Blog
Field notes on long-running work.
How we think about autonomous agents, the runtime underneath them, and the business model that keeps us honest.
The work that finishes itself
Most automation hands work back to a person the moment it gets hard. We built HorizonStack to do the opposite — to stay on a job until it is actually done.
Dana Whitfield · 6 min read
- EngineeringFebruary 18, 2026
Inside Meridian
A look at the harness underneath every HorizonStack agent: how triggers, policy, queued work, and contact form a loop that runs for weeks.
Marcus Adeyemi · 9 min read
- CompanyJanuary 27, 2026
Pricing agents on outcomes, not tokens
Why we bill for jobs that close rather than for minutes, messages, or model calls — and what that changes about how we build.
Dana Whitfield · 5 min read
- ExplainerMarch 4, 2026
What is a long-running agent?
A plain-language definition of long-running agents, how they differ from chatbots and RPA, and where they earn their keep.
Marcus Adeyemi · 7 min read
- ExplainerMarch 19, 2026
Agents vs. RPA vs. chatbots: a practical comparison
When to reach for a long-running agent, when RPA is enough, and when a chatbot is the right tool — with a decision framework.
Dana Whitfield · 8 min read
- EngineeringApril 8, 2026
Compliance guardrails for autonomous agents
How autonomous agents stay inside regulatory lines: deterministic policy, disclosures, audit logs, and knowing when to escalate.
Marcus Adeyemi · 9 min read