents with minimal friction. It supports the full lifecycle of agent development—from prototype to production. You can integrate it with frontend and backend stacks (e.g., React, Next.js, Node) or run agents as standalone services. If you're a JavaScript or TypeScript developer looking to build an agentic or AI-powered product without starting from first principles, Mastra provides the scaffolding, tools, and integrations to accelerate that process. ⏱️ CHAPTERS 00:00 Cold open 00:30 Welcome & London next week 01:48 Grok 4.5: benchmarks, data grabbing & ZDR 04:51 Meta's Muse Image & Muse Spark 1.1 07:21 GPT-5.6 Sol, Terra & Luna 10:38 The cost picture: Sol vs Fable 11:51 Sol's horror stories: deleted files & canceled subs 13:43 Subscribe break 13:53 GPT Live, Apple sues OpenAI & a brutal 24 hours 16:06 Fable extended — can Anthropic even drop it? 19:17 The Anthropic backlash 20:25 China restricts models, Z.ai chips & MiniMax M3 Pro 21:42 Software factories: Uber, Sierra & the harness 24:55 Cognition's SWE-1.7 25:19 Bun rewritten in Rust & the "asshole reviewer" 28:32 Quick hits: TypeScript 7 & 1X's Neo hands 29:48 Satya's "reverse information paradox" 31:35 Notion ShipOS 32:25 Inference Autotune 34:49 Outro
Transcribe →Is Sonnet 5 Even Worth It? Plus: Fable's Back, and War on Tokenmaxxing | This Week In AI
Jul 900:25:01Tap to summarizeAnthropic lived rent-free in everyone's feed this week. Claude Sonnet 5 launched as "the most agentic Sonnet yet" — and within 24 hours, the government lifted the export controls on Fable 5 and Mythos 5. Shane and Abhi unpack both, plus the catch: Fable came back wrapped in new classifiers that quietly fall it back to Opus for coding, and it's coming off subscription plans on July 7. That kicks off the routing wars — Devin Fusion, sidekick, mid-session routing — and a real question: if you're already using Fable to plan and something cheaper to code, you're the router. Then a stacked week of open and Chinese models (ZCode for GLM-5.2, DeepSeek V4's peak-hour pricing, Tencent's Hunyuan HY3), new inference silicon from Etched and Cerebras, an agent-memory result showing 5x more solved tasks, and the sleeper hot take: Palantir's "AI Sovereignty" manifesto and its shot at tokenmaxxing. Plus agentcn's Mastra recipes, Anthropic's own chip talks, and Anthropic's new "global workspace" research. AI Agents Hour is a weekly livestream by Mastra CPO Shane Thomas and CTO Abhi Aiyer. Mondays 12PM Pacific. 📚 READ MORE Claude Sonnet 5: https://x.com/claudeai/status/2072017450611142835 "Opus 4.8 is cheaper and better": https://x.com/Angaisb_/status/2072017407653060848 Export controls lifted: https://x.com/anthropicai/status/2072106151890809341 Fable back, with new classifiers: https://x.com/anthropicai/status/2072163884430229756 Fable off subscriptions July 7: https://x.com/trq212/status/2072814903170408784 Devin Fusion: https://x.com/cognition/status/2071624568465490170 Sidekick & mid-session routing: https://x.com/walden_yan/status/2071627241818399181 ZCode for GLM-5.2: https://x.com/zai_org/status/2072349453361557898 DeepSeek V4 & peak pricing: https://x.com/teortaxesTex/status/2071537369631625608 Tencent Hunyuan HY3: https://x.com/TencentHunyuan/status/2074148098876768478 Etched out of stealth: https://x.com/Etched/status/2071972062202343590 Gemma 4 31B on Cerebras: https://x.com/cerebras/status/2071776102633410684 Observational memory, 5x solves: https://x.com/hampsonw/status/2071728496909643917 Understand the code (Geoffrey Litt): https://x.com/geoffreylitt/status/2072522251300409556 Palantir's AI Sovereignty: https://x.com/PalantirTech/status/2072114267776491695 Primeagen's co-sign: https://x.com/ThePrimeagen/status/2072679880626512256 agentcn Mastra recipes: https://x.com/alaymanguy/status/2073500711737008270 Anthropic's chip talks: https://x.com/kimmonismus/status/2072685149708300360 Supermemory local: https://x.com/supermemory/status/2072086481997537301 Vercel + Shopify rebuild Hydrogen: https://x.com/vercel/status/2072041488012439803 "Th..thanks Dario": https://x.com/aljosa/status/2072175046324277644 📚 MASTRA RESOURCES Mastra: https://mastra.ai Mastra on X: https://x.com/mastra_ai Mastra Discord: https://mastra.ai/community/discord Mastra GitHub: https://github.com/mastra-ai Learn Mastra: https://mastra.ai/learn Principles of Building AI Agents (Book): https://mastra.ai/books/principles-of-building-ai-agents Patterns for Building AI Agents (New Book): https://mastra.ai/books/patterns-of-building-ai-agents ⏱️ CHAPTERS 00:00 Cold open 00:29 Welcome 01:08 Claude Sonnet 5 — and the Opus 4.8 counter 04:04 Fable's export controls lifted (and the coding catch) 05:51 Daily-driving Fable & the July 7 subscription cliff 08:07 The routing wars: Devin Fusion & mid-session routing 09:45 OSS & China: ZCode, DeepSeek V4, Hunyuan HY3 11:57 Inference hardware: Etched & Cerebras 12:58 Agent memory: 5x more solves 14:40 You still have to understand the code 16:18 Tokenmaxxing & Palantir's AI Sovereignty 20:57 agentcn's Mastra recipes & Anthropic's chip talks 21:40 Supermemory local & Vercel/Shopify Hydrogen 22:11 The "Thanks Dario" meme 22:29 Anthropic's "global workspace" research 23:57 Outro
Transcribe →The Software Factory: Dex Horthy on Shipping Fast Without AI Slop
Jul 700:22:36Tap to summarizeDex Horthy has spent the past year studying what happens to a codebase when you let AI write all of it and stop reading the output. His answer has a name: slop. And it's not just about sloppy prompting. In this episode, the HumanLayer founder Dex Horthy joins Shane and Abhi to draw the software factory on a whiteboard — the whole pipeline that builds the thing that builds the thing — and show exactly where teams tear out the parts that were holding quality up. He moves from the naive factory to the fully agentic one to the "lights-off" version where humans stop reading code entirely, then explains why that dream quietly rots a codebase after three to six months. Dex, Shane, and Abhi also talk about how coding models are trained. Every benchmark rewards one thing: did the test pass? Nowhere in the loop does a model get punished for leaving behind code that a smaller model, or a human six months later, can't work in. That gap is why architecture planning, sprint planning, and code review still can't be skipped. You'll also hear why you're not drowning in PRs so much as drowning in bad ones, what a longer-horizon benchmark that actually catches slop would look like, and how many times Dex has watched the HumanLayer launch film. 🔗 DEX HORTHY / HUMANLAYER Dex on X: https://x.com/dexhorthy HumanLayer: https://humanlayer.dev AI Agents Hour is a weekly livestream by Mastra CPO Shane Thomas and CTO Abhi Aiyer. Mondays 12PM Pacific. 📚 MASTRA RESOURCES Mastra: https://mastra.ai Mastra on X: https://x.com/mastra_ai Mastra Discord: https://mastra.ai/community/discord Mastra GitHub: https://github.com/mastra-ai Learn Mastra: https://mastra.ai/learn Principles of Building AI Agents: https://mastra.ai/books/principles-of-building-ai-agents Patterns for Building AI Agents: https://mastra.ai/books/patterns-of-building-ai-agents WHAT IS MASTRA Mastra is an open-source TypeScript framework designed for building and shipping AI-powered applications and agents with minimal friction. It supports the full lifecycle of agent development — from prototype to production. CHAPTERS 0:00 Cold open 0:14 Bringing on Dex Horthy 1:16 The HumanLayer launch film 3:31 Cutting through the AI hype 4:07 Defending the terms that matter 5:26 "I want to build a software factory" 5:58 What a software factory actually is 9:08 Agentic code review, testing, and feedback loops 13:35 The lights-off software factory 14:32 Why code quality decays over time 15:36 What RL actually rewards 17:29 Don't sacrifice planning or review 18:05 Getting the PR to 95% before review 19:21 Drowning in bad PRs, not PRs 20:22 Building longer-horizon benchmarks 21:46 Trying HumanLayer
Transcribe →GPT-5.6 Needs Government Approval Now, npm Finally Gets More Security, and Fable Returns | This Week In AI
Jul 200:22:36Tap to summarizeGPT-5.6 is here — Sol, Terra, and Luna — but for the first time, the US government decides who gets access, approving customers one by one. Shane and Abhi dig into what that precedent means for competition, open models, and every lab that comes next. Then a stacked week: OpenAI's own Jalapeño chip, open models pushing hard (Qwen AgentWorld, DeepSeek DSpark, Hermes mixture-of-agents), the model-routing wave (AI SDK 7, Devin Fusion), Claude Tag's viral moment and the Trojan-horse worry, voice agents on Vercel, Coinbase on keeping AI spend flat, a16z's $200M "seed" for Mirendil, npm finally freezing accounts after the Axios and Mastra attacks, Meta's think-to-text Brain2Qwerty, the DevRel Google fired over a viral CLI, and Fable's return after 15 days offline. AI Agents Hour is a weekly livestream by Mastra CPO Shane Thomas and CTO Abhi Aiyer. Mondays 12PM Pacific. 📚 READ MORE GPT-5.6 Sol/Terra/Luna: https://x.com/OpenAI/status/2070555272230384038 Restricted rollout: https://x.com/steph_palazzolo/status/2070241787180966279 Jalapeño chip: https://x.com/OpenAI/status/2069770172802773292 Qwen AgentWorld: https://x.com/Alibaba_Qwen/status/2069720365442719867 DeepSeek DSpark: https://x.com/Yuchenj_UW/status/2070928299744972814 Hermes Mixture of Agents: https://x.com/NousResearch AI SDK 7: https://x.com/vercel/status/2070155382488764566 Devin Fusion: https://x.com/cognition/status/2071624568465490170 Tau harness: https://x.com/_alejandroao Claude Tag (Karpathy): https://x.com/karpathy Claude Tag Trojan horse: https://x.com/ashwingop Open Tag: https://x.com/ataiiam GLM-5.2 in the wild: https://x.com/cline/status/2069171146994729078 Local 1-bit GLM: https://x.com/UnslothAI/status/2069418532375564484 Coinbase on AI spend: https://x.com/brian_armstrong Coinbase gateway: https://x.com/markletree Exa Connect: https://x.com/ExaAILabs/status/2069842203577651283 Mirendil $200M: https://x.com/a16z/status/2069869327411749012 npm account freeze: https://x.com/AikidoSecurity Brain2Qwerty: https://x.com/AIatMeta/status/2071566924803395741 Google Workspace CLI: https://x.com/JPoehnelt 📚 MASTRA RESOURCES Mastra: https://mastra.ai Mastra on X: https://x.com/mastra_ai Mastra Discord: https://mastra.ai/community/discord Mastra GitHub: https://github.com/mastra-ai Learn Mastra in the world's first MCP-Based Course: https://mastra.ai/course Principles of Building AI Agents (Book): https://mastra.ai/books/principles-of-building-ai-agents Patterns for Building AI Agents (New Book): https://mastra.ai/books/patterns-of-building-ai-agents WHAT IS MASTRA? Mastra is an open-source TypeScript framework designed for building and shipping AI-powered applications and agents with minimal friction. It supports the full lifecycle of agent development—from prototype to production. You can integrate it with frontend and backend stacks (e.g., React, Next.js, Node) or run agents as standalone services. If you're a JavaScript or TypeScript developer looking to build an agentic or AI-powered product without starting from first principles, Mastra provides the scaffolding, tools, and integrations to accelerate that process. ⏱️ CHAPTERS 00:00 Cold open: npm finally freezes accounts 00:29 Welcome 00:37 GPT-5.6 Sol, Terra, Luna — and government-gated access 04:25 OpenAI's Jalapeño chip 05:34 Open models: Qwen AgentWorld, DeepSeek, Hermes MoA 07:11 AI SDK 7 07:43 Devin Fusion & the model-routing wave 10:26 Harness interop: OpenCode & LangChain 11:01 Tau harness 11:26 Claude Tag's viral moment & the Trojan horse 12:49 Open Tag 13:12 Voice agents land on Vercel 13:53 The infra bill: Coinbase on keeping AI spend flat 15:15 Exa Connect & a16z's $200M "seed" 16:32 npm freezes accounts 18:08 Meta's Brain2Qwerty: think-to-text 19:11 Google fired its viral CLI star 21:48 Fable returns 22:00 Outro
Transcribe →Mastra Got Hacked. Here's What We Learned
Jun 2600:18:57Tap to summarizeMastra got hacked. In this special edition of Security Corner, Shane Thomas and Abhi Aiyer break down exactly what happened when a supply chain attack hit Mastra's npm packages — an attack that appears to trace back to hackers in North Korea. They're joined by Ismail Pelaseyed, co-founder and CTO of Superagent, for the outside view on how these campaigns actually work. You'll hear how a single, ordinary-looking call turned into a full npm account takeover, the small oversight that turned a scare into a genuine crisis, and why a malicious package was still live on the registry weeks after it was reported. Ismail makes the case that getting hacked is a side effect of success — and that the real problem runs deeper than any one team. You'll learn why he thinks npm and PyPI have dropped the ball on security, how AI now lets attackers one-shot a convincing phishing app, and what every maintainer should be doing to harden their pipeline before a trusted contributor becomes the way in. It's the unfiltered version, told by the people who lived through it. Connect with Ismail Pelaseyed: https://x.com/pelaseyed https://superagent.sh Connect with the hosts: https://x.com/smthomas3 https://x.com/abhiaiyer 📚 MASTRA RESOURCES https://mastra.ai https://x.com/mastra_ai https://mastra.ai/community/discord https://github.com/mastra-ai https://mastra.ai/course https://mastra.ai/books/principles-of-building-ai-agents https://mastra.ai/books/patterns-of-building-ai-agents Mastra is an open-source TypeScript framework designed for building and shipping AI-powered applications and agents with minimal friction. It supports the full lifecycle of agent development — from prototype to production. CHAPTERS 0:00 Intro: a special Security Corner 0:55 The supply chain attack on Mastra 1:29 How they got in: a fake Teams call 2:27 The npm account takeover 3:54 EasyDjS and the scramble to fix it 5:49 Why success makes you a target 9:20 How AI supercharges phishing 9:59 Hardening against compromised contributors 11:10 Open source under strain: IBM's $5B bet 12:27 npm and PyPI keep dropping the ball 14:31 Inside the fake package, and how Socket caught it 16:20 The fear-selling problem in security 18:02 Superagent!
Transcribe →GPT-5.5 Beats Fable, Cursor Takes On GitHub & Midjourney Scans Your Body | This Week in AI
Jun 2500:20:09Tap to summarizeFable is gone, and the race to replace it is already on. Shane and Abhi open with the fallout — reports that Mythos breached classified NSA systems, the ID-verification future that may follow, and David Sacks' "you asked for it" take. Then the story flips: OpenAI's GPT-5.5 Cyber lands state-of-the-art on Cyber Gym, beating Fable and Mythos on the underlying benchmarks, with OpenAI Daybreak shipping alongside it. GPT-5.6 leaks, then slips to mid-July. GLM 5.2 arrives near Opus 4.8, tops Design Arena, and runs on a Mac. Cursor goes after GitHub with Origin, adds mobile and its own frontier model on Colossus after the SpaceX deal closes. Vercel ships Eve, Fred Schott counters with Flue, Claude lands in Slack, and Midjourney — yes, the image company — unveils a full-body medical scanner headed for spas. AI Agents Hour is a weekly livestream by Mastra CPO Shane Thomas and CTO Abhi Aiyer. Mondays 12PM Pacific. 📚 READ MORE The Fable story, Mythos & the NSA: https://x.com/kimmonismus/status/2068605229965234238 GPT-5.5 Cyber beats Fable (Sam Altman): https://x.com/sama/status/2069121360744550796 OpenAI Daybreak: https://x.com/OpenAI/status/2069104283824640023 GPT-5.6 leak (Mark K): https://x.com/mark_k/status/2069169916432003105 GPT-5.6 delayed: https://x.com/synthwavedd/status/2069432791184650426 GLM 5.2 (Z.ai): https://x.com/zai_org/status/2066938937344495629 GLM 5.2 tops Design Arena: https://x.com/Designarena/status/2066940737011560652 GLM 5.2 vs Opus (Cline): https://x.com/cline/status/2069171146994729078 1-bit GLM 5.2 runs local (Unsloth): https://x.com/UnslothAI/status/2069418532375564484 Cursor Origin: https://x.com/cursor_ai/status/2069149296436330776 Cursor Compile keynote: https://x.com/cursor_ai/status/2067012220832329782 Swyx on Origin: https://x.com/swyx/status/2066928345246470204 Vercel Eve: https://x.com/vercel/status/2067180054979936413 Flue 1.0 (Fred Schott): https://x.com/FredKSchott/status/2066962296119959581 Codex record-replay to skills: https://x.com/OpenAIDevs/status/2067681320281723113 Claude Tag in Slack: https://x.com/claudeai/status/2069468693017268244 Replit in Slack: https://x.com/Replit/status/2067661213278900402 Midjourney Medical: https://x.com/midjourney/status/2067421950314688759 Sakana Fugu: https://x.com/SakanaAILabs/status/2068861630327443966 📚 MASTRA RESOURCES Mastra: https://mastra.ai Mastra on X: https://x.com/mastra_ai Mastra Discord: https://mastra.ai/community/discord Mastra GitHub: https://github.com/mastra-ai Principles of Building AI Agents (Book): https://mastra.ai/books/principles-of-building-ai-agents Patterns for Building AI Agents (New Book): https://mastra.ai/books/patterns-of-building-ai-agents WHAT IS MASTRA? Mastra is an open-source TypeScript framework designed for building and shipping AI-powered applications and agents with minimal friction. It supports the full lifecycle of agent development—from prototype to production. You can integrate it with frontend and backend stacks (e.g., React, Next.js, Node) or run agents as standalone services. If you're a JavaScript or TypeScript developer looking to build an agentic or AI-powered product without starting from first principles, Mastra provides the scaffolding, tools, and integrations to accelerate that process. ⏱️ CHAPTERS 00:00 Cold open 00:30 Welcome 01:02 Fable fallout: did Mythos breach the NSA? 03:10 GPT-5.5 Cyber beats Fable & OpenAI Daybreak 05:23 GPT-5.6 delayed, Gemini 3.5 Pro slips 05:52 GLM 5.2: near-Opus, and it runs on a Mac 09:52 Subscribe break 09:55 Cursor's Origin takes on GitHub 11:12 Cursor mobile, frontier model & the SpaceX deal 11:52 Vercel's Eve, Flue & Codex skills 13:48 Claude Tag & Replit in Slack 14:57 Midjourney goes medical: full-body scanners 16:42 Sakana Fugu & the rise of model routing 19:37 Outro
Transcribe →