--- title: Improve AI output with continuous improvement | daily.dev description: A technique called continuous improvement uses two AI agents looping together to refine outputs: one agent generates a response and a second critiques it,... canonical: https://daily.dev/posts/improve-ai-output-with-continuous-improvement-hmjak43yt twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Improve AI output with continuous improvement | daily.dev og:description: A technique called continuous improvement uses two AI agents looping together to refine outputs: one agent generates a response and a second critiques it,... og:url: https://daily.dev/posts/improve-ai-output-with-continuous-improvement-hmjak43yt og:image: https://api.daily.dev/og/posts/HmJAK43YT.png og:image:alt: Improve AI output with continuous improvement og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Improve AI output with continuous improvement **[Firebase](https://daily.dev/sources/firebase-blog)** · 32 min read · 0 upvotes · 0 comments ## Summary A technique called continuous improvement uses two AI agents looping together to refine outputs: one agent generates a response and a second critiques it, feeding results back for revision, typically over 2-4 iterations. The Google Agent Development Kit's LoopAgent primitive is demonstrated through a personal-training example, with a receptionist agent gathering user constraints and a lifting_friend/lifting_pro agent pair iterating on a workout plan. Tradeoffs include roughly double the latency and cost compared to single-shot inference, making it best suited for high-fidelity, infrequent tasks rather than daily operations. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How do I use a LoopAgent in Google's Agent Development Kit to implement a critique and revise loop between two agents? Define a LoopAgent with a name, description, max_iterations, and a sub_agents list containing two agents, such as a generator and a critic. The generator produces output, the critic reviews it and either calls exit_loop when satisfied or returns feedback, and the generator revises based on that feedback until max_iterations is reached or exit_loop is called. _Developers wiring up agent critique loops can browse ADK patterns and updates on daily.dev._ ### How many iterations does an AI critique loop typically need to reach a good output? Typically only 2 to 3 iterations are needed for a looping critique-and-revise setup between two AI models to reach a strong output, rather than requiring many repeated passes. This applies to continuous improvement patterns where one model generates content and a second model critiques it before the first revises. _Anyone tuning agent loop iteration counts can track practical AI agent techniques on daily.dev._ ### What are the cost and latency tradeoffs of using a multi-agent critique loop instead of a single AI prompt? A multi-agent critique loop roughly doubles both cost and latency compared to a single inference request, since the model must iterate with another model multiple times before producing a final answer. It works best for one-time or infrequent high-fidelity tasks rather than daily operating workloads, where the added cost and delay outweigh the benefit of a single well-tuned prompt. _Teams weighing agent loop cost against single-prompt speed can follow AI agent design tradeoffs on daily.dev._ ## Similar posts on daily.dev - [Human judgment in the agent improvement loop](https://daily.dev/posts/human-judgment-in-the-agent-improvement-loop-iwssfql5p) · LangChain · 1 upvotes · 0 comments - [The Signals Loop: Fine-tuning for world-class AI apps and agents](https://daily.dev/posts/the-signals-loop-fine-tuning-for-world-class-ai-apps-and-agents-ii8dbe85f) · Microsoft Azure · 0 upvotes · 0 comments - [Continuous AI in practice: What developers can automate today with agentic CI](https://daily.dev/posts/continuous-ai-in-practice-what-developers-can-automate-today-with-agentic-ci-bq4ihvy3k) · GitHub Blog · 2 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#prompt-engineering](https://daily.dev/tags/prompt-engineering), [#google-gemini](https://daily.dev/tags/google-gemini) [View this post on daily.dev](https://daily.dev/posts/improve-ai-output-with-continuous-improvement-hmjak43yt) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Improve AI output with continuous improvement","url":"https://daily.dev/posts/improve-ai-output-with-continuous-improvement-hmjak43yt","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/improve-ai-output-with-continuous-improvement-hmjak43yt"},"datePublished":"2026-08-23T12:20:08.230Z","dateModified":"2026-08-23T12:42:48.331Z","description":"A technique called continuous improvement uses two AI agents looping together to refine outputs: one agent generates a response and a second critiques it,...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/aa6aa2545570c504f20d518cc9f41d9b?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/aa6aa2545570c504f20d518cc9f41d9b?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Firebase","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Firebase","logo":"https://media.daily.dev/image/upload/s--i395oqZH--/f_auto,q_auto/v1787487448/logos/firebase-blog?_a=BAMAMicg0","url":"https://daily.dev/sources/firebase-blog"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/improve-ai-output-with-continuous-improvement-hmjak43yt","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,prompt-engineering,google-gemini","timeRequired":"PT32M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Firebase","item":"https://daily.dev/sources/firebase-blog"},{"@type":"ListItem","position":3,"name":"Improve AI output with continuous improvement"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/improve-ai-output-with-continuous-improvement-hmjak43yt#faq","mainEntity":[{"@type":"Question","name":"How do I use a LoopAgent in Google's Agent Development Kit to implement a critique and revise loop between two agents?","acceptedAnswer":{"@type":"Answer","text":"Define a LoopAgent with a name, description, max_iterations, and a sub_agents list containing two agents, such as a generator and a critic. The generator produces output, the critic reviews it and either calls exit_loop when satisfied or returns feedback, and the generator revises based on that feedback until max_iterations is reached or exit_loop is called. Developers wiring up agent critique loops can browse ADK patterns and updates on daily.dev."}},{"@type":"Question","name":"How many iterations does an AI critique loop typically need to reach a good output?","acceptedAnswer":{"@type":"Answer","text":"Typically only 2 to 3 iterations are needed for a looping critique-and-revise setup between two AI models to reach a strong output, rather than requiring many repeated passes. This applies to continuous improvement patterns where one model generates content and a second model critiques it before the first revises. Anyone tuning agent loop iteration counts can track practical AI agent techniques on daily.dev."}},{"@type":"Question","name":"What are the cost and latency tradeoffs of using a multi-agent critique loop instead of a single AI prompt?","acceptedAnswer":{"@type":"Answer","text":"A multi-agent critique loop roughly doubles both cost and latency compared to a single inference request, since the model must iterate with another model multiple times before producing a final answer. It works best for one-time or infrequent high-fidelity tasks rather than daily operating workloads, where the added cost and delay outweigh the benefit of a single well-tuned prompt. Teams weighing agent loop cost against single-prompt speed can follow AI agent design tradeoffs on daily.dev."}}]} ```