April 25, 2026
Bots in the walls or chaos in the halls?
Agents Aren't Coworkers, Embed Them in Your Software
Fans want calm background bots; skeptics yell hype and “negative value”
TLDR: The piece argues AI helpers should live inside apps and work quietly with clear goals and change alerts. The comments split hard: some love the low-noise vision, while others blast a factual slip and declare they prefer fast, reliable, non-agent software—calling today’s agents “negative value.”
The hot take du jour: stop treating AI assistants like chatty coworkers and build apps so bots quietly do the work in the background. The author channels Mark Weiser’s “calm computing” and says fewer chats, more quiet progress using clear commands, goal-setting, and data change alerts. Comment section? Explosive. The top quip—“not yet coworkers*”—set the tone: eye-rolls with a wink.
Supporters cheered the vibe of “ambient agents,” with one dev saying they’ve already built something like it and it works. But a credibility crater opened fast: a commenter flagged timeline mix-ups about which agent tools came first and declared they “lost trust” in paragraph one. That fueled the skeptics, led by a broadside: “I’d pay more for fast, deterministic software without agents… this stuff is negative value.”
Another sore spot: the article’s “calm” tips (command lines, config files, set-it-and-forget-it loops) read to critics like chores, not magic—“sounds more like making the human work for the bot.” Meanwhile, the database angle—streaming only what changed so bots don’t scan everything—got grudging nods for practicality. Verdict: half the crowd wants bots to fade into the wallpaper; the other half wants them out of the house entirely.
Key Points
- •The article advocates embedding AI agents into software via structured interfaces rather than relying on conversational, human-like interactions.
- •It references Mark Weiser’s vision of calm technology—systems that fade into the background and minimize attention demands.
- •Recommended patterns include robust CLIs, declarative specs (configs/schemas/manifests), and reconciliation loops popularized by Kubernetes.
- •Feldera, an incremental query engine, applies these patterns using a CLI, SQL, and a desired-state control plane for pipelines.
- •For data workflows, the article promotes change data capture (CDC) streams so agents react to precise updates instead of polling snapshots.