--- title: Writing Events Is All You Need | daily.dev description: A benchmark comparing three data storage models for AI agent reasoning: PostgreSQL with current state only (28%), PostgreSQL with CDC (59%), and KurrentDB with... canonical: https://daily.dev/posts/writing-events-is-all-you-need-rjvxyoggv twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Writing Events Is All You Need | daily.dev og:description: A benchmark comparing three data storage models for AI agent reasoning: PostgreSQL with current state only (28%), PostgreSQL with CDC (59%), and KurrentDB with... og:url: https://daily.dev/posts/writing-events-is-all-you-need-rjvxyoggv og:image: https://api.daily.dev/og/posts/rJvxyogGV.png og:image:alt: Writing Events Is All You Need 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. # Writing Events Is All You Need **[Event Store](https://daily.dev/sources/eventstore)** · 6 min read · 0 upvotes · 0 comments ## Summary A benchmark comparing three data storage models for AI agent reasoning: PostgreSQL with current state only (28%), PostgreSQL with CDC (59%), and KurrentDB with full event sourcing (85%). Tested across 300 questions in five tiers — from simple current-state queries to adversarial false-premise detection — across five domains. The core finding: the bottleneck for agentic AI accuracy is not the model but the data format. State-only storage destroys intent at write time, making it unrecoverable at read time. Event sourcing, where each atomic unit is an immutable named event with full context, gives agents everything needed to answer historical, causal, and temporal questions. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How does event sourcing compare to CDC and plain state storage for AI agent accuracy? Event sourcing scored 85% accuracy on a 300-question benchmark, compared to 59% for change data capture and 28% for state-only storage. The test used PostgreSQL for state-only and CDC models and KurrentDB for the event-sourced model, across five domains including e-commerce, banking, and software development, queried via each database's official MCP server. _Weighing event sourcing against CDC for agent-facing data pipelines is easier with daily.dev's engineering coverage._ ### Why do AI agents fail to explain why a database record changed when using change data capture? CDC records before-and-after field values but not the reason a change occurred, so agents can see that a status changed but not why. In the benchmark, CDC scored around 70% on change-history questions but roughly 0% on questions requiring intent or reasoning, since only named events like OrderCancelledByCustomer carry the causal context needed to answer 'why' questions. _Developers picking a data model for agent workflows can track findings like this on daily.dev._ ### What is the five-tier framework used to test whether a database can support AI agent reasoning? The framework tests current state (Tier 1), change history (Tier 2), intent and reason (Tier 3), cross-event temporal ordering (Tier 4), and adversarial false-premise detection (Tier 5). PostgreSQL alone passes only Tier 1, CDC passes Tiers 1-2, and only a full event stream store passed all five tiers in the 300-question benchmark across five real-world domains. _daily.dev helps engineers evaluating agent-ready data architectures follow benchmarks like this one._ ## Similar posts on daily.dev - [We'll End Up Where We Started. Except with Different Technology.](https://daily.dev/posts/we-ll-end-up-where-we-started-except-with-different-technology--5h7gm44uu) · Event Store · 0 upvotes · 0 comments - [Event Sourcing: Contextual AI Requires a Contextual Data Model](https://daily.dev/posts/event-sourcing-contextual-ai-requires-a-contextual-data-model-jzrwepa6o) · Event Store · 0 upvotes · 0 comments - [The hidden reason your AI assistant feels so sluggish](https://daily.dev/posts/the-hidden-reason-your-ai-assistant-feels-so-sluggish-v9jbcylq0) · The New Stack · 0 upvotes · 0 comments - [AI Infrastructure Starts With Data: Why Databases Are Critical in the Agentic AI Era](https://daily.dev/posts/ai-infrastructure-starts-with-data-why-databases-are-critical-in-the-agentic-ai-era-fv6einegp) · BigData Boutique blog · 2 upvotes · 0 comments - [Agent State Management: Redis vs Postgres for AI Memory](https://daily.dev/posts/agent-state-management-redis-vs-postgres-for-ai-memory-6f36iuftz) · SitePoint · 1 upvotes · 0 comments --- Tags: [#architecture](https://daily.dev/tags/architecture), [#postgresql](https://daily.dev/tags/postgresql), [#agentic-ai](https://daily.dev/tags/agentic-ai), [#change-data-capture](https://daily.dev/tags/change-data-capture) [View this post on daily.dev](https://daily.dev/posts/writing-events-is-all-you-need-rjvxyoggv) ```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":"Writing Events Is All You Need","url":"https://daily.dev/posts/writing-events-is-all-you-need-rjvxyoggv","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/writing-events-is-all-you-need-rjvxyoggv"},"datePublished":"2026-08-04T14:30:30.971Z","dateModified":"2026-09-14T06:40:22.845Z","description":"A benchmark comparing three data storage models for AI agent reasoning: PostgreSQL with current state only (28%), PostgreSQL with CDC (59%), and KurrentDB with...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/f5a2d2b5d110575876eac61573af6e71?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/f5a2d2b5d110575876eac61573af6e71?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Event Store","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":"Event Store","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/6e7940e0a29e4b3cbfef11831ec252c2","url":"https://daily.dev/sources/eventstore"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/writing-events-is-all-you-need-rjvxyoggv","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"architecture,postgresql,agentic-ai,change-data-capture","timeRequired":"PT6M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Event Store","item":"https://daily.dev/sources/eventstore"},{"@type":"ListItem","position":3,"name":"Writing Events Is All You Need"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/writing-events-is-all-you-need-rjvxyoggv#faq","mainEntity":[{"@type":"Question","name":"How does event sourcing compare to CDC and plain state storage for AI agent accuracy?","acceptedAnswer":{"@type":"Answer","text":"Event sourcing scored 85% accuracy on a 300-question benchmark, compared to 59% for change data capture and 28% for state-only storage. The test used PostgreSQL for state-only and CDC models and KurrentDB for the event-sourced model, across five domains including e-commerce, banking, and software development, queried via each database's official MCP server. Weighing event sourcing against CDC for agent-facing data pipelines is easier with daily.dev's engineering coverage."}},{"@type":"Question","name":"Why do AI agents fail to explain why a database record changed when using change data capture?","acceptedAnswer":{"@type":"Answer","text":"CDC records before-and-after field values but not the reason a change occurred, so agents can see that a status changed but not why. In the benchmark, CDC scored around 70% on change-history questions but roughly 0% on questions requiring intent or reasoning, since only named events like OrderCancelledByCustomer carry the causal context needed to answer 'why' questions. Developers picking a data model for agent workflows can track findings like this on daily.dev."}},{"@type":"Question","name":"What is the five-tier framework used to test whether a database can support AI agent reasoning?","acceptedAnswer":{"@type":"Answer","text":"The framework tests current state (Tier 1), change history (Tier 2), intent and reason (Tier 3), cross-event temporal ordering (Tier 4), and adversarial false-premise detection (Tier 5). PostgreSQL alone passes only Tier 1, CDC passes Tiers 1-2, and only a full event stream store passed all five tiers in the 300-question benchmark across five real-world domains. daily.dev helps engineers evaluating agent-ready data architectures follow benchmarks like this one."}}]} ```