Coverage atlas · Epoch 01

6277 systems. 10 cohorts. One longitudinal map.

You cannot catch a flare without the baseline that came before it. So we record the frontier wide and early, every score frozen, with hash-pinned inputs and open raw provenance. Absence here means one thing: no baseline yet. The instrument is the coverage, and the coverage is what compounds.

Robotics & Embodied AI

Robotics & Embodied AI · 16 systems · 4 with citation axis · 5 with direct-dependents axis

Cosmos dora-rs Genesis Gymnasium Habitat-Lab Isaac Lab Isaac-GR00T LeRobot LIBERO ManiSkill MuJoCo openpi OpenVLA RDT-1B RoboCasa robosuite

AI Drug Discovery

AI for Science · 19 systems · 14 with citation axis · 1 with direct-dependents axis

AlphaFold3 BioEmu Boltz Chai-1 DeepChem DiffDock DiffSBDD ESM / ESMFold Foldseek gnina Graphormer LigandMPNN OpenFold ProteinMPNN Protenix RFdiffusion SaProt TorchDrug Uni-Mol

Coding Agents

AI Agents · 20 systems · 0 with citation axis · 10 with direct-dependents axis

Agentless Aider ChatDev Cline Codex CLI Continue Crush Gemini CLI Goose Kilo Code mini-swe-agent Open Interpreter OpenCode OpenHands Qwen Code SWE-agent SWE-bench Tabby Void Zed

Agent Frameworks

AI Agents · 16 systems · 3 with citation axis · 12 with direct-dependents axis

AgentScope AutoGen CAMEL CrewAI DSPy Google ADK LangGraph Letta Mastra MetaGPT Microsoft Agent Framework OpenAI Agents SDK PydanticAI Qwen-Agent smolagents Strands Agents

GUI & Computer-Use Agents

AI Agents · 7 systems · 0 with citation axis · 2 with direct-dependents axis

Agent S Agent-E Browser Use OmniParser OSWorld ShowUI UI-TARS Desktop

Inference & Serving

AI Infrastructure · 14 systems · 0 with citation axis · 5 with direct-dependents axis

AIBrix FlashInfer KTransformers llama.cpp llm-d LMCache LMDeploy Mooncake Nano-vLLM Ollama SGLang TensorRT-LLM vLLM Xinference

Post-training & Alignment

Foundation Models · 16 systems · 0 with citation axis · 3 with direct-dependents axis

AReaL ART Axolotl LLaMA-Factory ms-swift NeMo-RL Open R1 Open-Instruct OpenRLHF prime-rl ROLL Search-R1 Trinity-RFT TRL Unsloth verl

Multimodal Foundation Models

Foundation Models · 13 systems · 0 with citation axis · 0 with direct-dependents axis

DeepSeek-OCR DeepSeek-VL2 Eagle 2.5 FastVLM GLM-V InternVL Kimi-VL LLaVA MiniCPM-V moondream Ovis2.5 Qwen3-VL SmolVLM

Media Generation

Multimodal & Media Generation · 15 systems · 0 with citation axis · 2 with direct-dependents axis

ACE-Step ComfyUI Diffusers Fish Speech FLUX.1 HunyuanVideo Kokoro-82M LTX-Video Open-Sora Qwen-Image Sana SkyReels-V2 Stable Diffusion 3.5 Wan2.2 YuE

Known gaps

Closed-source systems. Private SaaS leaves few public, hard-to-game signals; this pilot tracks OSS / research-native systems where the data is rich and tamper-resistant.
Systems without a paper. Where no academic source is configured, citation-source availability is recorded as source_not_measured.
Own packages. This axis measures the trajectory of a system's own packages: declared in its own tree, published under that name, and not attributed by deps.dev to another repository. A package with no repository link is unverified and stays. The self-name exclusion list is fixed at the v2h.1 panel. Systems with no admitted package are not measured on this axis. Points that cannot be derived exactly are withheld, and the series restarts after a withheld point.
New systems. Weekly history is reported as an exact count against the four-observation minimum. Each weekly snapshot adds one observation.
Beyond AI. The taxonomy reserves room beyond AI; this pilot measures three AI frontiers only.

The record gets stronger the longer it runs. Only a continuous baseline can catch a flare against its own past, and ours starts here.

See the measurement power funnel →

See the full snapshot →