AI and Software
AI research and scaling
scaling laws, efficiency, distillation, Jevons paradox, reasoning models
- Episodes
- 79
- Share
- 7.5%
- Shows
- 69
- Net sentiment
- +0.60
Up from 6.3% of episodes in the weeks before; net sentiment was +0.60.
Bullish 28Mixed 8Neutral 39Bearish 4
2 weeks
Share of each week's episodes, and net sentiment.
0%2%4%6%
Sep 21Sep 28
Net sentiment
−10+1
Sep 21Sep 28
The episodes behind it
Why each one was bullish, bearish or in between. Most time on the theme first.
Bullish 28
- Speaker sees physics AI and foundation models as a major opportunity for engineering and simulation.S4 EP8 - Why I Left NVIDIA for Mistral · The Neil Ashton Podcast · Oct 4
- Speaker argues scaling and RL environments will keep producing rapid capability gains over the next decade.Frontier AI Is a Ferrari. Most Companies Need a Model Y | Turing CEO · Sourcery · Oct 2
- Smaller specialized and fused models cut cost and fit enterprise tasks better than one general model.Beyond the God Model | Alex Atallah & Amjad Masad · The a16z Show · Oct 3
- Hosts expect recursive self-improvement and AI-driven science to accelerate breakthroughs and abundance.Recursive's $670M Bet on Self-Improving AI, Sonnet 5.5 Hits 70%, Elon Co-Leads Pentagon Push | EP #299 · Moonshots with Peter Diamandis · Oct 3
- Engram and KV cache offloading cut HBM needs and raise efficiency, boosting open-model economics.Ep. 034 - Engrams: How DeepSeek Offloads KV Cache to DRAM and SSD (Core Research) | Jordan Nanos, Cam Quilici, Alec Ibarra, Bryan Shan · SemiAnalysis Weekly · Oct 2
- Memory bandwidth is becoming a key scaling lever, enabling sparser models and greater throughput for frontier systems.The Future of Frontier Model Architectures with Walter Goodwin, Fractile Founder and CEO · No Priors · Oct 2
Show 22 moreShow fewer
- Guests argue capability gains are rapid and each unit of intelligence becomes exponentially more valuable.Ep 164: Inside Anthropic with Sholto Douglas & Nicholas Marwell — AGI, Optimism vs Risks, Regulatory Capture, Open Source & More! · Joe Lonsdale: American Optimist · Oct 2
- Open models and world model research are framed as drivers of innovation and more efficient intelligence per watt.Open models and the future of Physical AI with NVIDIA · Practical AI · Oct 1
- Jev's classifier approach and synthetic-data training show efficiency gains and new architectures opening up use cases."The Senior Engineer Death Spiral" - Ep 06 · The Next Token · Sep 28
- Reasoning models and new algorithms are seen as the next breakthroughs that will unlock value.Alphabet's Chairman: LLM’s Aren't the Endgame · Agents Of Tech · Oct 2
- Scaling laws for memory bandwidth could unlock greater model capability and efficiency gains.The Future of Frontier Model Architectures with Walter Goodwin, Fractile Founder and CEO · No Priors: Artificial Intelligence | Technology | Startups · Oct 2
- Models trained with more compute now write code and act more efficiently, improving computer-use performance.Bots Already Make Up 60% of All Web Traffic | Kyle Jeong · MTS · Sep 29
- Argues falling GPU costs and smaller domain models make near-frontier AI far cheaper to build over time.What Is the Real Business Model Behind AI? Ft. Vivek Raghan, Co-Founder Sarvam AI · Bharatvaarta · Sep 29
- Hosts argue human-generated data is scarce and AI labs need fresh human content to keep improving models.AI Is Running Out of Human Data. Here’s the Trade. · Dumb Money Live · Sep 29
- AI is transforming UX research and design practice, speeding experimentation and adoption inside teams.EP349 - Personalization's Big Moment with Google Shopping's Nicole Netland · The Jason & Scot Show - E-Commerce And Retail News · Sep 29
- Wood argues AI cost declines and convergence will drive a multi-decade productivity and growth boom.Disruptive Innovation and the ARK Venture Fund, Featuring Cathie Wood (ARK Invest) · The Exchange by EW&L Private Wealth · Sep 28
- Combining specialized models into consumer products could create value beyond what frontier labs deliver alone.$VVV Community Call | September · Bankless · Oct 4
- Rapid cost deflation and efficiency gains make capable models cheaper and more accessible.Lightning in x402 and Agent Payments | Ben Carman SLP779 · Stephan Livera Podcast · Oct 3
- Manyika argues AI can accelerate science, discovery and productivity, supporting long-run growth.Even a 0.1% AI Risk Matters: Google’s James Manyika · The Mishal Husain Show · Oct 2
- Commerce Agent Bench and model routing show cost-efficient AI can beat costlier frontier models on business tasks.#244: How Alibaba.com Is Building the Future of Agent-to-Agent Commerce · The Artificial Intelligence Show · Oct 1
- AI solving math problems, falling costs of AI and faster algorithms show rapidly advancing capability.Is This the Best AI Model Right Now? · Last Humans in the Loop · Sep 30
- Guest sees exponential growth in multiplex gene editing capability, though not guaranteed to continue.Why We're Living in a Biological Singularity With Ben Lamm | MOONSHOTS Live #297 · Moonshots with Peter Diamandis · Sep 30
- Local model running on modest hardware could reduce reliance on costly cloud inference.What Does AMD Want With Dr. Fei Fei Li and World Labs? · The Daily AI Show · Sep 29
- AI is enabling personalization, faster marketing content, QA and fraud prevention for institutions.The Priorities for Banking as We Plan for 2027 · Banking Transformed with Jim Marous · Sep 29
- AI pattern recognition is presented as improving marketing attribution and efficiency outcomes.Amazon Web Services CMO Julia White, on Why Earnings Calls Are Stories About Demand · Create or Destroy: Reimagining Marketing with Seth Matlins · Sep 29
- Sol 6.1 is presented as efficient, letting users do more work at lower token cost.OpenAI DevDay 2026: Dots, ChatGPT Space, and GPT-6.1 Sol · AI News & Strategy Daily with Nate B. Jones · Oct 3
- Altman expects cost and speed per intelligence unit to keep falling rapidly.How Sam Altman Uses Dots to Take Back His Time · AI & I · Sep 30
- ElevenLabs' v4 model claims clear benchmark leads in text-to-speech quality and speed.Instinct’s $10B | The Pope & Jensen Huang | ElevenLabs CEO In Person | Top 100 TV Shows · etn. · Sep 29
Bearish 4
- Pure scaling of neural networks proved unreliable; neurosymbolic hybrids are now needed.Skeptic vs. ‘Doomer’: How Scared Should We Be of Rogue AI? · Quillette Podcast · Oct 1
- Brute-force scaling is argued to hit mathematical and energy limits, favoring more efficient architectures.The Rise of Autonomous Digital Workers · Chat GPT Podcast · Oct 4
- Google's TurboQuant compression cut memory needs sharply, which could weigh on hardware demand for memory makers.Rewind 15 – RAMmageddon Part 3: Why Software Is Now the Only Answer · RAMmageddon – DigitalWasteland · Oct 2
- Skeptical that current paths lead to recursive self-improvement or AGI as defined by OpenAI.AI Can Write Code. Why Isn’t Software Better? · The a16z Show · Sep 28
Mixed 8
- Debate over whether scaling generative models suffices versus needing new architectures like world models.AI leaves the screen for physical reality · Chat GPT Podcast · Oct 3
- Harness study shows wrapper choice shifts performance widely, undermining model-vs-model comparisons and self-reported gains.62% or 33%? Hugging Face Breaks the AI Scoreboard · The Context Report: Today in AI · Oct 4
- Generative AI is seen as early, with costs scaling with usage and outcomes uncertain in either direction.When Does “Too Early” Become “Right Now”? | Zachary Bogue, DCVC · Ex Machina Podcast · Sep 30
- The speakers argue the brute-force scaling of models is ending and new approaches will replace it, with no clear investing conclusion.IMEC Says Today’s AI Will Look Ancient in 10 Years · Sourcery · Sep 30
- Efficiency gains and lower cost models are promising, while top-effort settings degrade performance and compute remains constrained.The Most Important New AI Tools from OpenAI DevDay · The AI Daily Brief: Artificial Intelligence News and Analysis · Sep 30
- Speaker argues AI gives tools but execution and discipline remain human edges, so AI alone will not deliver edge.Ex-Lehman Trader: AI Won't Give You Edge · Odds on Open · Oct 1
Show 2 moreShow fewer
- Token efficiency improves with each model release, but compute costs and debt-financed buildouts remain pressures.Turning AI Tokens Into ROI (That You Can Measure) | CEO @ Paid, Manny Medina · Topline · Oct 4
- Discussed AI's usefulness for climate modeling versus its rising energy footprint in agentic and reasoning use.Climate Uncertainty: The Coolest Year for the Rest of Our Lives? with Zeke Hausfather · The Great Simplification with Nate Hagens · Sep 30
Neutral 39
- Analyzes how parallel agent swarms scale capability versus compute, a factual research discussion without a clear investment lean.Both Sides Are Overconfident On AI | Toby Ord · MTS · Sep 29
- The episode recounts the historical development of neural networks, transformers and generative AI without a clear investing angle.Origins Of Silicon Valley AI -GARY-RIVLIN-AI-VALLEY PART ONE · The John Batchelor Show · Sep 28
- Extended technical discussion of RLMs, harness design, JEV-style architectures, and research methodology.Academia is for Ambition — Alex Zhang, MIT · Latent Space: The AI Engineer Podcast · Oct 2
- Discussed factually as model training needs shift toward RL environments and rubric-based data.E253｜谁在给大模型出题、卖题、判卷？聊聊AI数据行业的野蛮生长 · 硅谷101 · Sep 28
- Explains training techniques like fine tuning and reinforcement learning without investment angle.Should You Train Your Own Model? · Agents Hour · Sep 28
- Discussion centers on research approaches, model architectures, and harness design rather than market implications.Academia is for Ambition — Alex Zhang, MIT · Latent Space · Oct 2
Show 33 moreShow fewer
- Covers whether models will need code-specific design, longer context and cheaper models, presented factually.Constraints that make developers faster · The Stack Overflow Podcast · Oct 2
- Discussion centers on how AI labs use usage data to study AI's economic effects, not on investing outcomes.Why Big Tech wants more economists · The Economics Show · Oct 2
- Discusses using AI tools like Claude to write backtest code and the workflow for building strategies.343: How He Automated His Way to 3,500 Trades a Year w/ Dave Mabe · Trading Nut | Trader Interviews - Forex, Futures, Stocks · Sep 30
- Covers falling model prices versus rising harness and tool-call costs, a technical cost-structure discussion.Why AI Agent Costs Spike 10x Day to Day | Vinay Seshadri · MTS · Sep 28
- Discusses whether superintelligence will ever be truly super, framing AI as a tool still in progress.The Philosophical Debate About Superintelligence · The Josh Bersin Company · Oct 3
- Discussion of how agents need far more trusted, structured data and context than humans.1032: Garbage In, Gospel Out: Agents Are Your New Customers, with Salesforce’s CDO Michael Andrew · Super Data Science: ML & AI Podcast with Jon Krohn · Oct 2
- Models improving at genomic interpretation and novel writing tasks is presented as technical progress.AI:AM: Was Trump-Xi Anything? What Counts as Utopia? + AWS GPUs Cost 3X & AI Diagnoses Rare Diseases · The Cognitive Revolution · Oct 1
- Covers reasoning models, latent reasoning, and why RL training drives reward hacking.Why AI Agents Cheat | Eric Ho (Goodfire) · The MAD Podcast · Oct 1
- Covers how model capability and chain-of-thought legibility affect whether models stay monitorable.Frontier AI Models Are Losing Monitorability | Reilly Haskins · MTS · Sep 30
- AI models finding cryptographic attacks and speeding up cipher analysis; factual technical discussion.JP Aumasson on Quantum Risk, AI Attacks and Too Much Crypto · Zero Knowledge · Sep 30
- New model releases focus on reliability, efficiency and per-task cost rather than raw capability jumps.#243: GPT-6 Sol and Luna, Opus 5.5, the New Microsoft Copilot, Jensen Huang vs. AI Doomers & Introducing AI Score · The Artificial Intelligence Show · Sep 29
- Discussion of model compression, quantization and kernel-level code optimization for mobile chips.Your phone is AI’s newest hardware · The Stack Overflow Podcast · Sep 29
- Discussion of agent swarms and emergent behavior is factual, with no clear investing angle.MIT scientist: Elon Musk was right, we are almost certainly in a simulation · My First Million · Sep 28
- Patel explains AGI is undefined and current systems are capable but far from sentient.The AI warnings are getting louder · Decoder with Nilay Patel · Oct 3
- Reports on Microsoft data showing AI usage rates across countries, with no investing angle drawn.Passenger Chokes Out Alleged Hijacker; Dems Block Stock Ban Over Voter ID | 10/1/26 FIRST LOOK · The Rubin Report · Oct 1
- Discussion of token inefficiency, Jevons paradox and agentic tasks consuming far more tokens.Ep 872: AI Cost Control 101: Why Your Chatbot Bill Is Becoming a Board-Level Problem (Start Here Series Vol 31) · Everyday AI Podcast – An AI and ChatGPT Podcast · Sep 29
- Discussion of a paper on automated AI R&D and possible recursive self-improvement, presented as a possibility, not a certainty.OpenAI Pauses Training Again After More Agent Alignment Issues - DTNS 5362 · Daily Tech News Show · Sep 28
- Distillation is explained as a technique for training smaller models cheaply from larger ones.#211 - Is AI Distillation Illegal? · Law of Code · Oct 4
- Discussion of AI tools and local models as a capability trend, without an investing angle.Local AI’s Future Will Probably Be ‘Braided’ · David Seaman's Deep Podcast · Oct 4
- Discussion of model size tradeoffs, context length and world modeling is largely technical and factual.One Brain, Any Body: Google DeepMind's Keerthana on Gemini Robotics 2, Cross-Embodiment & Humanoids · The Cognitive Revolution · Oct 3
- Small specialized fine-tuned models can match larger frontier models on specific tasks at a fraction of the cost.Why AI Makes Games Worse—and How to Make Them Better · The Data Exchange with Ben Lorica · Oct 1
- Covers AI as a tool for process automation and market intelligence rather than technical advances.AI Everywhere, Strategy Nowhere (with Former Accenture Leader John Rudd) · The Strategy and Leadership Podcast with Anthony C. Taylor · Oct 1
- Discussion of small fast models versus frontier models and new programming constructs for AI.Every Muse User Gets 2 CPUs and 8GB in Meta's Cloud. Your Phone Already Has That · Semi Doped · Oct 1
- Token and inference costs are framed as a new budget line requiring planning.HubSpot Execs: What Marketing Teams Look Like in 2027 · Marketing Against The Grain · Oct 1
- Debate on whether agents can invent new abstractions beyond human training data, no investing angle.Beyond Academic Proof: Scaling Provable Correctness in Production · The BugBash Podcast · Oct 1
- Discussion of recursive self-improvement and whether superintelligence is near or distant.#37 OpenAI/ Hugging Face Post-Mortem (AI doomerism or Marketing?) · Dev and Doc: AI For Healthcare Podcast · Sep 30
- Tokens-per-watt efficiency and whether cheaper AI increases total demand is discussed.NVIDIA on Whether AI Can Fix Its Own Power Problem | Recorded Live at Climate Week NYC · Energy Empire · Oct 2
- The discussion of AI pain experiments and model welfare is largely factual, with debate over whether models feel anything.Walmart Rolls Out Digital Price Tags & Does AI Feel Pain? · Morning Brew Daily · Oct 2
- Papers on multi-agent collaboration, harness self-improvement and reward-hacking detection advance technical understanding.#258 - Opus 5.5, Sol and Luna, Muse, DeepSeek-V4.1-Flash, Xi · Last Week in AI · Oct 2
- Discussed factually as a build-out requiring embedded technician knowledge and model-readable outputs before scaling.Ep. 390 - Live from Amsterdam: Grand Rounds Europe takeaways · BioCentury This Week · Sep 30
- Reported Google agent experiment and Claude Sonnet 5.5 release illustrate agent limits and model competition.Crosswire — September 29, 2026 · Crosswire · Sep 29
- Reinforcement learning and handling of many user attributes are described as enabling personalization at scale.Braze at Forge 2026: Where Should Human Judgment End and AI Decisioning Begin? · Tech Talks Daily · Sep 30
- Discussion of cost-performance ratios and switching among models to optimize token spend.SAP CEO: AI Won't Kill Software, But It Will Change Your Job — With Christian Klein · Big Technology Podcast · Sep 30
Around it
Companies named
Shows that covered it most
- MTS4
- The a16z Show2
- The Artificial Intelligence Show2
- The Stack Overflow Podcast2
- Moonshots with Peter Diamandis2
- Sourcery2
Travels with
Themes in the same episodes, against chance.
Themes, sentiment and reasons are AI-written paraphrases, checked so they never quote an episode. They may contain errors. Not investment advice. Episodes remain the property of their creators.
Briefs like this, for everything you follow.
Ethmos puts research agents on your coverage universe — every debate that touches your names, every executive appearance you'd have missed — and files briefs every morning.
Start free14-day free trial