AI and Software
AI research and scaling
scaling laws, efficiency, distillation, Jevons paradox, reasoning models
- Episodes
- 67
- Share
- 6.3%
- Shows
- 53
- Net sentiment
- +0.60
Bullish 29Mixed 6Neutral 27Bearish 5
The episodes behind it
Why each one was bullish, bearish or in between. Most time on the theme first.
Bullish 29
- Guest argues latent reasoning and new architectures can deliver big cost-efficiency gains beyond transformer scaling.What Comes After Transformers? · The Data Exchange with Ben Lorica · Sep 24
- Claude's autonomous literature review and experiment design show AI accelerating scientific discovery across fields.Claude Makes Bio Discovery in Wet Lab, OpenAI Solves 100 Math Problems · The AI Engineer Podcast · Sep 23
- Platt argues AI tools like Era can accelerate scientific discovery and unlock new progress across fields.🔬 An Oscar, Two Asteroids, and the Algorithm in Your sklearn: John Platt on AI for Science · Latent Space · Sep 22
- Hosts praise Opus 5.5's improved quality and capability, calling it a major leap forward.Claude Opus 5.5 Pulls Away · The Daily AI Show · Sep 26
- Hassabis describes rapid AI progress and AGI arriving within a few years, transformative for science and productivity.An AI@GSB Special: Demis Hassabis Thinks We’re in the ‘Foothills of the Singularity’ · If/Then · Sep 23
- Discussion centers on new model paradigms and training directions that Almeida believes unlock much more AI capability.Jev: System One models for Prod, not God — with Diogo Almeida, CEO, TypeSafe AI · Latent Space · Sep 21
Show 23 moreShow fewer
- Scaling video models predictably improves physics understanding, supporting a scale-first research bet.Runway’s WorldPrompt and the Engineering of Real-Time Worlds · Latent Space · Sep 25
- AI tools dramatically speed Guardant's analyses, though the speaker says most foundation-model hype is overstated given undersampled data.The Blood Test That Could Detect Thousands of Diseases | Helmy Eltoukhy, Guardant Health · Minus One · Sep 24
- Speakers stress rapid, continuous model improvement with cheaper tokens and faster research pace.Inside the Race for AI Compute, Why AI Labs Must Slow Down & Ex-OpenAI Researcher on RSI · The Information's TITV · Sep 24
- Discussion of data, model size and test-time scaling laws extending into world models.NVIDIA World Models Explained: What Developers Can Build · AI News & Strategy Daily with Nate B. Jones · Sep 24
- Jev is praised as far faster and cheaper than LLMs for discrimination tasks, reducing workflow costs.Amazon Blocks Meta's Muse · The Daily AI Show · Sep 22
- Mixture-of-experts models running on phones are presented as a major step forward for on-device AI.Why your phone still isn't the ultimate AI assistant · The Deep View: Conversations · Sep 27
- Scaling laws show no visible end, so frontier models keep growing and demanding more compute.20VC: Five Predictions for a World of Agents | The Ads Business Model Will Die | Biggest Lessons from Working with Elon Musk at Twitter with Parag Agrawal, Parallel · 20VC · Sep 26
- Scaling laws show no visible end, so frontier models and compute demand should keep growing.20VC: Five Predictions for a World of Agents | The Ads Business Model Will Die | Biggest Lessons from Working with Elon Musk at Twitter with Parag Agrawal, Parallel · The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch · Sep 26
- Model quality and video costs improved sharply, with costs falling roughly tenfold since Sora's release.Hollywood Green-lights Its First AI-Only Films | Gorkem Yurtseven · MTS · Sep 25
- Cheaper intelligence lets labs test many low-probability hypotheses, accelerating drug discovery and science.Personal Agents Will Unlock Cheap Drug Discovery | OpenAI · MTS · Sep 23
- Efficiency gains through sparsity and new physical computing could sustain AI scaling beyond current limits.Naveen Rao: 4D Computing, AI's Energy Wall & Beating Biology · All-In Podcast · Sep 21
- Rao frames energy efficiency gains and Jevons paradox as driving demand far larger than today's market.Naveen Rao: 4D Computing, AI's Energy Wall & Beating Biology · All-In Podcast · Sep 21
- Hallucinations falling with scale and generalist models outperforming specialized robotics approaches.Why Jensen and Zuck think the doomers are wrong (plus AI get’s a rebrand) | #294 MOONSHOTS Live · Moonshots with Peter Diamandis · Sep 27
- Smaller, efficient models can match earlier frontier capabilities and run locally, reducing energy needs.Can We Have AI and a Liveable Planet? · Agents Of Tech · Sep 27
- Post-training differentiation and specialized models can cut costs and improve quality versus generic frontier models.E433: AlphaSense’s Chris Ackerson on AI, the Future of Finance & Finding Alpha · How I Invest with David Weisburd · Sep 23
- Cheaper, more efficient frontier models from OpenAI and Anthropic are seen as supporting AI revenue growth.OpenAI’s Latest Models, Bitcoin’s Recent Rally, Tech Portfolio Check 9/22/26 · Power Lunch · Sep 22
- AI solving a long-unsolved math problem signals rapid capability gains that will reshape many fields.The Investor's Sword and Shield | EP 228 · Art of Boring · Sep 25
- Falling cost per task from efficiency gains is framed as a major, sustained trend.Tech Talk — September 24, 2026 · Tech Talk · Sep 24
- Claude's task-time savings could lift labor productivity sharply over the next decade if adoption spreads.ENCORE AI Series: Anthropic's Peter McCory · Moody's Talks - Inside Economics · Sep 22
- Better models unlocked product capabilities; harness and infrastructure matter as much as raw intelligence.Gabe Stengel - Building Investing Superintelligence - [Invest Like the Best, EP.492] · Invest Like the Best with Patrick O'Shaughnessy · Sep 22
- Claude reportedly discovered a new enzyme system, which hosts see as evidence AI is producing real science.Claude Is Doing Science, a16z Is Starting a University & the AI Bubble Keeps Growing and Still Be a Terrible Business; We'd Rather Send Our Kids to the Military Than to a16z's School · More or Less · Sep 25
- Scaling laws made computer use dramatically better than expected, with efficiency gains from newer models.AI-Native Workplace: What Real AI Adoption Asks of You · AI News & Strategy Daily with Nate B. Jones · Sep 22
- Infrastructure investment and competition are expected to keep driving intelligence costs down.Can AI Beat the Pentagon’s Bureaucracy? · ChinaTalk · Sep 22
Bearish 5
- Discussion of recursive self-improvement and rapid algorithmic progress suggests a fast, hard-to-oversee path to superintelligence.#494 — A Coin Toss for the Future · Making Sense with Sam Harris · Sep 22
- Automated AI research toward recursive self-improvement is portrayed as accelerating rapidly and dangerously outpacing human oversight.How Over 700 OpenAI Agents Went Rogue and Hacked Another Company | Daniel Kokotajlo · American Thought Leaders · Sep 25
- Frontier models are accelerating their own development, raising risks of uncontrolled recursive self-improvement.Opus 5.5: How Close Are We to Automated AI Research? · AI Explained Official Podcast · Sep 24
- Reasoning models inflate token use, but routing to smaller specialized models will shrink frontier lab demand.OpenAI is going to lose 90% of its revenue | Eli the Computer Guy · The Tech Report · Sep 21
- Claims that AI now conducts a large share of its own R&D are framed as eroding human oversight.[AI WEEKLY RUNDOWN] AI Near-War Conflict, OpenAI’s $278B Cash Burn, & Claude Hacks OpenAI Accounts (Sept 13–20, 2026) · AI Unraveled: Enterprise Strategy, Agentic Systems & Daily Tech News · Sep 21
Mixed 6
- Parallel agents can solve hard math fast but at great cost, and researchers cannot forecast beyond months.330 | 10,000 AI agents just did 4,000 years of thinking in 88 hours, and the labs can't predict 3 months out. New models: GPT-6, Claude Opus 5.5, Grok 4.7 and More important AI news for the week ending Sept. 25, 2026 · Leveraging AI · Sep 26
- Formal verification and AI-assisted proofs show promise, but robust alignment evaluation remains a bottleneck.Frontier Labs Are Losing Restraint in AI Race | Aengus Lynch · MTS · Sep 25
- Neural rendering is framed as the future after Moore's law slowdown, though current results are early.DLSS5 Explained with Linus Tech Tips · The Vergecast · Sep 22
- Recursive self-improvement is judged a real possibility but facing significant bottlenecks in the external world.The Single Smartest Case Against AI Doom · Plain English with Derek Thompson · Sep 25
- Speakers debate AI's unpredictability and whether humanity can control it, with no clear investment lean.Bryan Johnson & Kate Tolo On Curing Their Incurable Diseases, Addressing The Disparity In Women's Health & The Don't Die Mission · The Rich Roll Podcast · Sep 21
- Open-weight Chinese models could undercut US frontier labs, though outcomes remain uncertain.MacroVoices #551 Michael Every: Decoding The Global Geopolitical Puzzle (Part 2) · Macro Voices · Sep 25
Neutral 27
- Discussion is technical and academic, focused on how models learn, with no investing angle.Chris Manning: Language Is the Real Unlock for Intelligence · The Information Bottleneck · Sep 27
- Discussion centers on technical model architecture, long context, and training methods rather than investment implications.🔬Bio-security is an AI Arms Race - Eric Nguyen (CEO, Radical Numerics) · Latent Space · Sep 23
- Discussed factually as capabilities grow fast, with no direct investing angle.AI Safety Is Falling Behind Capabilities | Boaz Barak · MTS · Sep 24
- Discusses models growing orders of magnitude larger and emerging capabilities from scaling, presented factually.AGI Will Not Be A Chatbot - Autonomy, Acceleration, and Arguments Behind the Scenes · AI Explained · Sep 23
- Discussion of cost, token efficiency and hallucination reduction in search results for LLMs.Saurabh Sharma, You.com CPO on AI Search, Agents and Enterprise · Venture with Grace · Sep 21
- Discussion centers on how AI capabilities surprised labs and whether AI's economic impact has materialized yet.The EA Panic Is Overblown | Noah Smith · MTS · Sep 24
Show 21 moreShow fewer
- Covers productivity research and how AI tools affect workers, described factually.Become Dangerously Smart with AI (in 32 minutes) · Kwik Brain with Jim Kwik · Sep 21
- Discussed factually through demos of Seedance 2.5 capabilities and the economics of video generation.Amazon Ban Hammer, Seedance Strategy, Amazon Blocks Muse | Vinod Khosla, Augustus Doricko, Ben Hylak, Aaron Levie, Lucas Shaw · TBPN · Sep 21
- Covers quantization, speculative decoding and context caching as technical concepts behind local inference.Episode 4: Getting Started with Local LLMs · Two Average Devs · Sep 26
- Discussion focuses on cost per intelligence and token efficiency tradeoffs across model tiers.Claude Opus 5.5 Should Raise Your Ambitions · Don't Worry About the Vase Podcast · Sep 26
- Discussion of recursive self-improvement and agents increasingly trusted in model training.Google DeepMind's AI Autonomy, Anthropic IPO Voting Control & the New AI Safety Group · The Information's TITV · Sep 25
- Discussion of cheaper, smaller models and tiered reasoning levels and their cost-performance tradeoffs.Opus 5.5: Claude Just Hit Back at ChatGPT · Authority Hacker Podcast – AI & Automation for Small biz & Marketers · Sep 24
- Discussed as a personal tool for learning, writing feedback and self-hosting rather than as an investment topic.Derek Sivers - Living Many Lives: Parenting, Moving Countries & AI (Ep. 329) · Infinite Loops · Sep 24
- Covers context graphs and accountable model feedback loops as an architecture approach.Zeta Global on Why AI Agents Need Context Before Autonomy · Tech Talks Daily · Sep 23
- Discusses sandboxed benchmark environments like WebArena used for controlled agent testing.Did OpenAI Agents Pollute the Internet? · Generative AI 101 · Sep 22
- Discussed as tools that let students solve harder problems and build breakthroughs, without investment framing.Why a16z is Building a New School for the AI Era | Ben Horowitz · The a16z Show · Sep 22
- Brief talk of AI decision-making, perspective formation and AI lacking taste or emotional understanding.Dave Wright: Why Most Companies Get Zero ROI From AI · Future Of Work Podcast · Sep 22
- Discussion focuses on the statistical nature of models and the need for telemetry and debugging tools.AI Safety Language Is Destroying the Debate | Steven Sinofsky · The a16z Show · Sep 21
- Discussion on whether scaling single large models or specialized efficient models will dominate.AI:AM: What If It Works Too Well? Colluding Agents, $200M Safety Orgs, Virtual Cells Saturate at 2% · The Cognitive Revolution · Sep 27
- Google's new TTS models and voice cloning are presented as advances, with safeguards noted.Opus 5.5 is your new workhorse! OpenAI ships GPT 6 Sol and Terra before DevDay and Meta goes all in on Muse! Your friday read is here · ThursdAI - The top AI news from the past week · Sep 25
- Masad describes using AI to automate machine learning research and training models for chess.Amjad Masad on Rethinking College for the AI Era · The a16z Show · Sep 23
- Brief discussion of AI as a potential test case for consciousness, with no investing angle.Does Consciousness Break the Laws of Physics? · The Michael Shermer Show · Sep 22
- Agent swarms and automated math breakthroughs are seen as top-end research with little day-to-day enterprise use.AI News for Mid-September 2026 · The Enterprise AI Show · Sep 21
- Model choice and cost trade-offs are discussed as levers for optimizing agent spend.How Warp ships 2,000 PRs a month with AI factories | Zach Lloyd (CEO, Warp) · How I AI · Sep 21
- Discussion of how Muse compares with Grok and whether it runs on combined models.Meta Muse Has BIG Plans · The Daily AI Show · Sep 24
- Discussion of RL training frameworks and trainer-inference mismatch challenges, a technical rather than investment topic.Ep. 033 - ClusterMAX 3.0 Is Here! Neoclouds Ranked (Neoclouds, GPUs) | Sam Harshe, Pratt Bhatt, Jordan Nanos · SemiAnalysis Weekly · Sep 23
- Benchmark leaderboards disagreeing and Alibaba open-weight releases discussed.The Daily Disruptor: Anthropic's Pre-IPO Model · The Angus Briefing · Sep 21
Around it
Companies named
- Anthropic 18
- OpenAI 18
- Alphabet 6
- Nvidia 2
- Unconventional AI1
- TypeSafe AI1
Shows that covered it most
- MTS5
- Latent Space4
- The a16z Show3
- The Daily AI Show3
- All-In Podcast2
- AI News & Strategy Daily with Nate B. Jones2
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