The wiki

Retrieval-augmented generation connects a language model to external information. Its central engineering problem is choosing evidence that helps the model answer—and checking that the answer actually uses it.

Model Context Protocol gives AI applications a shared interface to external data and actions. It standardizes how applications request capabilities; builders decide how to use the returned context and which actions to allow.

Cursor is Anysphere's AI software development platform, combining an editor, coding models and agent execution environments. Its evolution shows how retrieval, tool use and reliable environments support coding agents.

Evals turn questions about AI quality into repeatable tests: what a system can do, where it fails, and whether a change makes it better.

Search engineer and AI investor, from Glean’s founding team and Glean Assistant to early-stage investing at Menlo Ventures.

The Stanford computer scientist and World Labs cofounder helped build ImageNet and its shared recognition benchmark. Her work also spans scene description, dataset fairness, and spatial intelligence.

AI researcher and educator whose work connects vision and language, practical neural-network engineering, and compact implementations that expose how models work.

Co-creator and lead maintainer of the Model Context Protocol, David Soria Parra builds shared infrastructure for connecting AI applications to tools and data.

The organization behind ChatGPT, the OpenAI API, and Codex has evolved from nonprofit research into large-scale model development, commercial deployment, and agent platforms.

The company behind Claude combines model development with alignment and interpretability research, coding tools, agent infrastructure, and evolving safeguards.

Andreessen Horowitz combines technology investment with operating support, experimental developer tools, and analysis of AI businesses across models, applications, and physical infrastructure.

Google connects AI research, custom computing infrastructure and widely used products, from Search to Gemini and Google Cloud.

A practical account of coding agents: how they use tools, recover from errors, reuse workflows, and turn human feedback into completed work.

Video generation, segmentation with memory, faster detectors, and the challenge of teaching small models to see precise details.

Statsbot’s early Slack interface exposed two problems: understanding a question and understanding the data. Cube’s semantic layer separated them—and foundation models reopened the conversational side.

Part 3 of 3: the closing exchange of Latent Space’s conversation on NVIDIA’s AI engineering.