Personal AI assistant (Dot) built around a long-term, self-evolving memory system.
New Computer builds Dot, a personal AI whose agentic memory system creates and pre-computes documents for later retrieval instead of indexing a static corpus, with optional meta-fields like status and due date used as retrieval filters. The team treats retrieval as an evaluation problem: cohorts of synthetic users generate privacy-safe queries, an in-house labeling tool feeds LangSmith datasets, and precision/recall/F1 experiments compare semantic, keyword, BM25, and meta-field filtering strategies before anything reaches users.
new.computerSources last verified Sep 21, 2026
LangSmith
by LangChain Evaluates memory retrieval and conversation prompts on labeled synthetic datasets.
Receipts · 1
“With LangSmith, New Computer has been able to test and improve their memory retrieval systems” — blog.langchain.dev, Jul 2, 2024
Source last verified Sep 21, 2026