[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"nav-notes-count":3,"content-note-multi-agent-ai-useful-pattern-or-extra-complexity":4},9,{"id":5,"title":6,"author":7,"authorAvatar":8,"authorRole":9,"body":10,"category":325,"date":326,"description":187,"excerpt":327,"extension":328,"image":329,"index":330,"meta":331,"navigation":345,"path":346,"pullQuote":347,"readingTime":348,"seo":349,"slug":350,"stem":351,"takeaways":352,"total":356,"__hash__":357},"notes\u002Fnotes\u002Fmulti-agent-ai-useful-pattern-or-extra-complexity.md","Multi-Agent AI: Useful Pattern or Extra Complexity?","Norbert Brett","\u002Fimages\u002Fnorbert-brett.jpg","Full-Stack & AI Systems Engineer",{"type":11,"value":12,"toc":320},"minimark",[13,18,32,83,97,104,156,160,163,180,190,207,220],[14,15,17],"h2",{"id":16},"when-multi-agent-ai-actually-earns-its-complexity","When Multi-Agent AI Actually Earns Its Complexity",[19,20,21,22,26,27,31],"p",{},"If you have ever watched a demo of four AI agents chatting with each other for five minutes to solve a task that one well-written prompt or a short script could finish in ten seconds, you understand why businesses are cautious about \"multi-agent\" hype. Months earlier, while prototyping the first version of ",[23,24,25],"strong",{},"Founder Suite"," (my 11-stage business validation pipeline), I was annoyed by the opposite extreme too: when I asked a ",[28,29,30],"em",{},"single"," AI session to simultaneously act as an enthusiastic product marketer, a skeptical market strategist, and a strict CFO, it averaged all three voices into polite, wishy-washy encouragement.",[33,34,35,40],"blockquote",{},[19,36,37],{},[23,38,39],{},"In Plain English — Single Agent vs. Multi-Agent vs. Deterministic Script:",[41,42,43,59,69],"ul",{},[44,45,46,49,50,54,55,58],"li",{},[23,47,48],{},"Deterministic Script (Python \u002F TypeScript Code)",": Traditional software rules where ",[51,52,53],"code",{},"2 + 2"," always equals ",[51,56,57],{},"4"," in under a millisecond. Zero guesswork, zero AI token cost.",[44,60,61,68],{},[23,62,63,64,67],{},"Single-Agent Mode (",[51,65,66],{},"solo",")",": One AI conversation handling a focused task from start to finish—like asking one experienced editor to polish a two-page memo.",[44,70,71,74,75,78,79,82],{},[23,72,73],{},"Multi-Agent System",": Splitting a large job across separate, specialized AI workers that either run ",[23,76,77],{},"at the same time in parallel"," (like 10 assistants each summarizing one chapter of a book) or ",[23,80,81],{},"check each other's work"," with fresh eyes.",[19,84,85,86,89,90,93,94,96],{},"Across the three AI systems I maintain in production (",[23,87,88],{},"Gemini Advisor",", ",[23,91,92],{},"Book-to-Skill",", and ",[23,95,25],{},"), adding multi-agent coordination is only worth the extra complexity when at least one of these three conditions is true:",[19,98,99],{},[100,101],"img",{"alt":102,"src":103},"Multi-Agent vs Solo vs Deterministic Python Decision Matrix","\u002Fimages\u002Fnotes\u002Fdiagrams\u002Fmulti-agent-ai-useful-pattern-or-extra-complexity.svg",[105,106,107,116,133],"ol",{},[44,108,109,112,113,115],{},[23,110,111],{},"Fresh Eyes for Quality Control (Context Isolation)",": In ",[23,114,88],{},", the reviewer agent is launched in a separate session so it never sees the builder's 50,000 tokens of debugging struggle. Because it only inspects the original contract and the final code diff in read-only mode, it catches shortcuts without bias.",[44,117,118,121,122,124,125,128,129,132],{},[23,119,120],{},"Parallel Speed on Independent Chunks",": When ",[23,123,92],{}," turns a 500-page technical manual into a modular AI skill, Chapter 2 and Chapter 12 do not depend on each other. Launching parallel waves of fast ",[51,126,127],{},"Flash"," subagents finishes the entire book in ",[23,130,131],{},"under 4 minutes"," instead of 45 minutes one by one.",[44,134,135,138,139,141,142,145,146,89,149,93,152,155],{},[23,136,137],{},"Genuine Disagreement Across Different Perspectives",": In stage one of ",[23,140,25],{}," (",[51,143,144],{},"founder-board","), three independent subagents critique a business idea at the same time through three strict, published lenses (",[28,147,148],{},"Offers",[28,150,151],{},"Monopoly",[28,153,154],{},"Product",") before their memos are compared. Because they cannot see each other's drafts while writing, they surface real blind spots instead of polite groupthink.",[14,157,159],{"id":158},"when-a-single-model-or-a-python-script-is-the-better-choice","When a Single Model or a Python Script Is the Better Choice",[19,161,162],{},"You see, the most expensive mistake in AI engineering today is paying language models to do work that regular code already does perfectly.",[19,164,165,166,168,169,171,172,175,176,179],{},"When I built ",[23,167,92],{}," and ",[23,170,25],{},", I drew a strict line between ",[23,173,174],{},"qualitative judgment"," (where AI agents help) and ",[23,177,178],{},"exact calculation"," (where AI models guess and hallucinate):",[181,182,188],"pre",{"className":183,"code":185,"language":186,"meta":187},[184],"language-text","Book-to-Skill Pipeline:\n  1. scripts\u002Fextract.py (Python)          -> Exact PDF\u002FEPUB parsing, SHA-256 fingerprint, chapter token counts\n  2. Parallel Flash Subagents             -> Summarize chapter frameworks & decision rules in parallel\n  3. tools\u002Fscan_generated_skill.py (Py)   -> Deterministic security & copyright-leak check\n\nFounder Validation Suite:\n  1. 3x Board Subagents + Flash Buyer Wave -> Surface qualitative positioning & customer objections\n  2. van_westendorp.py & unit_economics.py -> Exact price sensitivity bands & break-even unit math\n","text","",[51,189,185],{"__ignoreMap":187},[19,191,192,193,196,197,89,200,93,203,206],{},"Ask an AI model to calculate profit margins, count exact words in a 200-page PDF, or compute a 36-month cash-flow table, and it will invent numbers that ",[28,194,195],{},"look"," right but fail on a calculator. I felt relieved the day I stripped every arithmetic question out of our AI prompts and moved them into deterministic Python tools (",[51,198,199],{},"extract.py",[51,201,202],{},"unit_economics.py",[51,204,205],{},"van_westendorp.py",").",[19,208,209,210,213,214],{},"And for everyday software development, ",[23,211,212],{},"Rule #1"," in Gemini Advisor is simple: ",[23,215,216,217,219],{},"Default to ",[51,218,66],{}," (a single model) for small, low-risk tasks.",[221,222,223,240],"table",{},[224,225,226],"thead",{},[227,228,229,234,237],"tr",{},[230,231,233],"th",{"align":232},"left","What You Need Done",[230,235,236],{"align":232},"Best Tool for the Job",[230,238,239],{"align":232},"Why It Wins",[241,242,243,266,287,305],"tbody",{},[227,244,245,251,256],{},[246,247,248],"td",{"align":232},[23,249,250],{},"Financial math, price bands, file hashing, or syntax checks",[246,252,253],{"align":232},[23,254,255],{},"Deterministic Python \u002F TypeScript Script",[246,257,258,261,262,265],{"align":232},[51,259,260],{},"0"," hallucinations, runs in ",[51,263,264],{},"\u003C50ms",", and gives the exact same answer every time.",[227,267,268,276,284],{},[246,269,270],{"align":232},[23,271,272,273,67],{},"Updating 1–2 files with clear instructions (",[51,274,275],{},"risk=\"low\"",[246,277,278],{"align":232},[23,279,280,281,283],{},"Single AI Model (",[51,282,66],{}," mode)",[246,285,286],{"align":232},"No time or money wasted passing messages between multiple agents.",[227,288,289,294,302],{},[246,290,291],{"align":232},[23,292,293],{},"Summarizing 15 book chapters or simulating 100 buyer cards",[246,295,296],{"align":232},[23,297,298,299,301],{},"Parallel ",[51,300,127],{}," Subagent Waves",[246,303,304],{"align":232},"Cuts waiting time from 45 minutes to 4 minutes at low token cost.",[227,306,307,312,317],{},[246,308,309],{"align":232},[23,310,311],{},"Multi-file feature or security-sensitive code refactor",[246,313,314],{"align":232},[23,315,316],{},"Planner -> Builder -> Read-Only Auditor",[246,318,319],{"align":232},"Locks file boundaries and verifies the final diff with unbiased fresh eyes.",{"title":187,"searchDepth":321,"depth":321,"links":322},2,[323,324],{"id":16,"depth":321,"text":17},{"id":158,"depth":321,"text":159},"Engineering Craft","October 19, 2026",null,"md","\u002Fimages\u002Fnotes\u002Fmulti-agent-ai-useful-pattern-or-extra-complexity.webp","07",{"sections":332},[333,339],{"marker":334,"heading":335,"body":336},"\u002F A","When Multi-Agent Orchestration Actually Earns Its Overhead",[337,338],"If you have ever built a multi-agent system that spent four minutes passing polite JSON messages back and forth only to produce a worse answer than a single prompt, you have met orchestration tax.","Across my own tools built in 2026, spawning subagents only pays off in three cases: isolating context windows so reviewers stay unbiased, embarrassing parallelism across independent slices, and enforcing adversarial analytical lenses.",{"marker":340,"heading":341,"body":342},"\u002F B","When to Default to Solo Mode or a Deterministic Python Script",[343,344],"In Gemini Advisor, the default route for localized changes is solo mode because coordinating a planner, builder, and reviewer on a 20-line CSS fix wastes tokens and latency.","Even more importantly, in Founder Suite and Book-to-Skill, agents are strictly forbidden from guessing token counts, SHA-256 hashes, or LTV\u002FCAC unit economics—those always run in deterministic Python CLI scripts.",true,"\u002Fnotes\u002Fmulti-agent-ai-useful-pattern-or-extra-complexity","Never spawn three conversational agents to debate a number that a 40-line Python script can calculate deterministically in two milliseconds.","7 min read",{"title":6,"description":187},"multi-agent-ai-useful-pattern-or-extra-complexity","notes\u002Fmulti-agent-ai-useful-pattern-or-extra-complexity",[353,354,355],"Never use AI agents to guess math, token counts, or financial margins—always use a deterministic Python script for exact numbers.","For small, focused tasks (1–2 files), a single model ('solo' mode) is faster, cheaper, and cleaner than a multi-agent crew.","Multi-agent workflows only earn their keep in three cases: independent code review, parallel waves of work, or structured disagreement.","10","mbvQx0BaVI2PbUWtr1SOwzElO00TWe4vORg-0lcn3Qo"]