1st Place: $5,000 Cash
2nd Place: $2,000 Cash
3rd Place: $1,000 Cash
Build AI agents that use a person's memory to deliver something genuinely useful to them.
The key constraint: the value should flow to the user, not to a corporation. Projects might include agents that help users reflect on their own patterns, plan based on their history, process their past decisions, or generate personalized creative work.
<aside> <img src="/icons/light-bulb_orange.svg" alt="/icons/light-bulb_orange.svg" width="40px" />
This track is where the AI conversation logs shine. Each persona's conversations.jsonl contains multi-turn AI sessions covering career decisions, mental health, relationships, finances, and creative work — exactly the kind of intimate, contextual data that makes personalized AI companions possible. Combined with the lifelog.jsonl, you have a 2-year narrative arc for each person: what they did, what they felt, what they asked for help with, and how they grew.
Recommended personas: p02 (Maya Patel) for high-stakes emotional support use cases, and p05 (Theo Nakamura) for productivity, creative work, and self-coaching scenarios.
</aside>
<aside> <img src="/icons/light-bulb_orange.svg" alt="/icons/light-bulb_orange.svg" width="40px" />
Quick load — conversations + lifelog combined (Python):
python
`import json
with open("persona_p02/conversations.jsonl") as f: conversations = [json.loads(line) for line in f]
with open("persona_p02/lifelog.jsonl") as f: lifelog = [json.loads(line) for line in f]
timeline = sorted(conversations + lifelog, key=lambda x: x["ts"])
mh_entries = [e for e in timeline if "mental_health" in e["tags"]]`
See QUICKSTART.md in the dataset folder for full examples including cross-file queries.
</aside>
<aside> <img src="/icons/light-bulb_orange.svg" alt="/icons/light-bulb_orange.svg" width="40px" />