--- title: AI agent tradeoffs: what evals catch and reading traces... description: A developer built a conference schedule assistant using an AI agent with two model tiers — a cheap open model for anonymous users and a frontier model for... canonical: https://daily.dev/posts/ai-agent-tradeoffs-what-evals-catch-and-reading-traces-reveal-tzdqlhlxk twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AI agent tradeoffs: what evals catch and reading traces reveal | daily.dev og:description: A developer built a conference schedule assistant using an AI agent with two model tiers — a cheap open model for anonymous users and a frontier model for... og:url: https://daily.dev/posts/ai-agent-tradeoffs-what-evals-catch-and-reading-traces-reveal-tzdqlhlxk og:image: https://api.daily.dev/og/posts/tZdQlHlXk.png og:image:alt: AI agent tradeoffs: what evals catch and reading traces reveal 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. # AI agent tradeoffs: what evals catch and reading traces reveal **[Sentry Engineering](https://daily.dev/sources/sentry)** · 10 min read · 0 upvotes · 0 comments ## Summary A developer built a conference schedule assistant using an AI agent with two model tiers — a cheap open model for anonymous users and a frontier model for signed-up users. The cheap model hallucinated famous speakers (including a non-existent person) from tool output that contained no speaker data at all, then confidently fabricated a source for the invented names. The post walks through how reading agent traces — not just eval scores — revealed the failure: a tool ran, so the answer looked grounded, but the tool returned tracks, not speakers. The author outlines a groundedness eval (asserting every named entity appears in tool output or the database) that would have caught this in CI, and explains the menu of fixes considered: better model, fixed routing, tighter prompt, or accepting the failure for anonymous traffic. The key lesson is that evals catch known regressions but can't make cost-vs-quality tradeoff decisions, and that reading raw traces builds the intuition needed to know what evals to write in the first place. Sentry's MCP integration is mentioned as a way to surface relevant traces programmatically. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Similar posts on daily.dev - [Debugging our AI search assistant with agent tracing](https://daily.dev/posts/debugging-our-ai-search-assistant-with-agent-tracing-zjgtgjbcw) · Sentry Engineering · 1 upvotes · 0 comments - [AI agent evaluations are part of the product](https://daily.dev/posts/ai-agent-evaluations-are-part-of-the-product-u3ha51aoc) · The New Stack · 0 upvotes · 0 comments - [I shipped an AI Agent. It broke in 3 days. Here’s the eval I should have written first.](https://daily.dev/posts/i-shipped-an-ai-agent-it-broke-in-3-days-here-s-the-eval-i-should-have-written-first--pq4qs5n7v) · gitconnected · 7 upvotes · 2 comments - [Making AI work through eval hygiene](https://daily.dev/posts/making-ai-work-through-eval-hygiene-9yhow3tph) · InfoWorld · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#observability](https://daily.dev/tags/observability) [View this post on daily.dev](https://daily.dev/posts/ai-agent-tradeoffs-what-evals-catch-and-reading-traces-reveal-tzdqlhlxk) ```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":"AI agent tradeoffs: what evals catch and reading traces reveal","url":"https://daily.dev/posts/ai-agent-tradeoffs-what-evals-catch-and-reading-traces-reveal-tzdqlhlxk","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-agent-tradeoffs-what-evals-catch-and-reading-traces-reveal-tzdqlhlxk"},"datePublished":"2026-07-09T06:35:55.207Z","dateModified":"2026-07-09T06:36:42.644Z","description":"A developer built a conference schedule assistant using an AI agent with two model tiers — a cheap open model for anonymous users and a frontier model for...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/af48bf930a4d5e313d7b0e9154a2aa85?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/af48bf930a4d5e313d7b0e9154a2aa85?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Sentry Engineering","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":"Sentry Engineering","logo":"https://media.daily.dev/image/upload/s--qqXwMThM--/f_auto/v1710505439/logos/sentry","url":"https://daily.dev/sources/sentry"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ai-agent-tradeoffs-what-evals-catch-and-reading-traces-reveal-tzdqlhlxk","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,observability","timeRequired":"PT10M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Sentry Engineering","item":"https://daily.dev/sources/sentry"},{"@type":"ListItem","position":3,"name":"AI agent tradeoffs: what evals catch and reading traces reveal"}]} ```