Latent Space: The AI Engineer Podcast cover art

Latent Space: The AI Engineer Podcast

Latent Space: The AI Engineer Podcast

Written by: swyx + Alessio
Listen for free

About this listen

 The podcast by and for AI Engineers! In 2024, over 2 million readers and listeners came to Latent Space to hear about news, papers and interviews in Software 3.0. We cover Foundation Models changing every domain in Code Generation, Multimodality, AI Agents, GPU Infra and more, directly from the founders, builders, and thinkers involved in pushing the cutting edge. Striving to give you both the definitive take on the Current Thing down to the first introduction to the tech you'll be using in the next 3 months! We break news and exclusive interviews from OpenAI, Anthropic, Gemini, Meta (Soumith Chintala), Sierra (Bret Taylor), tiny (George Hotz), Databricks/MosaicML (Jon Frankle), Modular (Chris Lattner), Answer.ai (Jeremy Howard), et al. Full show notes always on https://latent.spaceLatent.Space Economics Leadership Management & Leadership
Episodes
  • Artificial Analysis: The Independent LLM Analysis House — with George Cameron and Micah Hill-Smith

    Artificial Analysis: The Independent LLM Analysis House — with George Cameron and Micah Hill-Smith
    Jan 9 2026
    don’t miss George’s AIE talk: https://www.youtube.com/watch?v=sRpqPgKeXNk —- From launching a side project in a Sydney basement to becoming the independent gold standard for AI benchmarking—trusted by developers, enterprises, and every major lab to navigate the exploding landscape of models, providers, and capabilities—George Cameron and Micah-Hill Smith have spent two years building Artificial Analysis into the platform that answers the questions no one else will: Which model is actually best for your use case? What are the real speed-cost trade-offs? And how open is "open" really? We discuss: The origin story: built as a side project in 2023 while Micah was building a legal AI assistant, launched publicly in January 2024, and went viral after Swyx's retweet Why they run evals themselves: labs prompt models differently, cherry-pick chain-of-thought examples (Google Gemini 1.0 Ultra used 32-shot prompts to beat GPT-4 on MMLU), and self-report inflated numbers The mystery shopper policy: they register accounts not on their own domain and run intelligence + performance benchmarks incognito to prevent labs from serving different models on private endpoints How they make money: enterprise benchmarking insights subscription (standardized reports on model deployment, serverless vs. managed vs. leasing chips) and private custom benchmarking for AI companies (no one pays to be on the public leaderboard) The Intelligence Index (V3): synthesizes 10 eval datasets (MMLU, GPQA, agentic benchmarks, long-context reasoning) into a single score, with 95% confidence intervals via repeated runs Omissions Index (hallucination rate): scores models from -100 to +100 (penalizing incorrect answers, rewarding \"I don't know\"), and Claude models lead with the lowest hallucination rates despite not always being the smartest GDP Val AA: their version of OpenAI's GDP-bench (44 white-collar tasks with spreadsheets, PDFs, PowerPoints), run through their Stirrup agent harness (up to 100 turns, code execution, web search, file system), graded by Gemini 3 Pro as an LLM judge (tested extensively, no self-preference bias) The Openness Index: scores models 0-18 on transparency of pre-training data, post-training data, methodology, training code, and licensing (AI2 OLMo 2 leads, followed by Nous Hermes and NVIDIA Nemotron) The smiling curve of AI costs: GPT-4-level intelligence is 100-1000x cheaper than at launch (thanks to smaller models like Amazon Nova), but frontier reasoning models in agentic workflows cost more than ever (sparsity, long context, multi-turn agents) Why sparsity might go way lower than 5%: GPT-4.5 is ~5% active, Gemini models might be ~3%, and Omissions Index accuracy correlates with total parameters (not active), suggesting massive sparse models are the future Token efficiency vs. turn efficiency: GPT-5 costs more per token but solves Tau-bench in fewer turns (cheaper overall), and models are getting better at using more tokens only when needed (5.1 Codex has tighter token distributions) V4 of the Intelligence Index coming soon: adding GDP Val AA, Critical Point, hallucination rate, and dropping some saturated benchmarks (human-eval-style coding is now trivial for small models) — Artificial Analysis Website: https://artificialanalysis.ai (https://artificialanalysis.ai (\"https://artificialanalysis.ai\")) George Cameron on X: https://x.com/grmcameron (https://x.com/grmcameron (\"https://x.com/grmcameron\")) Micah-Hill Smith on X: https://x.com/_micah_h (https://x.com/_micah_h (\"https://x.com/_micah_h\")) Chapters 00:00:00 Introduction: Full Circle Moment and Artificial Analysis Origins 00:01:08 Business Model: Independence and Revenue Streams 00:04:00 The Origin Story: From Legal AI to Benchmarking 00:07:00 Early Challenges: Cost, Methodology, and Independence 00:16:13 AI Grant and Moving to San Francisco 00:18:58 Evolution of the Intelligence Index: V1 to V3 00:27:55 New Benchmarks: Hallucination Rate and Omissions Index 00:33:19 Critical Point and Frontier Physics Problems 00:35:56 GDPVAL AA: Agentic Evaluation and Stirrup Harness 00:51:47 The Openness Index: Measuring Model Transparency 00:57:57 The Smiling Curve: Cost of Intelligence Paradox 01:04:00 Hardware Efficiency and Sparsity Trends 01:07:43 Reasoning vs Non-Reasoning: Token Efficiency Matters 01:10:47 Multimodal Benchmarking and Community Requests 01:14:50 Looking Ahead: V4 Intelligence Index and Beyond
    Show More Show Less
    1 hr and 18 mins
  • [NeurIPS Best Paper] 1000 Layer Networks for Self-Supervised RL — Kevin Wang et al, Princeton

    [NeurIPS Best Paper] 1000 Layer Networks for Self-Supervised RL — Kevin Wang et al, Princeton
    Jan 2 2026
    From undergraduate research seminars at Princeton to winning Best Paper award at NeurIPS 2025, Kevin Wang, Ishaan Javali, Michał Bortkiewicz, Tomasz Trzcinski, Benjamin Eysenbach defied conventional wisdom by scaling reinforcement learning networks to 1,000 layers deep—unlocking performance gains that the RL community thought impossible. We caught up with the team live at NeurIPS to dig into the story behind RL1000: why deep networks have worked in language and vision but failed in RL for over a decade (spoiler: it's not just about depth, it's about the objective), how they discovered that self-supervised RL (learning representations of states, actions, and future states via contrastive learning) scales where value-based methods collapse, the critical architectural tricks that made it work (residual connections, layer normalization, and a shift from regression to classification), why scaling depth is more parameter-efficient than scaling width (linear vs. quadratic growth), how Jax and GPU-accelerated environments let them collect hundreds of millions of transitions in hours (the data abundance that unlocked scaling in the first place), the "critical depth" phenomenon where performance doesn't just improve—it multiplies once you cross 15M+ transitions and add the right architectural components, why this isn't just "make networks bigger" but a fundamental shift in RL objectives (their code doesn't have a line saying "maximize rewards"—it's pure self-supervised representation learning), how deep teacher, shallow student distillation could unlock deployment at scale (train frontier capabilities with 1000 layers, distill down to efficient inference models), the robotics implications (goal-conditioned RL without human supervision or demonstrations, scaling architecture instead of scaling manual data collection), and their thesis that RL is finally ready to scale like language and vision—not by throwing compute at value functions, but by borrowing the self-supervised, representation-learning paradigms that made the rest of deep learning work. We discuss: The self-supervised RL objective: instead of learning value functions (noisy, biased, spurious), they learn representations where states along the same trajectory are pushed together, states along different trajectories are pushed apart—turning RL into a classification problem Why naive scaling failed: doubling depth degraded performance, doubling again with residual connections and layer norm suddenly skyrocketed performance in one environment—unlocking the "critical depth" phenomenon Scaling depth vs. width: depth grows parameters linearly, width grows quadratically—depth is more parameter-efficient and sample-efficient for the same performance The Jax + GPU-accelerated environments unlock: collecting thousands of trajectories in parallel meant data wasn't the bottleneck, and crossing 15M+ transitions was when deep networks really paid off The blurring of RL and self-supervised learning: their code doesn't maximize rewards directly, it's an actor-critic goal-conditioned RL algorithm, but the learning burden shifts to classification (cross-entropy loss, representation learning) instead of TD error regression Why scaling batch size unlocks at depth: traditional RL doesn't benefit from larger batches because networks are too small to exploit the signal, but once you scale depth, batch size becomes another effective scaling dimension — RL1000 Team (Princeton) 1000 Layer Networks for Self-Supervised RL: Scaling Depth Can Enable New Goal-Reaching Capabilities: https://openreview.net/forum?id=s0JVsx3bx1 Chapters 00:00:00 Introduction: Best Paper Award and NeurIPS Poster Experience 00:01:11 Team Introductions and Princeton Research Origins 00:03:35 The Deep Learning Anomaly: Why RL Stayed Shallow 00:04:35 Self-Supervised RL: A Different Approach to Scaling 00:05:13 The Breakthrough Moment: Residual Connections and Critical Depth 00:07:15 Architectural Choices: Borrowing from ResNets and Avoiding Vanishing Gradients 00:07:50 Clarifying the Paper: Not Just Big Networks, But Different Objectives 00:08:46 Blurring the Lines: RL Meets Self-Supervised Learning 00:09:44 From TD Errors to Classification: Why This Objective Scales 00:11:06 Architecture Details: Building on Braw and SymbaFowl 00:12:05 Robotics Applications: Goal-Conditioned RL Without Human Supervision 00:13:15 Efficiency Trade-offs: Depth vs Width and Parameter Scaling 00:15:48 JAX and GPU-Accelerated Environments: The Data Infrastructure 00:18:05 World Models and Next State Classification 00:22:37 Unlocking Batch Size Scaling Through Network Capacity 00:24:10 Compute Requirements: State-of-the-Art on a Single GPU 00:21:02 Future Directions: Distillation, VLMs, and Hierarchical Planning 00:27:15 Closing Thoughts: Challenging Conventional Wisdom in RL Scaling
    Show More Show Less
    28 mins
  • [State of Code Evals] After SWE-bench, Code Clash & SOTA Coding Benchmarks recap — John Yang

    [State of Code Evals] After SWE-bench, Code Clash & SOTA Coding Benchmarks recap — John Yang
    Dec 31 2025
    From creating SWE-bench in a Princeton basement to shipping CodeClash, SWE-bench Multimodal, and SWE-bench Multilingual, John Yang has spent the last year and a half watching his benchmark become the de facto standard for evaluating AI coding agents—trusted by Cognition (Devin), OpenAI, Anthropic, and every major lab racing to solve software engineering at scale. We caught up with John live at NeurIPS 2025 to dig into the state of code evals heading into 2026: why SWE-bench went from ignored (October 2023) to the industry standard after Devin's launch (and how Walden emailed him two weeks before the big reveal), how the benchmark evolved from Django-heavy to nine languages across 40 repos (JavaScript, Rust, Java, C, Ruby), why unit tests as verification are limiting and long-running agent tournaments might be the future (CodeClash: agents maintain codebases, compete in arenas, and iterate over multiple rounds), the proliferation of SWE-bench variants (SWE-bench Pro, SWE-bench Live, SWE-Efficiency, AlgoTune, SciCode) and how benchmark authors are now justifying their splits with curation techniques instead of just "more repos," why Tau-bench's "impossible tasks" controversy is actually a feature not a bug (intentionally including impossible tasks flags cheating), the tension between long autonomy (5-hour runs) vs. interactivity (Cognition's emphasis on fast back-and-forth), how Terminal-bench unlocked creativity by letting PhD students and non-coders design environments beyond GitHub issues and PRs, the academic data problem (companies like Cognition and Cursor have rich user interaction data, academics need user simulators or compelling products like LMArena to get similar signal), and his vision for CodeClash as a testbed for human-AI collaboration—freeze model capability, vary the collaboration setup (solo agent, multi-agent, human+agent), and measure how interaction patterns change as models climb the ladder from code completion to full codebase reasoning. We discuss: John's path: Princeton → SWE-bench (October 2023) → Stanford PhD with Diyi Yang and the Iris Group, focusing on code evals, human-AI collaboration, and long-running agent benchmarks The SWE-bench origin story: released October 2023, mostly ignored until Cognition's Devin launch kicked off the arms race (Walden emailed John two weeks before: "we have a good number") SWE-bench Verified: the curated, high-quality split that became the standard for serious evals SWE-bench Multimodal and Multilingual: nine languages (JavaScript, Rust, Java, C, Ruby) across 40 repos, moving beyond the Django-heavy original distribution The SWE-bench Pro controversy: independent authors used the "SWE-bench" name without John's blessing, but he's okay with it ("congrats to them, it's a great benchmark") CodeClash: John's new benchmark for long-horizon development—agents maintain their own codebases, edit and improve them each round, then compete in arenas (programming games like Halite, economic tasks like GDP optimization) SWE-Efficiency (Jeffrey Maugh, John's high school classmate): optimize code for speed without changing behavior (parallelization, SIMD operations) AlgoTune, SciCode, Terminal-bench, Tau-bench, SecBench, SRE-bench: the Cambrian explosion of code evals, each diving into different domains (security, SRE, science, user simulation) The Tau-bench "impossible tasks" debate: some tasks are underspecified or impossible, but John thinks that's actually a feature (flags cheating if you score above 75%) Cognition's research focus: codebase understanding (retrieval++), helping humans understand their own codebases, and automatic context engineering for LLMs (research sub-agents) The vision: CodeClash as a testbed for human-AI collaboration—vary the setup (solo agent, multi-agent, human+agent), freeze model capability, and measure how interaction changes as models improve — John Yang SWE-bench: https://www.swebench.com X: https://x.com/jyangballin Chapters 00:00:00 Introduction: John Yang on SWE-bench and Code Evaluations 00:00:31 SWE-bench Origins and Devon's Impact on the Coding Agent Arms Race 00:01:09 SWE-bench Ecosystem: Verified, Pro, Multimodal, and Multilingual Variants 00:02:17 Moving Beyond Django: Diversifying Code Evaluation Repositories 00:03:08 Code Clash: Long-Horizon Development Through Programming Tournaments 00:04:41 From Halite to Economic Value: Designing Competitive Coding Arenas 00:06:04 Ofir's Lab: SWE-ficiency, AlgoTune, and SciCode for Scientific Computing 00:07:52 The Benchmark Landscape: TAU-bench, Terminal-bench, and User Simulation 00:09:20 The Impossible Task Debate: Refusals, Ambiguity, and Benchmark Integrity 00:12:32 The Future of Code Evals: Long Autonomy vs Human-AI Collaboration 00:14:37 Call to Action: User Interaction Data and Codebase Understanding Research
    Show More Show Less
    18 mins
No reviews yet