{"id":107681,"date":"2026-07-29T19:51:14","date_gmt":"2026-07-29T19:51:14","guid":{"rendered":"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/"},"modified":"2026-07-29T19:51:14","modified_gmt":"2026-07-29T19:51:14","slug":"prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer","status":"publish","type":"post","link":"https:\/\/youzum.net\/fr\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/","title":{"rendered":"Prompt Engineering vs Loop Engineering vs Graph Engineering: What Changes at Each Layer"},"content":{"rendered":"<p class=\"wp-block-paragraph\">Three terms now compete for the same line in AI engineering job descriptions. Prompt engineering is the established one. Loop engineering entered the AI vocabulary in late 2025 and dominated developer discussion through June 2026. Graph engineering followed roughly six weeks later.<\/p>\n<p class=\"wp-block-paragraph\">They get used interchangeably. <strong>Should they be?<\/strong><\/p>\n<p class=\"wp-block-paragraph\">The three are not competing techniques. They are three different <strong>units of control<\/strong>, stacked. A prompt controls one model response. A loop controls one agent\u2019s behavior cycle. A graph controls the organization of many agents. Each layer preserves the layer beneath it. A prompt does not disappear once a loop is built around it. it stops being the thing typed by hand.<\/p>\n<p class=\"wp-block-paragraph\">This article separates the three: what gets designed at each layer, what the published claim says about when the higher layers pay for themselves, and where the skepticism is warranted.<\/p>\n<p><title>One task, three layers<\/title><\/p>\n<div>\n<div class=\"hd\">\n<p class=\"eyebrow\">Interactive explainer \u00b7 01<\/p>\n<h2>One task, three layers \u2014 watch what changes<\/h2>\n<p>The three terms are not competing techniques. They are three different units of control. Here is the same job handled at each layer, step by step. Orange dots mark the moments a human is required.<\/p>\n<\/div>\n<div class=\"taskbar\">\n    <span class=\"tl\">Task<\/span><br \/>\n    <span class=\"tv\">Fix the failing tests in the auth module, then open a pull request.<\/span><br \/>\n    <button class=\"btn primary\">Run all three<\/button><br \/>\n    <button class=\"btn\">Reset<\/button>\n  <\/div>\n<div class=\"lanes\">\n<div class=\"lane p\">\n<div class=\"lh\">\n<p class=\"k\">Layer 1<\/p>\n<h3>Prompt engineering<\/h3>\n<p class=\"sub\">You control one model response. You are the loop.<\/p>\n<\/div>\n<ul class=\"steps\"><\/ul>\n<div class=\"lstat\">Your turns <b>0<\/b> \u00b7 Model calls <b>0<\/b><\/div>\n<\/div>\n<div class=\"lane l\">\n<div class=\"lh\">\n<p class=\"k\">Layer 2<\/p>\n<h3>Loop engineering<\/h3>\n<p class=\"sub\">You control one agent\u2019s cycle. The loop does the prompting.<\/p>\n<\/div>\n<ul class=\"steps\"><\/ul>\n<div class=\"lstat\">Your turns <b>0<\/b> \u00b7 Model calls <b>0<\/b><\/div>\n<\/div>\n<div class=\"lane g\">\n<div class=\"lh\">\n<p class=\"k\">Layer 3<\/p>\n<h3>Graph engineering<\/h3>\n<p class=\"sub\">You control how many agents are organised.<\/p>\n<\/div>\n<ul class=\"steps\"><\/ul>\n<div class=\"lstat\">Your turns <b>0<\/b> \u00b7 Nodes <b>0<\/b> \u00b7 Parallel <b>0<\/b><\/div>\n<\/div>\n<\/div>\n<div class=\"cmp\">\n<h4>What actually differs<\/h4>\n<table>\n<thead>\n<tr>\n<th>\u00a0<\/th>\n<th class=\"c-p\">Prompt<\/th>\n<th class=\"c-l\">Loop<\/th>\n<th class=\"c-g\">Graph<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<th>Unit of control<\/th>\n<td>One model response<\/td>\n<td>One agent\u2019s behaviour cycle<\/td>\n<td>An organisation of agents<\/td>\n<\/tr>\n<tr>\n<th>What you write<\/th>\n<td>Instructions, examples, output format<\/td>\n<td>Trigger, tools, stop condition, retry budget<\/td>\n<td>Nodes, edges, shared state, failure routes<\/td>\n<\/tr>\n<tr>\n<th>Who says \u201cagain?\u201d<\/th>\n<td>A human, every turn<\/td>\n<td>A verifier the loop calls itself<\/td>\n<td>A routing rule written in advance<\/td>\n<\/tr>\n<tr>\n<th>Where it breaks<\/th>\n<td>Ambiguous or overstuffed instruction<\/td>\n<td>It cannot tell done from stuck<\/td>\n<td>Context never crossed an edge you forgot to draw<\/td>\n<\/tr>\n<tr>\n<th>Enough when<\/th>\n<td>One shot, a person reads the result<\/td>\n<td>Repetitive, machine-checkable, one domain<\/td>\n<td>Cross-domain work with parallel branches<\/td>\n<\/tr>\n<\/tbody>\n<\/table><\/div>\n<div class=\"ft\">\n    <span>Illustrative walkthrough \u00b7 step counts are not benchmarks<\/span><br \/>\n    <span><b>Built by Marktechpost<\/b><\/span>\n  <\/div>\n<\/div>\n<p class=\"wp-block-paragraph\">\n<h2 class=\"wp-block-heading\"><strong>The stack, in order<\/strong><\/h2>\n<\/p><p class=\"wp-block-paragraph\">Each step in the progression was named in practice before it appeared in vendor documentation.<\/p>\n<ol class=\"wp-block-list\">\n<li><strong>Prompt engineering<\/strong> covers writing and structuring the instruction for a single call. <a href=\"https:\/\/www.anthropic.com\/engineering\/effective-context-engineering-for-ai-agents\">Anthropic\u2019s guidance<\/a> is to separate a system prompt into labeled sections \u2014 background information, instructions, tool guidance, output description \u2014 delineated with XML tags or Markdown headers. The recommendation is to supply the minimal set of information that fully specifies the expected behavior. Minimal does not mean short.<\/li>\n<li><strong>Context engineering<\/strong> came next. Anthropic describes it as the natural progression of prompt engineering. The question moves from finding the right words to deciding what configuration of tokens belongs in the window at all. Context is a finite resource, and the engineering problem is optimizing the utility of those tokens against model constraints.<\/li>\n<li><strong>Harness engineering<\/strong> covers the environment a single agent runs inside: files, tools, memory, feedback.<\/li>\n<li><strong>Loop engineering<\/strong> sits one floor above the harness. A June 2026 arXiv paper on agentic AI in building engineering, <a href=\"https:\/\/arxiv.org\/abs\/2606.25139\">Buildrix<\/a>, sets out the same four-step progression explicitly \u2014 prompt, then context, then harness, then loop \u2014 with the final layer defining how a system repeatedly observes, acts, verifies and recovers.<\/li>\n<li><strong>Graph engineering<\/strong> is the newest label and the least settled. <a href=\"https:\/\/www.truefoundry.com\/blog\/graph-engineering-enterprise-guide\">One enterprise writeup<\/a> notes that the term\u2019s provenance is unresolved and that it collides with an older knowledge-graph usage of the same word. The underlying practice, graph-based orchestration, has a documented lineage in multi-agent systems research.<\/li>\n<\/ol>\n<h3 class=\"wp-block-heading\"><strong>Layer 1: Prompt Engineering<\/strong><\/h3>\n<p class=\"wp-block-paragraph\">The defining assumption is that a human is present at every iteration. A prompt is written, the model responds, the output is judged, the prompt is revised.<\/p>\n<p class=\"wp-block-paragraph\">That assumption is what breaks. High volume. Multi-step tasks. No human available to grade the output. Results that feed the next step automatically. Any one of these, and the prompt alone stops being sufficient.<\/p>\n<p class=\"wp-block-paragraph\">Nothing about the prompt got worse. The surrounding conditions changed.<\/p>\n<p class=\"wp-block-paragraph\">Prompt engineering also does not vanish inside the higher layers. Anthropic\u2019s <a href=\"https:\/\/www.anthropic.com\/engineering\/multi-agent-research-system\">multi-agent research writeup<\/a> reports that prompt engineering was the <strong>primary lever<\/strong> for fixing coordination failures. Early versions spawned 50 subagents for simple queries, and the fix was prompting rather than topology.<\/p>\n<p><title>Prompt engineering, explained<\/title><\/p>\n<div>\n<div class=\"hd\">\n<p class=\"eyebrow\">Interactive explainer \u00b7 02 \u00b7 Layer 1<\/p>\n<h2>What prompt engineering is<\/h2>\n<p class=\"def\"><b>Prompt engineering is designing the text of a single call to a model.<\/b> One input, one forward pass, one output \u2014 and a human who reads the result and decides whether to run it again. That last part is the assumption everything above this layer exists to remove.<\/p>\n<\/div>\n<div class=\"flowwrap\">\n<div class=\"scroller\">\n      <svg viewbox=\"0 0 720 150\" role=\"img\">\n<path d=\"M0,0 L8,4 L0,8 z\" fill=\"#9BD7FF\"><\/path>\n<path d=\"M0,0 L8,4 L0,8 z\" fill=\"#FF9E7D\"><\/path>\n<p>        Prompt\n        what you write<\/p>\n<p>        Model\n        one forward pass<\/p>\n<p>        Response\n        one output<\/p>\n<p>        you<\/p>\n<path d=\"M680,84 C680,128 300,138 90,110\" fill=\"none\" stroke=\"#FF9E7D\" stroke-width=\"1.4\"><\/path>\n        you judge it, then rewrite the prompt yourself\n<p>      <\/p><\/svg>\n    <\/div>\n<p class=\"swipe\">Swipe the diagram sideways \u2192<\/p>\n<p class=\"cap\">The cycle exists here too \u2014 but you are the one closing it, every single turn.<\/p>\n<\/div>\n<div class=\"split\">\n<div class=\"col\">\n<h3>Anatomy of a prompt<\/h3>\n<p class=\"sub\">Tap a section. Anthropic\u2019s guidance is to delineate parts with XML tags or Markdown headers rather than writing one undifferentiated block.<\/p>\n<p>      <button class=\"blk on\" data-k=\"bg\"><span class=\"tag\">&lt;background_information&gt;<\/span><span class=\"ln\">What the model needs to know before it starts<\/span><\/button><br \/>\n      <button class=\"blk\" data-k=\"ins\"><span class=\"tag\">&lt;instructions&gt;<\/span><span class=\"ln\">The task itself, and the rules for doing it<\/span><\/button><br \/>\n      <button class=\"blk\" data-k=\"tool\"><span class=\"tag\">## Tool guidance<\/span><span class=\"ln\">Which tool to reach for, and when<\/span><\/button><br \/>\n      <button class=\"blk\" data-k=\"out\"><span class=\"tag\">## Output description<\/span><span class=\"ln\">The exact shape the answer should take<\/span><\/button><br \/>\n      <button class=\"blk\" data-k=\"ex\"><span class=\"tag\">&lt;examples&gt;<\/span><span class=\"ln\">A few cases that cover the edges you actually hit<\/span><\/button>\n    <\/p><\/div>\n<div class=\"col\">\n<h3>What this section does<\/h3>\n<h4>Background information<\/h4>\n\n<p class=\"quote\">\n<\/p><\/div>\n<\/div>\n<div class=\"brk\">\n<h3>The assumption that breaks<\/h3>\n<p class=\"sub\">Prompt engineering assumes <b>a human is present at every iteration<\/b>. Tap a condition to see what happens when that stops being true.<\/p>\n<div class=\"chips\">\n      <button class=\"chip on\" data-c=\"vol\">high volume<\/button><br \/>\n      <button class=\"chip\" data-c=\"multi\">multi-step task<\/button><br \/>\n      <button class=\"chip\" data-c=\"nobody\">nobody available to grade it<\/button><br \/>\n      <button class=\"chip\" data-c=\"chain\">output feeds the next step<\/button>\n    <\/div>\n<div class=\"conseq\"><\/div>\n<\/div>\n<div class=\"ft\">\n    <span>Layer 1 of 3 \u00b7 the layer most tasks stop at<\/span><br \/>\n    <span><b>Built by Marktechpost<\/b><\/span>\n  <\/div>\n<\/div>\n<p class=\"wp-block-paragraph\">\n<h3 class=\"wp-block-heading\"><strong>Layer 2: loop engineering<\/strong><\/h3>\n<\/p><p class=\"wp-block-paragraph\">The framing is that a coding agent is a brute-force tool for finding solutions. The craft is designing the goal, the tools and the loop and not only the prompt.<\/p>\n<p class=\"wp-block-paragraph\">The term reached mainstream developer discussion in June 2026, after a widely shared post argued that engineers should stop prompting coding agents and start designing the loops that prompt them. The Claude Code team at Anthropic described the same shift on stage that week.<\/p>\n<p class=\"wp-block-paragraph\">The <a href=\"https:\/\/addyosmani.com\/blog\/loop-engineering\/\">most detailed public breakdown<\/a> identifies five primitives, plus a sixth element that holds them together:<\/p>\n<ol class=\"wp-block-list\">\n<li><strong>Automations<\/strong>: a schedule or event that performs discovery and triage without supervision<\/li>\n<li><strong>Worktrees<\/strong>: isolation so parallel agents cannot edit the same files<\/li>\n<li><strong>Skills<\/strong>: project knowledge written down once in a <code>SKILL.md<\/code>, rather than re-explained every session<\/li>\n<li><strong>Plugins and connectors<\/strong>: MCP-based access to the issue tracker, database or staging API<\/li>\n<li><strong>Sub-agents<\/strong>: a maker\/checker split, since the model that wrote the code grades it too generously<\/li>\n<li><strong>State<\/strong>: a markdown file or board outside the conversation, because the model forgets between runs<\/li>\n<\/ol>\n<p class=\"wp-block-paragraph\">Two in-session features are very important. <code>\/loop<\/code> re-runs on a cadence. <code>\/goal<\/code> runs until a written condition is actually true, with a separate small model checking after each turn \u2014 so the agent that wrote the code is not the agent grading it. Both Claude Code and the Codex app ship equivalents.<\/p>\n<p class=\"wp-block-paragraph\">The cycle is not the hard part. The <strong>stop condition<\/strong> is. A loop that cannot mechanically distinguish done from stuck does not fail loudly. It keeps spending tokens.<\/p>\n<p><title>Loop engineering, explained<\/title><\/p>\n<div>\n<div class=\"hd\">\n<p class=\"eyebrow\">Interactive explainer \u00b7 03 \u00b7 Layer 2<\/p>\n<h2>What loop engineering is<\/h2>\n<p class=\"def\"><b>Loop engineering is designing the system that prompts the agent, instead of prompting it yourself.<\/b> The prompt does not disappear \u2014 it becomes a component inside a cycle that decides when to fire, what counts as passing, and when to stop trying. The skill was named in September 2025 and became the dominant developer topic in June 2026.<\/p>\n<\/div>\n<div class=\"band\">\n<div class=\"side\">\n<h3>The cycle<\/h3>\n<p class=\"sub\">Four phases and a memory. Press run to watch one complete, then force a failure to see why the guardrails exist.<\/p>\n<div class=\"ring\">\n        <svg viewbox=\"0 0 300 244\" role=\"img\">\n<p>          0\n          ITERATION<\/p>\n<p>          trigger\n          cron \/ event \/ goal<\/p>\n<p>          act\n          tools + edits<\/p>\n<p>          verify\n          tests \/ 2nd model<\/p>\n<p>          decide\n          retry or exit<\/p>\n<path d=\"M197,22 C242,30 262,60 253,100\" fill=\"none\" stroke=\"#274B4B\" stroke-width=\"1.2\"><\/path>\n          <path d=\"M253,140 C262,180 242,210 197,218\" fill=\"none\" stroke=\"#274B4B\" stroke-width=\"1.2\"><\/path>\n          <path d=\"M103,218 C58,210 38,180 47,140\" fill=\"none\" stroke=\"#274B4B\" stroke-width=\"1.2\"><\/path>\n          <path d=\"M47,100 C38,60 58,30 103,22\" fill=\"none\" stroke=\"#274B4B\" stroke-width=\"1.2\"><\/path>\n        <\/svg>\n      <\/div>\n<div class=\"btns\">\n        <button class=\"btn primary\">Run the loop<\/button><br \/>\n        <button class=\"btn\">Force a stall<\/button><br \/>\n        <button class=\"btn\">Reset<\/button>\n      <\/div>\n<\/div>\n<div class=\"side\">\n<h3>Run log<\/h3>\n<p class=\"sub\">Everything below happens with nobody watching. That is the whole point, and the whole risk.<\/p>\n<div class=\"log\" aria-live=\"polite\">\n<div class=\"t-dim\">Idle. Press \u201cRun the loop\u201d.<\/div>\n<\/div>\n<p class=\"verdict\">A loop needs a stopping rule it can evaluate without you. Without one it does not fail loudly \u2014 it just keeps spending.<\/p>\n<\/div>\n<\/div>\n<div class=\"prims\">\n<h3>What you actually build<\/h3>\n<p class=\"sub\">A working loop needs <b>five primitives plus a place to remember things<\/b>. Both Claude Code and the Codex app now ship all six. Tap one.<\/p>\n<div class=\"grid\">\n      <button class=\"pc on\" data-p=\"auto\"><span class=\"n\">01<\/span><span class=\"t\">Automations<\/span><\/button><br \/>\n      <button class=\"pc\" data-p=\"wt\"><span class=\"n\">02<\/span><span class=\"t\">Worktrees<\/span><\/button><br \/>\n      <button class=\"pc\" data-p=\"sk\"><span class=\"n\">03<\/span><span class=\"t\">Skills<\/span><\/button><br \/>\n      <button class=\"pc\" data-p=\"mcp\"><span class=\"n\">04<\/span><span class=\"t\">Connectors<\/span><\/button><br \/>\n      <button class=\"pc\" data-p=\"sub\"><span class=\"n\">05<\/span><span class=\"t\">Sub-agents<\/span><\/button><br \/>\n      <button class=\"pc\" data-p=\"st\"><span class=\"n\">06<\/span><span class=\"t\">State<\/span><\/button>\n    <\/div>\n<div class=\"pd\"><\/div>\n<\/div>\n<div class=\"ft\">\n    <span>Layer 2 of 3 \u00b7 illustrative simulation, not measured<\/span><br \/>\n    <span><b>Built by Marktechpost<\/b><\/span>\n  <\/div>\n<\/div>","protected":false},"excerpt":{"rendered":"<p>Three terms now compete for the same line in AI engineering job descriptions. Prompt engineering is the established one. Loop engineering entered the AI vocabulary in late 2025 and dominated developer discussion through June 2026. Graph engineering followed roughly six weeks later. They get used interchangeably. Should they be? The three are not competing techniques. They are three different units of control, stacked. A prompt controls one model response. A loop controls one agent\u2019s behavior cycle. A graph controls the organization of many agents. Each layer preserves the layer beneath it. A prompt does not disappear once a loop is built around it. it stops being the thing typed by hand. This article separates the three: what gets designed at each layer, what the published claim says about when the higher layers pay for themselves, and where the skepticism is warranted. One task, three layers Interactive explainer \u00b7 01 One task, three layers \u2014 watch what changes The three terms are not competing techniques. They are three different units of control. Here is the same job handled at each layer, step by step. Orange dots mark the moments a human is required. Task Fix the failing tests in the auth module, then open a pull request. Run all three Reset Layer 1 Prompt engineering You control one model response. You are the loop. Your turns 0 \u00b7 Model calls 0 Layer 2 Loop engineering You control one agent\u2019s cycle. The loop does the prompting. Your turns 0 \u00b7 Model calls 0 Layer 3 Graph engineering You control how many agents are organised. Your turns 0 \u00b7 Nodes 0 \u00b7 Parallel 0 What actually differs \u00a0 Prompt Loop Graph Unit of control One model response One agent\u2019s behaviour cycle An organisation of agents What you write Instructions, examples, output format Trigger, tools, stop condition, retry budget Nodes, edges, shared state, failure routes Who says \u201cagain?\u201d A human, every turn A verifier the loop calls itself A routing rule written in advance Where it breaks Ambiguous or overstuffed instruction It cannot tell done from stuck Context never crossed an edge you forgot to draw Enough when One shot, a person reads the result Repetitive, machine-checkable, one domain Cross-domain work with parallel branches Illustrative walkthrough \u00b7 step counts are not benchmarks Built by Marktechpost The stack, in order Each step in the progression was named in practice before it appeared in vendor documentation. Prompt engineering covers writing and structuring the instruction for a single call. Anthropic\u2019s guidance is to separate a system prompt into labeled sections \u2014 background information, instructions, tool guidance, output description \u2014 delineated with XML tags or Markdown headers. The recommendation is to supply the minimal set of information that fully specifies the expected behavior. Minimal does not mean short. Context engineering came next. Anthropic describes it as the natural progression of prompt engineering. The question moves from finding the right words to deciding what configuration of tokens belongs in the window at all. Context is a finite resource, and the engineering problem is optimizing the utility of those tokens against model constraints. Harness engineering covers the environment a single agent runs inside: files, tools, memory, feedback. Loop engineering sits one floor above the harness. A June 2026 arXiv paper on agentic AI in building engineering, Buildrix, sets out the same four-step progression explicitly \u2014 prompt, then context, then harness, then loop \u2014 with the final layer defining how a system repeatedly observes, acts, verifies and recovers. Graph engineering is the newest label and the least settled. One enterprise writeup notes that the term\u2019s provenance is unresolved and that it collides with an older knowledge-graph usage of the same word. The underlying practice, graph-based orchestration, has a documented lineage in multi-agent systems research. Layer 1: Prompt Engineering The defining assumption is that a human is present at every iteration. A prompt is written, the model responds, the output is judged, the prompt is revised. That assumption is what breaks. High volume. Multi-step tasks. No human available to grade the output. Results that feed the next step automatically. Any one of these, and the prompt alone stops being sufficient. Nothing about the prompt got worse. The surrounding conditions changed. Prompt engineering also does not vanish inside the higher layers. Anthropic\u2019s multi-agent research writeup reports that prompt engineering was the primary lever for fixing coordination failures. Early versions spawned 50 subagents for simple queries, and the fix was prompting rather than topology. Prompt engineering, explained Interactive explainer \u00b7 02 \u00b7 Layer 1 What prompt engineering is Prompt engineering is designing the text of a single call to a model. One input, one forward pass, one output \u2014 and a human who reads the result and decides whether to run it again. That last part is the assumption everything above this layer exists to remove. Prompt what you write Model one forward pass Response one output you you judge it, then rewrite the prompt yourself Swipe the diagram sideways \u2192 The cycle exists here too \u2014 but you are the one closing it, every single turn. Anatomy of a prompt Tap a section. Anthropic\u2019s guidance is to delineate parts with XML tags or Markdown headers rather than writing one undifferentiated block. &lt;background_information&gt;What the model needs to know before it starts &lt;instructions&gt;The task itself, and the rules for doing it ## Tool guidanceWhich tool to reach for, and when ## Output descriptionThe exact shape the answer should take &lt;examples&gt;A few cases that cover the edges you actually hit What this section does Background information The assumption that breaks Prompt engineering assumes a human is present at every iteration. Tap a condition to see what happens when that stops being true. high volume multi-step task nobody available to grade it output feeds the next step Layer 1 of 3 \u00b7 the layer most tasks stop at Built by Marktechpost Layer 2: loop engineering The framing is that a coding agent is a brute-force tool for finding solutions. The craft is designing the goal, the<\/p>","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"pmpro_default_level":"","site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"_pvb_checkbox_block_on_post":false,"footnotes":""},"categories":[52,5,7,1],"tags":[],"class_list":["post-107681","post","type-post","status-publish","format-standard","hentry","category-ai-club","category-committee","category-news","category-uncategorized","pmpro-has-access"],"acf":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v25.3 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Prompt Engineering vs Loop Engineering vs Graph Engineering: What Changes at Each Layer - YouZum<\/title>\n<meta name=\"description\" content=\"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/youzum.net\/fr\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Prompt Engineering vs Loop Engineering vs Graph Engineering: What Changes at Each Layer - YouZum\" \/>\n<meta property=\"og:description\" content=\"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19\" \/>\n<meta property=\"og:url\" content=\"https:\/\/youzum.net\/fr\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/\" \/>\n<meta property=\"og:site_name\" content=\"YouZum\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/DroneAssociationTH\/\" \/>\n<meta property=\"article:published_time\" content=\"2026-07-29T19:51:14+00:00\" \/>\n<meta name=\"author\" content=\"admin NU\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u00c9crit par\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin NU\" \/>\n\t<meta name=\"twitter:label2\" content=\"Dur\u00e9e de lecture estim\u00e9e\" \/>\n\t<meta name=\"twitter:data2\" content=\"7 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/#article\",\"isPartOf\":{\"@id\":\"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/\"},\"author\":{\"name\":\"admin NU\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c\"},\"headline\":\"Prompt Engineering vs Loop Engineering vs Graph Engineering: What Changes at Each Layer\",\"datePublished\":\"2026-07-29T19:51:14+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/\"},\"wordCount\":1413,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#organization\"},\"articleSection\":[\"AI\",\"Committee\",\"News\",\"Uncategorized\"],\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/\",\"url\":\"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/\",\"name\":\"Prompt Engineering vs Loop Engineering vs Graph Engineering: What Changes at Each Layer - YouZum\",\"isPartOf\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#website\"},\"datePublished\":\"2026-07-29T19:51:14+00:00\",\"description\":\"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19\",\"breadcrumb\":{\"@id\":\"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/#breadcrumb\"},\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/youzum.net\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Prompt Engineering vs Loop Engineering vs Graph Engineering: What Changes at Each Layer\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/yousum.gpucore.co\/#website\",\"url\":\"https:\/\/yousum.gpucore.co\/\",\"name\":\"YouSum\",\"description\":\"\",\"publisher\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/yousum.gpucore.co\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"fr-FR\"},{\"@type\":\"Organization\",\"@id\":\"https:\/\/yousum.gpucore.co\/#organization\",\"name\":\"Drone Association Thailand\",\"url\":\"https:\/\/yousum.gpucore.co\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/\",\"url\":\"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png\",\"contentUrl\":\"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png\",\"width\":300,\"height\":300,\"caption\":\"Drone Association Thailand\"},\"image\":{\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/\"},\"sameAs\":[\"https:\/\/www.facebook.com\/DroneAssociationTH\/\"]},{\"@type\":\"Person\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c\",\"name\":\"admin NU\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/yousum.gpucore.co\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png\",\"contentUrl\":\"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png\",\"caption\":\"admin NU\"},\"url\":\"https:\/\/youzum.net\/fr\/members\/adminnu\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Prompt Engineering vs Loop Engineering vs Graph Engineering: What Changes at Each Layer - YouZum","description":"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/youzum.net\/fr\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/","og_locale":"fr_FR","og_type":"article","og_title":"Prompt Engineering vs Loop Engineering vs Graph Engineering: What Changes at Each Layer - YouZum","og_description":"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19","og_url":"https:\/\/youzum.net\/fr\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/","og_site_name":"YouZum","article_publisher":"https:\/\/www.facebook.com\/DroneAssociationTH\/","article_published_time":"2026-07-29T19:51:14+00:00","author":"admin NU","twitter_card":"summary_large_image","twitter_misc":{"\u00c9crit par":"admin NU","Dur\u00e9e de lecture estim\u00e9e":"7 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/#article","isPartOf":{"@id":"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/"},"author":{"name":"admin NU","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c"},"headline":"Prompt Engineering vs Loop Engineering vs Graph Engineering: What Changes at Each Layer","datePublished":"2026-07-29T19:51:14+00:00","mainEntityOfPage":{"@id":"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/"},"wordCount":1413,"commentCount":0,"publisher":{"@id":"https:\/\/yousum.gpucore.co\/#organization"},"articleSection":["AI","Committee","News","Uncategorized"],"inLanguage":"fr-FR","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/#respond"]}]},{"@type":"WebPage","@id":"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/","url":"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/","name":"Prompt Engineering vs Loop Engineering vs Graph Engineering: What Changes at Each Layer - YouZum","isPartOf":{"@id":"https:\/\/yousum.gpucore.co\/#website"},"datePublished":"2026-07-29T19:51:14+00:00","description":"\u0e01\u0e34\u0e08\u0e01\u0e23\u0e23\u0e21\u0e40\u0e01\u0e35\u0e48\u0e22\u0e27\u0e01\u0e31\u0e1a\u0e42\u0e14\u0e23\u0e19","breadcrumb":{"@id":"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/#breadcrumb"},"inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/youzum.net\/prompt-engineering-vs-loop-engineering-vs-graph-engineering-what-changes-at-each-layer\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/youzum.net\/"},{"@type":"ListItem","position":2,"name":"Prompt Engineering vs Loop Engineering vs Graph Engineering: What Changes at Each Layer"}]},{"@type":"WebSite","@id":"https:\/\/yousum.gpucore.co\/#website","url":"https:\/\/yousum.gpucore.co\/","name":"YouSum","description":"","publisher":{"@id":"https:\/\/yousum.gpucore.co\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/yousum.gpucore.co\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"fr-FR"},{"@type":"Organization","@id":"https:\/\/yousum.gpucore.co\/#organization","name":"Drone Association Thailand","url":"https:\/\/yousum.gpucore.co\/","logo":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/","url":"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png","contentUrl":"https:\/\/youzum.net\/wp-content\/uploads\/2024\/11\/tranparent-logo.png","width":300,"height":300,"caption":"Drone Association Thailand"},"image":{"@id":"https:\/\/yousum.gpucore.co\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.facebook.com\/DroneAssociationTH\/"]},{"@type":"Person","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/person\/97fa48242daf3908e4d9a5f26f4a059c","name":"admin NU","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/yousum.gpucore.co\/#\/schema\/person\/image\/","url":"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png","contentUrl":"https:\/\/youzum.net\/wp-content\/uploads\/avatars\/2\/1746849356-bpfull.png","caption":"admin NU"},"url":"https:\/\/youzum.net\/fr\/members\/adminnu\/"}]}},"rttpg_featured_image_url":null,"rttpg_author":{"display_name":"admin NU","author_link":"https:\/\/youzum.net\/fr\/members\/adminnu\/"},"rttpg_comment":0,"rttpg_category":"<a href=\"https:\/\/youzum.net\/fr\/category\/ai-club\/\" rel=\"category tag\">AI<\/a> <a href=\"https:\/\/youzum.net\/fr\/category\/committee\/\" rel=\"category tag\">Committee<\/a> <a href=\"https:\/\/youzum.net\/fr\/category\/news\/\" rel=\"category tag\">News<\/a> <a href=\"https:\/\/youzum.net\/fr\/category\/uncategorized\/\" rel=\"category tag\">Uncategorized<\/a>","rttpg_excerpt":"Three terms now compete for the same line in AI engineering job descriptions. Prompt engineering is the established one. Loop engineering entered the AI vocabulary in late 2025 and dominated developer discussion through June 2026. Graph engineering followed roughly six weeks later. They get used interchangeably. Should they be? The three are not competing techniques.\u2026","_links":{"self":[{"href":"https:\/\/youzum.net\/fr\/wp-json\/wp\/v2\/posts\/107681","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/youzum.net\/fr\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/youzum.net\/fr\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/youzum.net\/fr\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/youzum.net\/fr\/wp-json\/wp\/v2\/comments?post=107681"}],"version-history":[{"count":0,"href":"https:\/\/youzum.net\/fr\/wp-json\/wp\/v2\/posts\/107681\/revisions"}],"wp:attachment":[{"href":"https:\/\/youzum.net\/fr\/wp-json\/wp\/v2\/media?parent=107681"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/youzum.net\/fr\/wp-json\/wp\/v2\/categories?post=107681"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/youzum.net\/fr\/wp-json\/wp\/v2\/tags?post=107681"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}