Project memory · agent history
All sessions
Browse top-level agent sessions attached to this project. Subagent trees stay collapsed under their parent runs until you choose to inspect them.
5Main sessions
0Sub-agent runs
1Harnesses
59Tool calls
Harness mix
claude-code: 5
Top tools: node_context 5, wiki_page 5, search_facts 3, search_nodes 3, raw_source 2, ask 1, lint_report 1, timeline 1
| Session | Agent | Project | Date | Model | Msgs | Tools | Subagents |
|---|---|---|---|---|---|---|---|
| Head-to-head: 3D Gaussian Splatting vs Instant-NGP Compared 3DGS and Instant-NGP head-to-head on Mip-NeRF 360 outdoor average. Initial matrix listed PSNR/SSIM/FPS/train time but no VRAM column; the researcher asked for training mem | Claude Code | Tesserae | 2026-05-16 | claude-opus-4-7 | 11 | 11 | — |
| Investigate feed-forward vs per-scene reconstruction open question Investigated the inference-cost-vs-fidelity open question. Read the existing question doc, enumerated the LRM-family and SDS-on-splats families, and used `ask` for a synthesis pass | Claude Code | Tesserae | 2026-05-15 | claude-opus-4-7 | 9 | 11 | — |
| Concept trace — Multi-View Consistency across the corpus Traced 'Multi-View Consistency' through the corpus as a five-stage timeline anchored to NeRF (2020), NeuS (2021), DROID-SLAM (2021), 3DGS / MonoGS (2023), and 2DGS / MVSp | Claude Code | Tesserae | 2026-05-14 | claude-opus-4-7 | 11 | 12 | — |
| 3D Gaussian Splatting end-to-end lineage explainer Walked the 3D Gaussian Splatting dependency graph from the Kerbl 2023 anchor in both directions. Read NeRF as the only upstream prerequisite, then six direct extensions (SuGaR, 2DG | Claude Code | Tesserae | 2026-05-13 | claude-opus-4-7 | 11 | 12 | — |
| Weekly Gaussian Splatting digest — W19 Compiled W19 digest. First pass recycled LGM and missed surfacing an open question; the researcher pushed back twice. Agent dropped LGM, re-read the Splatting-SLAM trio (MonoGS, GS | Claude Code | Tesserae | 2026-05-12 | claude-opus-4-7 | 12 | 13 | — |