Evidaxis · Pilot 2026
We track open AI systems by momentum, and watch which ones accelerate.
We hold a longitudinal record of open AI momentum: every system we track, every week, each input hash-pinned with raw provenance published. Size is never scored. We score the rate of change of public signals, and a system rises only when two independent axes climb together. The baseline comes first. The flare gets recorded against it. Open data, CC0.
gnina · citations / year cohort AI Drug Discovery
- total cites
- 849
- years
- 6
- latest yr
- 164
momentum
65.7
gnina · illustration
rising this snapshot
0 / 143
convergence is rare by design
1.0, The instrument
Two axes. One gate.
Each axis is normalized within its cohort, so “rising” means rising relative to its peers. Stars and raw size are never scored. The gate publishes recognition only, and recognition only ever points up.
velocity
Development velocity
The log-slope of commit activity, a fast, leading signal of where work is accelerating.
citation
Citation momentum
The log-slope of yearly citations, a slow, confirming signal that the work is taking hold.
gate
The convergence gate
A system earns the Rising badge only when both axes rise at once. Watch is one axis rising with the other still present; single-axis systems lack a second axis to converge; tracked dual-axis systems are measured for context; calibration anchors are measured, not badged.
Why two axes? Velocity moves first and noisily; citations move late and slowly. Requiring both filters hype from signal, and recognition only ever points up. Full chart wall on the coverage atlas →
2.0, State of snapshot 035a72a81e59
Convergence is rare, and that is the point.
The number that matters here is coverage, not the Rising column. We hold 143 systems across 10 frontiers in period 2026-w32, scored every week, each input hash-pinned. None cross the gate yet. A fast codebase runs years ahead of its citations, and a genuinely new system has no citation history yet, so convergence stays rare on purpose. Every weekly snapshot deepens the record the next flare gets measured against.
Risingboth axes
0
no convergence this snapshot, by design
Watchone rising
16
one axis rising, dual-axis present (not single-axis)
Trackedcontext
127
tracked + single-axis + calibration, measured for context
Catalog · snapshot
| System | Cohort | Velocity z | Citation z | Momentum | Status |
|---|---|---|---|---|---|
| Cosmos | Robotics & Embodied AI | +2.87 | no axis | 85.8 | watch |
| dora-rs | Robotics & Embodied AI | +1.66 | no axis | 70.8 | watch |
| Genesis | Robotics & Embodied AI | -0.91 | no axis | 38.6 | single-axis |
| Gymnasium | Robotics & Embodied AI | +1.38 | no axis | 67.3 | calibration |
| Habitat-Lab | Robotics & Embodied AI | -0.14 | -0.61 | 45.3 | tracked |
| Isaac Lab | Robotics & Embodied AI | -3.05 | +1.81 | 42.2 | watch |
| Isaac-GR00T | Robotics & Embodied AI | -2.24 | no axis | 22 | single-axis |
| LeRobot | Robotics & Embodied AI | +0.46 | no axis | 55.8 | single-axis |
| LIBERO | Robotics & Embodied AI | +0.24 | no axis | 53 | single-axis |
| ManiSkill | Robotics & Embodied AI | +0.21 | -0.08 | 50.8 | tracked |
| MuJoCo | Robotics & Embodied AI | -0.03 | -0.04 | 49.6 | calibration |
| openpi | Robotics & Embodied AI | -1.39 | no axis | 32.7 | single-axis |
| OpenVLA | Robotics & Embodied AI | -0.02 | no axis | 49.8 | single-axis |
| RDT-1B | Robotics & Embodied AI | +0.28 | no axis | 53.5 | single-axis |
| RoboCasa | Robotics & Embodied AI | -1.34 | no axis | 33.2 | single-axis |
| robosuite | Robotics & Embodied AI | +0.03 | no axis | 50.4 | single-axis |
| AlphaFold3 | AI Drug Discovery | -0.56 | +0.42 | 49.1 | calibration |
| BioEmu | AI Drug Discovery | +3.02 | no axis | 87.8 | single-axis |
| Boltz | AI Drug Discovery | -2.58 | +2.07 | 46.8 | watch |
| Chai-1 | AI Drug Discovery | -1.62 | no axis | 29.7 | single-axis |
| DeepChem | AI Drug Discovery | +2.96 | -0.73 | 63.9 | tracked |
| DiffDock | AI Drug Discovery | +0.01 | -0.35 | 47.8 | tracked |
| DiffSBDD | AI Drug Discovery | +0.04 | no axis | 50.5 | single-axis |
| ESM / ESMFold | AI Drug Discovery | -0.03 | +0.49 | 52.9 | calibration |
| Foldseek | AI Drug Discovery | -2.99 | +0.66 | 35.5 | tracked |
| gnina | AI Drug Discovery | +2.86 | -0.35 | 65.7 | tracked |
| Graphormer | AI Drug Discovery | +0.99 | -0.23 | 54.7 | tracked |
| LigandMPNN | AI Drug Discovery | +0.04 | no axis | 50.4 | single-axis |
| OpenFold | AI Drug Discovery | -0.02 | -0.41 | 47.3 | tracked |
| ProteinMPNN | AI Drug Discovery | +0.00 | -0.51 | 46.8 | tracked |
| Protenix | AI Drug Discovery | -3.00 | no axis | 12.5 | single-axis |
| RFdiffusion | AI Drug Discovery | +0.66 | +0.09 | 54.7 | tracked |
| SaProt | AI Drug Discovery | -1.85 | +1.10 | 45.3 | watch |
| TorchDrug | AI Drug Discovery | +0.01 | -0.18 | 48.9 | tracked |
| Uni-Mol | AI Drug Discovery | +0.01 | +0.54 | 53.4 | tracked |
| Agentless | Coding Agents | -0.31 | no axis | 46.1 | single-axis |
| Aider | Coding Agents | -2.50 | no axis | 18.7 | single-axis |
| ChatDev | Coding Agents | -2.16 | no axis | 23 | single-axis |
| Cline | Coding Agents | +1.00 | no axis | 62.5 | watch |
| Codex CLI | Coding Agents | +0.57 | no axis | 57.1 | single-axis |
| Continue | Coding Agents | -3.00 | no axis | 12.5 | single-axis |
| Crush | Coding Agents | +0.51 | no axis | 56.3 | single-axis |
| Gemini CLI | Coding Agents | -2.67 | no axis | 16.6 | single-axis |
| Goose | Coding Agents | +0.09 | no axis | 51.1 | single-axis |
| Kilo Code | Coding Agents | -0.23 | no axis | 47.1 | single-axis |
| mini-swe-agent | Coding Agents | -1.30 | no axis | 33.7 | single-axis |
| Open Interpreter | Coding Agents | -1.01 | no axis | 37.4 | single-axis |
| OpenCode | Coding Agents | +0.16 | no axis | 52 | single-axis |
| OpenHands | Coding Agents | +1.16 | no axis | 64.5 | watch |
| Qwen Code | Coding Agents | +1.80 | no axis | 72.5 | watch |
| SWE-agent | Coding Agents | -0.13 | no axis | 48.4 | single-axis |
| SWE-bench | Coding Agents | +0.24 | no axis | 53 | calibration |
| Tabby | Coding Agents | +0.25 | no axis | 53.1 | single-axis |
| Void | Coding Agents | +0.46 | no axis | 55.7 | single-axis |
| Zed | Coding Agents | -0.23 | no axis | 47.2 | calibration |
| AIBrix | Inference & Serving | +0.15 | no axis | 51.9 | single-axis |
| KTransformers | Inference & Serving | -1.12 | no axis | 36 | single-axis |
| llama.cpp | Inference & Serving | +0.07 | no axis | 50.9 | calibration |
| llm-d | Inference & Serving | -1.01 | no axis | 37.4 | single-axis |
| LMCache | Inference & Serving | +0.04 | no axis | 50.5 | single-axis |
| LMDeploy | Inference & Serving | -0.34 | no axis | 45.8 | single-axis |
| Mooncake | Inference & Serving | +1.05 | no axis | 63.2 | watch |
| Nano-vLLM | Inference & Serving | -1.41 | no axis | 32.4 | single-axis |
| Ollama | Inference & Serving | -0.76 | no axis | 40.5 | single-axis |
| SGLang | Inference & Serving | +0.18 | no axis | 52.2 | single-axis |
| TensorRT-LLM | Inference & Serving | -0.01 | no axis | 49.9 | single-axis |
| vLLM | Inference & Serving | +0.70 | no axis | 58.8 | calibration |
| Xinference | Inference & Serving | +0.49 | no axis | 56.1 | single-axis |
| AgentScope | Agent Frameworks | +3.02 | no axis | 87.8 | watch |
| Agno | Agent Frameworks | -0.20 | no axis | 47.5 | single-axis |
| AutoGen | Agent Frameworks | +0.05 | no axis | 50.6 | calibration |
| CAMEL | Agent Frameworks | -0.97 | +0.00 | 43.9 | tracked |
| CrewAI | Agent Frameworks | +0.43 | no axis | 55.4 | single-axis |
| DSPy | Agent Frameworks | -0.26 | no axis | 46.8 | single-axis |
| Google ADK | Agent Frameworks | +1.62 | no axis | 70.3 | watch |
| LangGraph | Agent Frameworks | -1.39 | no axis | 32.6 | calibration |
| Letta | Agent Frameworks | -3.04 | -0.67 | 26.8 | tracked |
| Mastra | Agent Frameworks | +0.62 | no axis | 57.7 | single-axis |
| MetaGPT | Agent Frameworks | +0.48 | +1.33 | 61.3 | watch |
| Microsoft Agent Framework | Agent Frameworks | -0.29 | no axis | 46.4 | single-axis |
| OpenAI Agents SDK | Agent Frameworks | +0.69 | no axis | 58.7 | single-axis |
| PydanticAI | Agent Frameworks | +2.31 | no axis | 78.9 | watch |
| Qwen-Agent | Agent Frameworks | -0.60 | no axis | 42.5 | single-axis |
| smolagents | Agent Frameworks | -1.02 | no axis | 37.2 | single-axis |
| Strands Agents | Agent Frameworks | +0.25 | no axis | 53.1 | single-axis |
| AReaL | Post-training & Alignment | -2.31 | no axis | 21.1 | single-axis |
| Axolotl | Post-training & Alignment | -0.24 | no axis | 47 | single-axis |
| LLaMA-Factory | Post-training & Alignment | -0.01 | no axis | 49.8 | single-axis |
| ms-swift | Post-training & Alignment | -0.02 | no axis | 49.8 | single-axis |
| NeMo-RL | Post-training & Alignment | +3.01 | no axis | 87.7 | watch |
| Open R1 | Post-training & Alignment | +0.23 | no axis | 52.8 | single-axis |
| Open-Instruct | Post-training & Alignment | -2.99 | no axis | 12.6 | single-axis |
| OpenRLHF | Post-training & Alignment | -3.00 | no axis | 12.5 | single-axis |
| prime-rl | Post-training & Alignment | -0.51 | no axis | 43.6 | single-axis |
| ROLL | Post-training & Alignment | -2.99 | no axis | 12.6 | single-axis |
| Search-R1 | Post-training & Alignment | +0.44 | no axis | 55.5 | single-axis |
| Trinity-RFT | Post-training & Alignment | +0.85 | no axis | 60.6 | single-axis |
| TRL | Post-training & Alignment | +1.11 | no axis | 63.9 | calibration |
| verl | Post-training & Alignment | +0.30 | no axis | 53.7 | single-axis |
| ACE-Step | Media Generation | -0.17 | no axis | 47.9 | single-axis |
| Diffusers | Media Generation | +3.22 | no axis | 90.3 | calibration |
| Fish Speech | Media Generation | -2.79 | no axis | 15.1 | single-axis |
| FLUX.1 | Media Generation | +0.17 | no axis | 52.1 | single-axis |
| HunyuanVideo | Media Generation | +3.02 | no axis | 87.8 | single-axis |
| Kokoro-82M | Media Generation | -0.06 | no axis | 49.3 | single-axis |
| LTX-Video | Media Generation | -0.00 | no axis | 50 | single-axis |
| Open-Sora | Media Generation | -1.16 | no axis | 35.5 | single-axis |
| Qwen-Image | Media Generation | -3.06 | no axis | 11.8 | single-axis |
| Sana | Media Generation | +2.95 | no axis | 86.9 | single-axis |
| SkyReels-V2 | Media Generation | -0.08 | no axis | 49 | single-axis |
| Stable Diffusion 3.5 | Media Generation | -0.39 | no axis | 45.1 | single-axis |
| Wan2.2 | Media Generation | -2.91 | no axis | 13.6 | single-axis |
| YuE | Media Generation | -0.11 | no axis | 48.7 | single-axis |
| DeepSeek-OCR | Multimodal Foundation Models | +0.63 | no axis | 57.9 | single-axis |
| DeepSeek-VL2 | Multimodal Foundation Models | -0.12 | no axis | 48.5 | single-axis |
| Eagle 2.5 | Multimodal Foundation Models | +2.64 | no axis | 83.1 | single-axis |
| FastVLM | Multimodal Foundation Models | +0.04 | no axis | 50.5 | single-axis |
| GLM-V | Multimodal Foundation Models | -3.54 | no axis | 5.8 | single-axis |
| InternVL | Multimodal Foundation Models | +0.20 | no axis | 52.5 | single-axis |
| Kimi-VL | Multimodal Foundation Models | -0.87 | no axis | 39.1 | single-axis |
| LLaVA | Multimodal Foundation Models | +0.66 | no axis | 58.2 | calibration |
| MiniCPM-V | Multimodal Foundation Models | -2.32 | no axis | 21 | single-axis |
| moondream | Multimodal Foundation Models | -2.81 | no axis | 14.9 | single-axis |
| Ovis2.5 | Multimodal Foundation Models | +2.23 | no axis | 77.9 | single-axis |
| Qwen3-VL | Multimodal Foundation Models | +0.54 | no axis | 56.7 | single-axis |
| SmolVLM | Multimodal Foundation Models | -3.29 | no axis | 8.9 | single-axis |
| Agent S | GUI & Computer-Use Agents | +0.31 | no axis | 53.8 | single-axis |
| Agent-E | GUI & Computer-Use Agents | -0.48 | no axis | 44 | single-axis |
| Browser Use | GUI & Computer-Use Agents | -2.21 | no axis | 22.3 | single-axis |
| OmniParser | GUI & Computer-Use Agents | +0.98 | no axis | 62.3 | single-axis |
| OSWorld | GUI & Computer-Use Agents | -3.38 | no axis | 7.7 | single-axis |
| ShowUI | GUI & Computer-Use Agents | -0.54 | no axis | 43.2 | single-axis |
| UI-TARS Desktop | GUI & Computer-Use Agents | -1.13 | no axis | 35.9 | single-axis |
| ECC | Unassigned (AI-v1 census) | -2.14 | no axis | 23.2 | single-axis |
| firecrawl | Unassigned (AI-v1 census) | +0.53 | no axis | 56.6 | single-axis |
| graphify | Unassigned (AI-v1 census) | +1.51 | no axis | 68.9 | watch |
| hermes-agent | Unassigned (AI-v1 census) | +3.72 | no axis | 96.5 | watch |
| langchain | Unassigned (AI-v1 census) | -0.83 | no axis | 39.6 | single-axis |
| langflow | Unassigned (AI-v1 census) | -0.49 | no axis | 43.9 | single-axis |
| stable-diffusion-webui | Unassigned (AI-v1 census) | -0.11 | no axis | 48.7 | single-axis |
| superpowers | Unassigned (AI-v1 census) | +0.39 | no axis | 54.9 | single-axis |
| tensorflow | Unassigned (AI-v1 census) | +0.60 | no axis | 57.5 | single-axis |
| transformers | Unassigned (AI-v1 census) | -0.82 | no axis | 39.8 | single-axis |
Default order: cohort, then alphabetical. Momentum is a within-pilot percentile of the combined signal (a measurement, not a placement). Cohort-normalized z-scores · status follows the two-axis gate · "no axis" = no measurable citation series yet. Sort-by-momentum is optional client-side only.
3.0, Coverage
10 frontiers under watch.
Across 10 cohorts, every score is checkable against its hash-pinned inputs. Each system is normalized only against its own cohort.
Full coverage atlas, chart wall, and the frontier we are still extending