Labs, models, tools & papers
Every entry collects all the places it appears on YLEM: field guides, timeline milestones, lab reports and log notes.
Labs8
AnthropicLab behind Claude, Constitutional AI and the Model Context Protocol.8 mentionsDeepSeekChinese lab known for efficient open-weight models, including R1.1 mentionsGitHubCode host and maker of Copilot, the first mass-market AI coding assistant.2 mentionsGoogle DeepMindGoogle's combined AI lab: Gemini, AlphaFold and robotics research.9 mentionsMeta AIMeta's AI research group and publisher of the open-weight Llama family.4 mentionsMistral AIEuropean lab shipping open-weight and commercial models.1 mentionsOpenAILab behind the GPT series, ChatGPT and the o-series reasoning models.12 mentionsStability AIReleased Stable Diffusion as open weights in 2022.1 mentions
Models11
AlphaFoldDeepMind's protein structure prediction system.1 mentionsClaudeAnthropic's family of assistant models.5 mentionsDeepSeek-R1Open-weight reasoning model trained largely with reinforcement learning.3 mentionsGeminiGoogle DeepMind's natively multimodal model family.3 mentionsGPT seriesOpenAI's line of generative pre-trained transformers, from GPT-2 to GPT-4o and beyond.6 mentionsLlamaMeta's open-weight model family, the base of much of the open ecosystem.3 mentionsMixtralMistral's sparse mixture-of-experts models.2 mentionso-seriesOpenAI's reasoning models trained to think before answering, starting with o1.2 mentionsRT-2Vision-language-action model that turns web knowledge into robot actions.2 mentionsSoraOpenAI's text-to-video model.2 mentionsStable DiffusionOpen-weight latent diffusion model for images.2 mentions
Tools3
Standards1
Papers9
Attention Is All You NeedIntroduced the Transformer architecture.2 mentionsChain-of-Thought PromptingIntermediate reasoning steps improve multi-step problem solving.3 mentionsChinchillaCompute-optimal training: scale data with parameters.3 mentionsConstitutional AITraining harmlessness from AI feedback guided by written principles.3 mentionsFlashAttentionIO-aware exact attention that made long contexts practical.2 mentionsInstructGPTRLHF at scale; the recipe behind instruction-following assistants.3 mentionsReActInterleaves reasoning traces with tool actions; the template for agent loops.3 mentionsRetrieval-Augmented GenerationCombined a retriever with a generator; named the RAG pattern.3 mentionsScaling Laws for Neural Language ModelsShowed loss falls predictably with model size, data and compute.3 mentions