MTS

Carnegie Mellon professor and ClickHouse VP of Database Research Andy Pavlo breaks down why AI agents demand an entirely new database layer. He covers how nonstop query volumes and vibe coding break legacy systems, ClickHouse’s expansion into managed Postgres with the Xata acquisition, and his predictions for agent-written databases. Check out our sponsor Lovable: https://lovable.dev. Turn ideas i...

Calif founder and offensive security researcher Thai Duong discusses how frontier AI models are shrinking exploit development timelines from months to weeks, why AI labs and neo clouds represent the most vulnerable infrastructure, and the breakdown of traditional cyber defense against concurrent agent swarms. -- https://x.com/XorNinja -- MTS is a live news and interview show covering technology, b...

METR Chief of Staff Parv Mahajan breaks down their new report assessing AI R&D acceleration inside Anthropic. He discusses finding a sustained 2x acceleration since late 2025, why public models hide this kink, the absence of near-term training compute bottlenecks, and why Anthropic likely crossed its own acceleration warning threshold. Check out our sponsor Lovable: https://lovable.dev. Turn i...

Google DeepMind Director of AGI Economics Alex Imas breaks down why hockey-stick AI model capabilities will not trigger explosive double-digit GDP growth. He explains how competitive markets drive automated output prices down, why firms operate as mosaic superintelligence, and how minimal AGI shifts labor into new high-wage markets. Check out our sponsor Lovable: https://lovable.dev. Turn ideas in...

Raindrop co-founder and CTO Ben Hylak explains why smarter models make AI agent failures more catastrophic, from healthcare liabilities to high-stakes alignment breakdowns. He breaks down how recursive self-improvement requires simulating past production traces to safely test model switches and prevent silent failures. Check out our sponsor Lovable: https://lovable.dev. Turn ideas into software pe...

Wayve simulation and validation head Brandon Basso joins MTS to explain why end-to-end AI models are replacing modular self-driving checker stacks. He unpacks how physical AI learns safety intrinsically across global edge cases and predicts when eyes-off autonomous systems will reshape daily transit. Check out our sponsor Lovable: https://lovable.dev. Turn ideas into software people love. MTS is a...

Dan Lee, co-founder and CEO of Nooks, explains how their $500M non-dilutive capital facility from General Catalyst powers specialized AI agents to automate the revenue lifecycle. He outlines why sales is a persuasion problem rather than pure reasoning, details training open-weight models against Claude and ChatGPT, and debuts Node, an AI assistant for sales teams. Check out our sponsor Lovable: ht...

Stanford finance professor Joshua Rauh breaks down why California’s proposed Prop 40 billionaire wealth tax faces low odds and threatens to accelerate capital flight. He explains how state and federal proposals could tax unrealized startup equity, and argues California’s structural deficit stems from runaway spending rather than a lack of tax revenue. Check out our sponsor Lovable: https://lovable...

Google DeepMind Director of AGI Economics Alex Imas breaks down why hockey-stick AI model capabilities will not trigger explosive double-digit GDP growth. He explains how competitive markets drive automated output prices down, why firms operate as mosaic superintelligence, and how minimal AGI shifts labor into new high-wage markets. Check out our sponsor Lovable: https://lovable.dev. Turn ideas in...

Former a16z investor Robin Guo explains how LionBio is developing nasal delivery platforms to bypass the blood-brain barrier for GLP-1s and neuropeptides. He discusses how frontier AI models are accelerating target discovery, why traditional pharma timelines are artificially slow, and how preventative neurotherapeutics could reshape diseases like Alzheimer’s over the next decade. Check out our spo...

Security researcher Max Justicz explains how using OpenAI’s Daybreak Blue and an obliterated GLM 5.3 model enabled him to uncover severe zero-day exploits in apt and Android in just a few hours. He warns that frontier AI capabilities are lowering the technical bar for cyber offense, making widespread vulnerability discovery inevitable. Check out our sponsor Lovable: https://lovable.dev. Turn ideas...

AI researcher and writer Herbie Bradley joins MTS to break down Anthropic's amended S-1 filing on imminent RSI thresholds, why fast takeoff timelines are overblown, and why enterprise AI adoption is hitting organizational bottlenecks rather than capability walls. Check out our sponsor Lovable: https://lovable.dev. Turn ideas into software people love. MTS is a live news and interview show cove...

StackAI co-founder Bernardo Aceituno explains why open-weight models have caught up to frontier closed models for enterprise automation, arguing that AGI for enterprise tasks arrived in February 2026. He breaks down how soaring token bills and on-premise governance are driving 90% of enterprise workloads to open models, while outlining the remaining bottlenecks in visual understanding, memory, and...

Ruby Justice Thelot and Sofia Puccini explore how generative AI coding tools have eliminated the technical barrier for artists, igniting a resurgence of creative technology. They trace corporate-artist collaborations from Bell Labs and Xerox PARC to Sora and Anthropic, while unpacking the tension between corporate KPIs and experimental art. -- https://x.com/being_on_line Brought to you by: Mongo D...

HumanLayer founder and CEO Dexter Horthy breaks down why his multiplayer coding agent platform is joining Factory and open-sourcing its code base. He explains why agent swarms doing months of work in days force software planning to shift left, and outlines the engineering challenges of managing live multiplayer state, security, and governance across distributed agent harnesses. Check out our spons...

Extrapolate co-founder and former xAI researcher Qasim Wani argues recursive self-improvement is fundamentally a data bottleneck rather than a model training problem. He breaks down why today’s frontier models fail at automated AI research, how agents accidentally leak commit patches to reward-hack tasks, and why building high-quality environments is the true path to superintelligence. Check out o...

Palisade Research Executive Director Jeffrey Ladish analyzes METR’s report showing AI capabilities accelerating twice as fast as historical trends, driven by algorithmic efficiency and autonomous AI development. He discusses recursive self-improvement timelines, why Anthropic must un-redact internal evaluations, and how close autonomous agents are to compromising digital infrastructure. Check out...

Ruby Justice Thelot explores the cultural and technological counter-currents to modern hyper-digitization, from the death of Detroit techno pioneer Juan Atkins to the failure of tech-fashion house Coperni. He analyzes the corporate aesthetics overtaking social platforms and breaks down data behind the resurgence of analog media alongside the explosive commercial rise of paratechnology designed to...

Congressman Sam Liccardo breaks down the incoming Democratic majority’s agenda for artificial intelligence, from federal preemption and bipartisan safety standards to the pushback against frontier AI research pauses. He also outlines why mandatory kill switches remain technically unworkable and how Congress can leverage hyperscalers to rebuild the U.S. power grid while protecting local communities...

fal co-founder and CTO Gorkem Yurtseven breaks down the explosive demand driving their $1.65B raise at a $16B valuation. He explains why frontier LLMs are rapidly converging with 3D gaming and VFX pipelines, and shares his prediction for when we will see the first AI-assisted, feature-length animated film. Check out our sponsor Lovable: https://lovable.dev. Turn ideas into software people love. --...

MongoDB's head of AI platform Chirag Shah breaks down why production agents fail without dedicated memory and context architecture. He explains how MongoDB built Atlas Agent Engine, why teams are over-indexing on LLM selection, and how multi-agent architectures are replacing traditional full-stack software. Check out our sponsor Lovable: https://lovable.dev. Turn ideas into software people lov...

Former DOJ antitrust leadership Jonathan Kanter and Doha Mekki discuss exiting the Justice Department to launch FluxCap, an AI platform designed for complex litigation. They break down the shift from standard legal tech to specialized investigative AI, the post-election regulatory landscape, and why prosecutors must learn to interrogate algorithms. Check out our sponsor Lovable: https://lovable.de...

Former Anthropic pre-training researcher and Core Automation co-founder Rohan Anil explains how compute islands allow decentralized GPU clusters to train frontier AI models without colocation. He details why current scaling hits test-time learning bottlenecks, how butterfly mixing handles asynchronous training noise, and sets a 2027 timeline for models that continuously learn on the fly. Check out...

Baseten co-head of model training Charlie O'Neill breaks down why recursive self-improvement will be much weaker than fast-takeoff theories predict, pointing to real-world compute and RL bottlenecks. He also unpacks why automating machine learning research faces sharp diminishing returns and introduces Baseten's newly launched inference and agent platform tools. Check out our sponsor Lovab...

Ruby Justice Thelot analyzes how algorithmic curation and platform economics have fractured cultural monoculture, flattening mainstream pop while accelerating niche subcultures. He examines empirical data across 11 years of celebrity Wikipedia traffic, Billboard streaming dispersion, and the surging adult toy market to argue that monolithic monoculture is officially over. -- https://x.com/being_on...