Definition
Agentic loop
A model running repeatedly toward a validated goal rather than answering a single prompt.
From the episode “40,000 Models, One API Key, and a $25M Bet on Open Source”Hear it on the podcast1 episode
Learn the language of AI-native founders
Every week: the new terms, tools, and workflows founders are using, straight from the podcast.
Related terms in Agents & Automation
Agentic economyAn economy where autonomous AI agents transact, negotiate, and execute on behalf of businesses.Agentic FrameworkA software architecture in which autonomous digital agents handle discrete, well-defined tasks within a larger workflow — passing outputs to the next agent, escalating to human review when thresholds are exceeded, and operating continuously without manual triggering. Softheon's average customer runs 1,300 such agent processes. Eugene Sayan filed a patent for agentic frameworks applied to enterprise healthcare applications in 1998 — more than two decades before the current AI wave made the architecture mainstream.Agentic GatewayA middleware layer that sits between AI agents and the external APIs or services they need to call, handling authentication, payment routing, and access control so the agent operator never has to manage individual credentials. Mitchell Jones built Lava as an agentic gateway: install one MCP, load a wallet, and your Claude Code or Codex instance can call financial data, LLM providers, enrichment tools, and blockchain APIs without separate signups or secret key management. The gateway abstracts the plumbing so humans can focus on directing agents rather than provisioning them.Agentic IdentityA cryptographic identity assigned to an AI agent that (a) is distinct from the human who authorized it, (b) is linked to that human authorization through a verifiable chain, (c) is scoped to specific permissions and a bounded time window, and (d) is sealed in a tamper-evident attestation that persists after the agent terminates. Agentic identity solves the firefly problem: AI agents come and go in seconds, but the authorization record must outlive them to enable audit, compliance, and incident response. Jasson Casey's architecture: the authorization chain reads as 'this user on this device with this posture authorizes this agent on this device with this posture, with these permissions, for this period.' Both sides of the chain are cryptographically linked and sealed. Without this, an enterprise buyer asking 'what did your AI agent do and who authorized it?' has no answer.Agentic operating systemA connected layer of AI agents and automations that runs defined roles inside a business, rather than a collection of standalone tools a human has to coordinate.Agentic OrganizationA company where AI agents hold defined leadership roles, interact with each other and with human contractors, and operate on real budgets with real authority.Agentic PlatformSoftware infrastructure that hosts, orchestrates, and manages AI agents at enterprise scale, independent of any single model provider. Distinct from the agents themselves and the underlying models — the layer that connects, controls, and logs everything.Agentic Professional ServicesProfessional services — accounting, legal, medical, financial advice — delivered primarily through AI agents that handle the technical labor, with human professionals providing review, verification, and trust-layer touchpoints rather than doing the work throughout. Deduction.com is the canonical example: the AI agent handles the tax work, a licensed CPA reviews and signs. The human is not eliminated — they are repositioned to where their presence has the highest value.