The wiki

How instruction tuning, reward models, and DPO turn feedback into model updates—and why the resulting behavior remains difficult to predict.

A Kleiner Perkins incubation building AI-native enterprise quoting, with $27 million in announced financing and a Prompt, Quote, Approve workflow.

WebSim cofounder whose work spans generative browser experiences, tools for thought, and end-user programming.

Scrunch cofounder and CTO working on how AI agents retrieve, understand and use web content.

Runway principal research scientist developing real-time generative video, world models and robotics simulation.

Computer vision tools for preparing datasets, training models and turning image and video predictions into working applications.

AI researcher credited on Nomos, Hermes 4, and research into reducing communication during distributed model training.

Pinecone developer advocate whose technical writing examines how AI systems retrieve applicable knowledge, generate useful queries, and choose actions under explicit constraints.

Introspection co-founder focused on feedback loops, captured human expertise and infrastructure for self-improving agents.

OpenAI’s developer-experience leader on application architecture, real-time voice, and engineers working with agents.

Physician-scientist and Noetik co-founder focused on using multimodal foundation models to connect cancer biology, therapeutic targets and patient populations.

AI researcher with work on continual learning at Trajectory and contributions to Windsurf, SWE-1 and Google DeepMind’s agentic coding work.

Mathematician and coauthor of TinyStories and approximate-unlearning research, with earlier contributions to neural-network theory and high-dimensional probability.

Runway builds generative video tools and world models, moving from creative software toward interactive simulation.

Identified by Hg as Cognition’s founding president, Kaplan’s work spans autonomous coding agents, machine-learning data infrastructure and Tesla Autopilot.

A browser-testing problem leads to cheaper token processing, multilingual models, and a crucial distinction: processing a long input is not the same as remembering it.

Software engineer whose harness-engineering work makes requirements, tools, and feedback accessible to coding agents, now working on cloud-operation agents at Google Cloud.

AI researcher studying how evaluation metrics shape apparent capabilities and how model performance can be predicted with scale.

AI researcher working on financial question answering, agent-based report generation, and personal AI systems.

Machine-learning researcher working on protein language models, design and biological understanding.

Selecting an object is the beginning. SAM 2 connects tracking, correction and training data through an interface designed alongside the model.

Concept prompts let users name what they want to see. SAM 3’s architecture, data engine and deployment examples explain how that request becomes masks and tracks—and where human intent still matters.

OpenAI founding co-chair whose public views address AI deployment, supervised agents, and tools for better thinking.

Creator and host of the TWIML AI Podcast, with an analytical focus on AI applications, technology platforms, and the practical demands of enterprise deployment.

Impulse founder and coauthor of research on reducing repeated computation in video vision-language models.

Product leader and developer educator focused on AI platforms, agent interfaces, and the boundary between automated synthesis and human judgment.

Co-creator of Writely, which became Google Docs, and a Microsoft technology leader whose AI work combines model reasoning with conventional code.

A March 2024 conversation about combining model reasoning with software workflows—from a stubborn output format to a thousand-call textbook generator.

AI researcher whose collaborative work examines how training data and scaling choices affect multimodal and language-model performance.

Pydantic’s creator builds Python developer tools for typed agents, application observability and controlled execution of AI-generated code.