# I Want More from AI Than Another Agent | matgreten.dev

DevFeed: [I Want More from AI Than Another Agent | matgreten.dev](<https://devfeed.tech/articles/i-want-more-from-ai-than-another-agent-matgreten-dev-86866.md>)

Original publisher: [Read original article](<https://swamp-club.com/feed/p/986de6de-bbfd-4e70-839b-f21c4202c08a/go>)

Author: mgreten

Published: 2026-10-14T15:35:34Z

Content type: opinion

Language: en

Sources: [Swamp Club Feed](<https://devfeed.tech/sources/swamp-club-feed.md>)

Topics: [AI Agent](<https://devfeed.tech/topics/ai-agent.md>), [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>), [Functional programming](<https://devfeed.tech/topics/functional-programming.md>), [Pull Request](<https://devfeed.tech/topics/pull-request.md>), [Test automation](<https://devfeed.tech/topics/test-automation.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [ai](<https://devfeed.tech/tags/ai.md>), [behavior](<https://devfeed.tech/tags/behavior.md>), [operations](<https://devfeed.tech/tags/operations.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [workflows](<https://devfeed.tech/tags/workflows.md>)

## AI overview

The author argues that reusable, behavior-centric operations reduce repeated AI orchestration: define typed methods with validated inputs and outputs, then compose them into workflows that control sequence, conditions, and failure handling. In Swamp, workflows can call models for bounded tasks such as assessing PR comments, while code handles settled mechanics such as polling, running checks, and recording results; posting replies or applying changes remains a separate action. The author says this makes procedures easier to inspect and reuse across projects, but does not eliminate API failures, stale assumptions, model errors, or human responsibility. They also describe scaling back excessive checks: code can validate required fields, while judging whether a plan makes sense remains a human decision.

## Source excerpt

A developer argues for behavior-centric design over agent-centric workflows, advocating for reusable, typed operations with defined inputs and outputs that constrain when AI reasoning is invoked rather than letting models orchestrate entire processes.