Frontier Systems
AI, evidence, and coordination systems for institutions making consequential decisions.
The systems
We build frontier systems that expand what people and organizations can do.
Outbreak response
LOVS for BDBV-2026
For responders and analysts deciding where the Bundibugyo virus response needs support next: a daily brief and map, built from official data.
Running. Updated as official situation reports are published.
Open the brief and map, LOVS for BDBV-2026 (opens in a new tab)Field programs
Impact Data Bridge
A worked example for program teams acting on what communities report: each observation keeps who made it, and that person hears the decision.
A public worked example, on illustrative data.
Open the worked example, Impact Data BridgeAgents on the web
Agent Internet Runtime
For developers whose agents use the web, and site owners deciding what agents may do there: open-source tools and the agent.json standard.
Open source.
Visit agentinternetruntime.com, Agent Internet Runtime (opens in a new tab)
Outbreak response
LOVS for BDBV-2026
For responders and analysts deciding where the Bundibugyo virus response needs support next: a daily brief and map, built from official data.
Explore this system (opens in a new tab)
LOVS spatial map ·
BDBV spatial overlay: Arcede LOVS · Health zones: GRID3 COD Health Zones v8.0 (CC BY 4.0; CIESIN/Columbia, DRC Ministere de la Sante, GRID3). Marker positions: WorldPop Global 2 R2025A 2026 (CC BY 4.0). Uganda districts: geoBoundaries (public domain) · OpenFreeMap © OpenMapTiles Data from OpenStreetMap
Running. Updated as official situation reports are published.
Ambitious in purpose. Precise in practice.
Selected
All newsArticle · 20 Oct 2026 · Frans Moore
What AI does to a state depends on where it is aimed
I modelled AI inside three states: the UK, Kenya and the DRC. Where it was aimed, and whether safeguards held, decided what it did.
Apolitical
LOVS: mapping disease outbreaks (opens in a new tab)
A case study of how LOVS turns official outbreak data into spatial maps, forecasts and searchable evidence, on the Apolitical Navigator.
Arcede
When AI Agents Act Without Permission
An AI agent can complete a task correctly and still take an action nobody authorized. What organizations should ask before putting one to work.
Arcede
Working with philanthropy
An argument to funders for putting frontier systems into public-benefit work, where the hours they free go back to the team rather than out of headcount.
Foundations
A conversation with Arcede
Start with the work.
What is your team trying to make possible? Where does the work get stuck? Start there. A useful conversation does not require a technical brief.
Please do not include confidential or sensitive information in an initial message.
or email info@arcede.com