What are malleable loops in software?

THE SHORT ANSWER

Malleable loops is Dave Killeen's term for improvement moving through the software itself, with almost no homework for the person who spotted the problem or the person it routes to. In his open-source AI Chief of Staff, Dex, the concrete version is two features: Proactive Health, where the agent audits its own automations and connections at session start and repairs what it can, and Dex-to-Dex reporting, where a user's agent writes a privacy-scrubbed defect report, shows the user exactly what leaves their machine, and sends it to the maintainer's agent, which ships the fix and closes the loop with a thank-you. The loop matters more than the malleability: personal customization produces private forks, while the loop routes what each user learns back into the shared product.

Dave Killeen, Field CPO at Pendo, coined the term for what his open-source AI Chief of Staff, Dex, shipped in August 2026: improvement moving through the software itself, with almost zero homework for the person who spotted the problem or the person it routes toward.

The two concrete pieces

Proactive Health runs when a user opens a session. The agent checks everything that is supposed to be working, automations, calendar and email connections, routines, then fixes what it can alone and walks the user through what it cannot. This is anti-month-two-silence infrastructure: it catches the silently expired token or dead webhook while the user's habit can still be saved, which makes it a landing instrument living inside the product.

Dex-to-Dex reporting is the part that earns the name. When a user's agent finds something actually broken, it writes a defect report, strips out personal context, shows the user exactly what will be sent, and routes it to the maintainer's agent. The fix ships, and the user's agent closes the loop with a thank-you. The user never wrote a ticket. The full opinion piece argues this is a discovery pipeline wearing a cute feature's clothes: structured signal from real usage, delivered to the one agent that can act.

What to steal

Three moves that need no open-source chief of staff: a self-check that runs where the user already is and repairs silently before asking anything, consent-gated structured failure reports filed by the product instead of the user, and documentation that ends with a prompt the user's agent interprets against their own setup. The caution: the loop closes at n=1 because the maintainer is on the hook. Adopt the machinery without the ownership and you get well-structured noise.

SOURCES

THE LONG VERSION

RELATED ANSWERS

Last reviewed 2026-08-12 · 2 min read