individual
OtherTechnical AI safety researchAI safety fieldbuilding
| BlueDot Impact | $450 | 100% | 3 |
| Date ↓ | Funder | Via | Amount | Cause | Source | Purpose |
|---|---|---|---|---|---|---|
| Sep 2026 | BlueDot Impact | $100 | Other | bluedot | Self-repair in tabular foundation models | |
| Sep 2026 | BlueDot Impact | $300 | Other | bluedot | Can a linear probe detect agent's intent to escape from a sandbox before the action executes, if so, how early? This topic is inspired from the recent incident of OpenAI model hacking HuggingFace. | |
| Jun 2026 | BlueDot Impact | $50 | Interpretability, Training pipelines, Local groups | bluedot | I'm self studying mechanistic interpretability by following curriculum in ARENA 3.0 (https://github.com/callummcdougall/ARENA_3.0), in order to prepare for future research in tabular foundation models |