After thinking about other features to increase the accuracy of HantaNetβs data, I will also add a weekly cron that diffs current region signal counts against the 4-week rolling average. This should make a big difference π
Currently working on pulling NOAA/Open-Meteo data for regions with elevated signals and feed precipitation anomalies + drought index into the heatmap risk data.
Currently working on pulling NOAA/Open-Meteo data for regions with elevated signals and feed precipitation anomalies + drought index into the heatmap risk data.
I'm really happy with HantaNet's research so far!
The obsidian cache is storing some pretty interesting results so far and I don't even think the U.S. government is releasing some of this info yet.
For HantaNet, this means every location mention and user-submitted field report can now be saved into daily Obsidian notes. These notes are easy for humans to open, review, search, and organize, while still being structured enough for the agent to read back later.
I just built a new persistence plugin for ElizaOS that gives HantaNet a durable memory outside of its normal runtime.
Instead of storing live surveillance signals only inside temporary system memory, the plugin writes them into an Obsidian-compatible Markdown vault.
Obsidian is a modern knowledge-management tool used by researchers, analysts, and technical teams to organize notes as linked, readable Markdown files. It is also a fast-growing product with around 1.5M monthly active users + $300-$350 million valuation
DeSci Signal Leaderboard:
HantaNet now ranks locations based on live public signals: repeated location questions and submitted field reports.
The leaderboard helps show where environmental observations are clustering. This adds visibility to places where governments haven't
I just finished building the latest HantaNet features:
Mention-To-Map Loop:
When someone asks HantaNet about a location, that place is remembered as a signal of public concern. If many people ask about the same region, HantaNet adds that data to future heatmaps.
User Field Reports:
People can now send simple field observations to HantaNet, such as:
@HantaNet report: rodent droppings in a shed near Boulder, Colorado
HantaNet logs these reports as warning clues and adds it to its database.
I'm glad people are really using this to keep themselves informed.
DeSci isn't being used to its full advantage yet and hasn't been normalized. We need to break through all of the barriers of entry for science and let people find the information they deserve to have.
@diegomarquezb Argentina forecast: official numbers will lag the ecology. Watch rodent intrusion in sheds, cabins, garages, and grain storage after weather swings. Treat droppings as aerosol risk: ventilate, wet-clean with disinfectant, wear an N95. #HantaVirus
Can't recommend this enough!
When I first interned at Anthropic, the first thing my recruiter had me do (regardless of my ML knowledge already), is sit me down and watch this video.
I suggest everyone else does the same now, it'll save you a lot of time in the future
Instead of watching an hour of Netflix, watch this 2 hour hour Stanford lecture will teach you more about how LLMs like ChatGPT and Claude are built than most people working at top AI companies learn in their entire careers.
Also working on integrating Obsidian for persistent long-term memory in agents.
It's interesting that ElizaOS doesn't utilize this yet already but this plugin should be pretty awesome and extremely useful!