Key Takeaways
- Claw4Science grew from 54 to 107 projects in the second half of March — a 98% increase
- Skill hubs grew from 19 to 24 (+26%)
- Blog posts grew from 10 to 25 (+150%)
- Notable arrivals: Karpathy's autoresearch (60K stars), Alibaba DeepResearch, NVIDIA NemoClaw, AgentLaboratory
- EdgeClaw 2.0 launched with ClawXMemory and ClawXRouter
- Underserved categories: agriculture, environmental science, materials science
The Numbers
| Metric | Start of March | End of March | Change |
|---|---|---|---|
| Projects | 54 | 107 | +98% |
| Skill hubs | 19 | 24 | +26% |
| Blog posts | 10 | 25 | +150% |
Timeline
Week 1 (Mar 18–22): Claw4Science launched with 54 projects.
Week 2 (Mar 23–25): Discovery phase. GitHub page-by-page search found ScienceClaw (4 variants), PaperClaw (6), DrugClaw (2), ResearchClaw (4).
Week 3 (Mar 26–29): The floodgates. Cross-referencing added Karpathy's autoresearch (60K stars), Alibaba DeepResearch, AgentLaboratory. Also NVIDIA NemoClaw, Hermes Agent (21K stars).
Week 4 (Mar 30–31): Consolidation. Hit 107 projects and 24 skill hubs.
Notable Arrivals
| Project | Stars | Why Notable |
|---|---|---|
| autoresearch (Karpathy) | 60K | Biggest name in AI doing research automation |
| DeepResearch (Alibaba NLP) | 5K+ | Major tech company entry |
| NemoClaw (NVIDIA) | 200+ | NVIDIA's official entry |
| AgentLaboratory | 4K+ | Full autonomous research lab |
| EdgeClaw 2.0 | 343 | Memory engine + cost-saving router |
What's Coming Next
- Agriculture: Zero projects currently — wide open for the first mover
- Environmental science: Only PantheonOS touches this space
- Materials science: ChemGraph is alone — opportunity for specialization
- Project consolidation: Similar projects may start collaborating or merging
Summary
March 2026 was the month the OpenClaw science ecosystem went from "interesting niche" to "you can't ignore this anymore." 107 projects, 24 skill hubs, major corporate entries (NVIDIA, Alibaba), and a doubling of the directory in two weeks. The question is no longer "will AI agents matter for science?" — it's "which ones will you use?"