The wiki

AI-assisted scattering calculations show how a well-posed question, a simplifying assumption, and expert verification can turn an unwieldy expression into a research result.

AI educator and community organizer whose work connects foundational research, reasoning models and practical questions about interpretability.

Crixet founder whose software became Prism at OpenAI, with earlier work making technical ideas understandable through interactive visualization.

Part 2 of 6: VideoPoet turns video generation into sequence modeling; Tali Dekel explores controllable editing through neural atlases and motion features; Sander Dieleman begins a geometric introduction to diffusion.

Creator of Moondream and LNQA, focused on compact vision-language models, practical deployment and reasoning grounded in images.

AWS software development manager whose work and public demonstrations span AI-assisted development, testing, and observability.

Kepler co-founder and CEO working on verifiable financial AI and forward deployed engineering that feeds customer insight back into the product.

Together AI co-founder focused on open, user-owned AI, efficient computing, and the economics of production inference.

HackerRank’s co-founder and CEO on assessing AI-assisted developers, the limits of coding benchmarks, and hiring for demonstrated skill.

Microsoft’s Visual Studio Code evolved from a cross-platform editor and extension platform into an AI editor, opening its Copilot Chat code and using agents in its own development workflow.

Cognition co-founder and product leader whose writing explains context engineering, coordinated agents, and the infrastructure needed for autonomous software development.

Warp evolved from a Rust terminal into an agentic development environment and infrastructure for automating software development.

Warp’s transition from a developer terminal to an environment for coding with multiple AI agents, followed by cloud orchestration and software-factory infrastructure.

A browser-based AI creation platform that grew from generating imagined websites into tools for building, sharing, and collaboratively updating games and web apps.

AI evaluation researcher and Arena co-founder whose work combines crowdsourced human preferences with automated benchmarks for model development.

Recovering text from embeddings, aligning model representations, and measuring memorization reveal the gap between information stored and information we can use.

A practical guide to managed clusters, reliability and security from SemiAnalysis’s ClusterMAX 3.0 discussion.

A speaker-management contest tested AI-built software against the details of real event operations—and exposed the gap between a demo, an evaluation and an owned service.

Part 2 follows the choices behind useful model improvements: architecture, evaluation, training budgets, context, serving costs—and the research habits that sustain them.

Goal-reaching agents benefit from depth when the learning objective, architecture, and supply of experience scale together.

Shrinking model advantages put more weight on product design, integration and execution. The Unsupervised Learning × Latent Space crossover explores where that work creates value—and where technical limits still matter.

Human task duration makes AI capability easier to interpret. Productivity, autonomous research and catastrophic risk depend on more than the work a benchmark captures.

Part 1 follows five research bets—from longer inference to online models—then examines Sora's physical consistency, Groq's utilization economics, and Gemini's million-token context.

Omar Qazi’s public commentary spans local language-model inference and Tesla’s Full Self-Driving system.

Imbue’s approach starts with a recruiting product’s trust problem and asks what it takes to make models dependable enough to delegate work to.

Pushmeet Kohli and Sal Candido on protein structure, biological data, and the work required to make models useful for science and medicine.

Falling model prices weakened OpenPipe’s distillation business. Reinforcement learning offered a new opportunity: improve agents through experience, provided their training environments behave like production.

Narrow tests and leaked repository knowledge make higher scores harder to trust. The next coding evaluations need harder work and credible judgments of quality.

Creator of Verifiers and a Prime Intellect researcher working on reinforcement learning for agents that use tools across multiple turns.

Exa co-founder whose work and public arguments center on neural search and the retrieval needs of AI systems.