The wiki

Thinking Machines Lab develops model-customization infrastructure, real-time interaction models, and the open-weight Inkling family.

AI researcher whose work spans Llama post-training, learning from human preferences, tool-using language models and practical agent evaluation.

Positron AI co-founder focused on the economics of transformer inference, memory-system design, and shipping adaptable hardware before committing to custom silicon.

Hugging Face co-founder and Chief Science Officer whose work spans open-source machine-learning tools and the question of how AI might produce new scientific knowledge.

Developer relations engineer focused on AI audio, voice interfaces, and practical examples that help developers ship applications.

Amp co-founder and co-creator, software engineer, and author who explains coding agents through working implementations and accounts of model-driven engineering decisions.

Machine learning systems researcher behind XGBoost, Apache TVM and MLC-LLM, and co-creator of Apache MXNet.

Researcher behind bitsandbytes and coauthor of QLoRA, whose work spans memory-efficient language models and open coding agents.

Synthetic children’s stories make coherence, elementary reasoning and model internals accessible to small experiments. Which lessons transfer to larger models remains an open question.

Steve Ruiz’s canvas company grew from digital ink and an open-source whiteboard into a developer SDK, an AI prototyping environment, and tools for agents to draw and script.

Good-looking arrows, digital ink, and reusable canvas interactions led to a workspace for expressing intent—and revising what a model produces.

Thursday Nights in AI is a San Francisco fireside-chat and interview series co-hosted by Outset Capital and Imbue, covering AI products, research, and engineering.

An AI cloud company combining open-model research, inference, fine-tuning and customer-owned model development with distributed GPU infrastructure.

Together connects data selection, model architecture and GPU systems to lower the cost of building and serving models. Its goal of 5,000 tokens per second reaches beyond faster chat.

AutoGPT’s founder described using programming techniques from games to chain language-model predictions, and advocated open source as a way to broaden access to AI.

The engineering behind reliable clusters, predictable training runs, and evaluations that measure useful coding capabilities.

Gradient’s context-extension work combines positional scaling, distributed training and carefully constructed data. The harder test is whether a model can use the extra information.

How Hugging Face’s M4 team connected pretrained models, built interleaved web data, debugged unstable attention and planned a smaller successor.

How training budgets, numerical precision, memory and communication shape the practical work of running large language models.

The FlashAttention and Mamba researcher designs machine-learning algorithms around GPU memory, parallel execution, and the cost of generating tokens.

A search infrastructure company built around object storage, with vector and full-text retrieval for AI applications.

A Canadian search company connecting large datasets to AI through vector and full-text retrieval built on object storage.

The University of California, Berkeley develops AI infrastructure and data systems through collaborative research labs, including work on efficient model serving, cloud computing and LLM-powered data processing.

Shared inventorship in Intuit-assigned automated prompt improvement, financial AI, and public views on domain expertise and data.

A Singapore-founded UI-testing company whose low-code framework expanded into AI-assisted authoring and agentic test creation.

Redpoint’s AI content hub combines practitioner interviews, deployment analysis and surveys of AI builders.

AI engineer at Google Labs whose NotebookLM work spans conversational audio, visual overviews and an agentic research workspace.

Co-founder of Exafunction and leader of Codeium’s development into Windsurf, whose work spans GPU infrastructure, coding models and agentic developer tools.

Cofounder of Exafunction and Codeium who helped lead the move from GPU infrastructure to agentic coding in Windsurf, then joined Google DeepMind.

From one-command deployment to v0 and durable agents, Vercel builds tools for creating and operating web software.