The wiki
A field guide to the people, organizations, and ideas in our reporting.
From Coding Agents to Inference Economics — Part 2 of 3
Part 2 follows the work of making AI useful: turning intent into tests, choosing long context or retrieval, and separating fast responses from sustainable inference.
From Computer Use to Agents in Production — 2024 Year in Review, Part 3 of 3
The final part of Latent Space’s 2024 review examines what keeps agents from everyday use, revisits the year’s lasting technical lessons, and sets out the production test facing 2025.
From Denoising Images to Inferring 3D Worlds — Part 3 of 6
Part 3 of this ICML 2024 reading edition explains diffusion’s training and sampling mechanics, follows DreamFusion through CAT3D, and introduces Ricky TQ Chen’s transport-based account of flow matching.
From Easy GPU Access to Disaggregated Inference — Part 1 of 3
Part 1 follows Brev’s approach to GPU access, NVIDIA’s “speed of light” planning, and Dynamo’s controls for serving models across machines.
From enterprise search to the quote desk
What Glean, Windsurf, and Roadrunner reveal about building enterprise AI: permissions, customer trust, sales leadership, and the commercial rules behind every quote.
From GPU Scarcity to AI That Works on Your Data — Part 3 of 3
Part 3 closes the hardware discussion with cloud economics, chip competition and overordering, then follows Milind’s work from video understanding to applied AI at Capital One.
From Image Captioning to Editable Worlds
Marble turns text and images into editable 3D scenes. Its camera control makes that useful today; learning how worlds behave remains a larger challenge.
From impressive outputs to practical AI
A July 2023 conversation between Latent Space and Practical AI explores what turns model capabilities into useful systems: operations, evaluation, local data, customization and user experience.
From Llama’s training budget to the compute an agent needs
How model size, data quality, human preferences, and tool use shaped Llama—and the research still needed to make agents more capable.
From Memory-Price Forecasts to the Future of Analyst Software — Part 2 of 3
Agents can assemble financial research faster than an analyst can, but judgment, market regimes, and trusted data still matter. Part 2 of 3 follows a forecasting experiment into the economics of AI infrastructure and Microsoft’s software dilemma.
From MLOps to LLMOps: The Data Work Behind the Model
Fast experiments, reproducible trials, changing inputs, and production feedback belong in the same engineering workflow.
From o1’s Thinking Budget to Working Applications — Part 3 of 3
Part 3 of 3 follows the closing Q&A, research AMA recap, and two demos: reasoning limits, immigration drafting, and a prototype that turns o1’s plans into executable code.
From On-Policy Learning to IMO Gold: How Much Can One Model Learn?
Learning from rewarded attempts, building a competition checkpoint, and confronting the gaps in planning, data efficiency, and research judgment.
From Persistent Pixels to Playable Worlds — ICML 2024, Part 1 of 6
Sora explores what video generation can learn about the world. Genie adds frame-by-frame actions. Part 1 follows their mechanisms, demonstrations, and limits on the path toward agent training.
From Probability Flows to Reusable Representations — ICML 2024, Part 4 of 6
Part 4 follows flow matching into crystals and discrete tokens, then examines efficient image synthesis, controllable speech, synthetic pretraining data, and DeCAF’s legacy in computer vision.
From reasoning benchmarks to the whole software engineering job
Why reinforcement learning’s uneven generalization makes workflow context, product design, and learning from mistakes central to Cursor’s ambitions.
From Reusable Vision Features to Actions Learned from Video — Part 5 of 6
Part 5 of 6 follows DeCAF’s transfer-learning legacy, Lucas Beyer’s case for captioning and reward tuning, Ashley Edwards’s path from robot imitation to Genie, and Seung Jae Lee’s introduction to tokenized behavior generation.
From scaling models to building reliable agents
David Luan’s account of OpenAI, Google and Adept connects research organization to a practical challenge: making models operate software reliably and learn from the people they help.
From the Reversal Curse to DSPy: ICLR 2024, Part 5 of 5
Part 5 of 5 examines why reversing a learned fact differs from reversing one supplied in a prompt, then explores how DSPy optimizes language-model programs—and why it still needs a useful evaluation metric.
From Transformer Experiments to Unreliable GPU Clusters — Part 1 of 3
Part 1 of Latent Space’s conversation with Yi Tay follows the path from general-purpose model research to Reka, where research judgment meets the difficulty of keeping rented compute running.
From Video Models to Video Agents
Building video models, preserving history, and combining generation with reasoning and editing tools: a reading edition of the Latent Space conversation with Ethan He.
Gabriele Corso
Boltz co-founder and CEO whose research spans molecular docking, generalization to unfamiliar protein pockets, and open biomolecular interaction models.
Galileo Labs
Galileo develops evaluation models, observability tools, and guardrails for generative AI applications and agents.
Garry Tan
Y Combinator’s president and CEO advocates AI-native companies and shares gstack, his workflow for organizing AI-assisted software development.
Gavriel Cohen
Creator of NanoClaw and co-founder and CEO of NanoCo, building AI agents around scoped access, container isolation and a customizable core.
GDM
The organization formed from DeepMind and Google Brain, developing Gemini models and AI systems for scientific research.
Gemini in 2025: From Thinking Budgets to Real-Time Voice
Controlling reasoning is one part of building with Gemini. Live voice adds session limits, tool execution, conversational timing and the decision to stay silent.
General Intuition: From Gameplay Highlights to Playable Worlds
Medal’s clips pair what players saw with what they did. General Intuition aims to use that experience to teach agents to act, then reconstruct situations where they can learn through interaction.
Genesis Molecular AI
An AI drug-discovery company combining molecular generation, physical simulation and protein–ligand structure prediction for its own pipeline and pharmaceutical partnerships.
Geoffrey Huntley
Creator of the Ralph Loop, coding-agent educator, and advocate for autonomous software development guided by senior engineering judgment.
Public beta. Model source and teaching reviews passed; human editorial review is pending. Original articles and recordings remain separate.