<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/understanding-the-modern-ai-stack-ebantprlw" -->

---
title: Understanding the Modern AI Stack | daily.dev
description: A conceptual framework maps modern AI engineering vocabulary onto familiar full-stack development concepts, organizing terms into five layers: model (LLMs as...
canonical: https://daily.dev/posts/understanding-the-modern-ai-stack-ebantprlw
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: Understanding the Modern AI Stack | daily.dev
og:description: A conceptual framework maps modern AI engineering vocabulary onto familiar full-stack development concepts, organizing terms into five layers: model (LLMs as...
og:url: https://daily.dev/posts/understanding-the-modern-ai-stack-ebantprlw
og:image: https://api.daily.dev/og/posts/EbANtPrLw.png
og:image:alt: Understanding the Modern AI Stack
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.

# Understanding the Modern AI Stack

**[Telerik](https://daily.dev/sources/telerik)** · 9 min read · 0 upvotes · 0 comments

## Summary

A conceptual framework maps modern AI engineering vocabulary onto familiar full-stack development concepts, organizing terms into five layers: model (LLMs as stateless API runtimes), knowledge (embeddings, vector databases, RAG), action (agents, tools, memory), coordination (orchestration), and trust (evaluation, observability). Using a fictional customer support assistant called TaskFlow as a running example, the piece explains how LLMs are stateless with limited context windows, how RAG grounds responses in real documentation, how tool use and MCP let agents act on systems, how orchestration frameworks like LangGraph coordinate multi-step flows, and how evaluation sets and observability tracing replace traditional deterministic testing and monitoring for non-deterministic AI outputs.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://www.telerik.com/blogs/understanding-modern-ai-stack>

## Questions this post answers

### What is the difference between short-term and long-term memory in an AI agent system?

Short-term memory is the running conversation history resent with each request so the model can follow references like 'move those ones too,' while long-term memory persists across sessions in separate storage the application controls, such as remembering a user's sprint length preference and retrieving that fact into context at the right moment later.

_Developers designing agent memory systems can follow ongoing AI engineering explainers on daily.dev._

### Why can't a large language model remember previous conversations on its own?

A large language model is stateless and retains nothing between requests, so any information it needs, such as conversation history, a user's data, or relevant documentation, must be reassembled into each new request. Some provider APIs, like OpenAI's Responses API with a previous_response_id parameter, handle this bookkeeping server-side, but the full history is still fed to the model and billed as input tokens each call.

_Anyone architecting chat memory around stateless models can track these patterns via daily.dev._

### How do you test or evaluate an AI system when its outputs are not deterministic?

Traditional assertions like expect(output).toBe(expected) don't work because the same input can produce different, equally valid responses, so teams build evaluation sets instead: collections of representative inputs paired with criteria for judging outputs via string checks, rules, or another model acting as a grader. These evaluations run whenever a prompt changes, a model is swapped, or retrieval is adjusted, serving the role a regression suite serves in conventional software.

_Teams building non-deterministic system evaluations can follow related AI engineering practices on daily.dev._

## Similar posts on daily.dev

- [The AI Stack in 2026: 11 Layers Behind Production-Ready AI Systems](https://daily.dev/posts/the-ai-stack-in-2026-11-layers-behind-production-ready-ai-systems-jdcq77ovf) · Medium · 1 upvotes · 0 comments
- [The AI Agent Tech Stack Explained](https://daily.dev/posts/the-ai-agent-tech-stack-explained-yfyxgvmzt) · Machine Learning Mastery · 6 upvotes · 2 comments
- [The AI Engineering Master Stack for 2026\!](https://daily.dev/posts/the-ai-engineering-master-stack-for-2026--a21pct7m4) · Daily Dose of Data Science \| Avi Chawla \| Substack · 4 upvotes · 0 comments

---

Tags: [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#mcp](https://daily.dev/tags/mcp), [#rag](https://daily.dev/tags/rag), [#vector-search](https://daily.dev/tags/vector-search)

[View this post on daily.dev](https://daily.dev/posts/understanding-the-modern-ai-stack-ebantprlw)

```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":"Understanding the Modern AI Stack","url":"https://daily.dev/posts/understanding-the-modern-ai-stack-ebantprlw","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/understanding-the-modern-ai-stack-ebantprlw"},"datePublished":"2026-08-25T15:28:07.500Z","dateModified":"2026-08-25T15:37:43.765Z","description":"A conceptual framework maps modern AI engineering vocabulary onto familiar full-stack development concepts, organizing terms into five layers: model (LLMs as...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/b83e6573037525bc78570b48599b4cd0?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/b83e6573037525bc78570b48599b4cd0?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Telerik","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":"Telerik","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/5ee6d4bef8ac4f84886d06f5881963e7","url":"https://daily.dev/sources/telerik"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/understanding-the-modern-ai-stack-ebantprlw","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,ai-agents,mcp,rag,vector-search","timeRequired":"PT9M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Telerik","item":"https://daily.dev/sources/telerik"},{"@type":"ListItem","position":3,"name":"Understanding the Modern AI Stack"}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/understanding-the-modern-ai-stack-ebantprlw#faq","mainEntity":[{"@type":"Question","name":"What is the difference between short-term and long-term memory in an AI agent system?","acceptedAnswer":{"@type":"Answer","text":"Short-term memory is the running conversation history resent with each request so the model can follow references like 'move those ones too,' while long-term memory persists across sessions in separate storage the application controls, such as remembering a user's sprint length preference and retrieving that fact into context at the right moment later. Developers designing agent memory systems can follow ongoing AI engineering explainers on daily.dev."}},{"@type":"Question","name":"Why can't a large language model remember previous conversations on its own?","acceptedAnswer":{"@type":"Answer","text":"A large language model is stateless and retains nothing between requests, so any information it needs, such as conversation history, a user's data, or relevant documentation, must be reassembled into each new request. Some provider APIs, like OpenAI's Responses API with a previous_response_id parameter, handle this bookkeeping server-side, but the full history is still fed to the model and billed as input tokens each call. Anyone architecting chat memory around stateless models can track these patterns via daily.dev."}},{"@type":"Question","name":"How do you test or evaluate an AI system when its outputs are not deterministic?","acceptedAnswer":{"@type":"Answer","text":"Traditional assertions like expect(output).toBe(expected) don't work because the same input can produce different, equally valid responses, so teams build evaluation sets instead: collections of representative inputs paired with criteria for judging outputs via string checks, rules, or another model acting as a grader. These evaluations run whenever a prompt changes, a model is swapped, or retrieval is adjusted, serving the role a regression suite serves in conventional software. Teams building non-deterministic system evaluations can follow related AI engineering practices on daily.dev."}}]}
```

