[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"blog-why-9-agents-not-one-model":3},{"message":4,"data":5},"查询成功",{"id":6,"slug":7,"title":8,"excerpt":9,"coverImage":10,"status":11,"isPinned":12,"isFeatured":12,"publishedAt":13,"viewCount":14,"readingTime":15,"createdAt":16,"updatedAt":17,"category":18,"author":22,"content":28,"metaTitle":29,"metaDesc":30,"faqSchema":31,"tags":47,"relatedPosts":68},"9b5e6df0-6d3d-44a4-9186-e8c06fecca5c","why-9-agents-not-one-model","Why Custyle Built a 9-Agent Crew Instead of One Giant Model","Models keep getting bigger. Custyle still runs nine specialist agents instead of one. Here is the measured case for that call - and the honest case against it.","https:\u002F\u002Fpub-7cfffa54b5844016944f48b05a9f2282.r2.dev\u002Fblog\u002Fwhy-9-agents-not-one-model\u002Fcover.png","PUBLISHED",false,"2026-08-07T17:36:23.192Z",99,15,"2026-08-07T17:36:23.195Z","2026-09-21T22:39:32.691Z",{"id":19,"name":20,"slug":21},"17b15c13-7387-4a1d-a613-66cc6667c805","Learn","learn",{"id":23,"name":24,"avatarUrl":25,"authorBio":26,"authorTitle":27},"739d1411-ca20-4f09-ac9f-b3210195345d","Custyle Lab","https:\u002F\u002Fcustyle-s3.s3.us-east-1.amazonaws.com\u002Fcommon-dev\u002Fhomepage\u002Fimages\u002Fauthors\u002Fcustyle-lab.png","Custyle Lab explores how AI is changing the way people create, express, and wear what they care about. Deep dives into tools, trends, and techniques.","Research & Guides","# Why Custyle Built a 9-Agent Crew Instead of One Giant Model\n\n> **TL;DR:** Custyle runs nine specialist AI agents instead of one large model. The reason is measured, not stylistic: every frontier model tested by Chroma degrades as its context grows, merch needs real production data a model cannot invent, and a crew can be replayed step by step. A single model wins plenty of tasks — this article covers those too.\n\n## Table of Contents\n\n- [The Sentence That Hides Nine Jobs](#the-sentence-that-hides-nine-jobs)\n- [So Why Not Just Use One Big Model?](#so-why-not-just-use-one-big-model)\n- [Reason 1: Nine Jobs, Nine Different Skills](#reason-1-nine-jobs-nine-different-skills)\n- [How One Sentence Moves Through the Crew](#how-one-sentence-moves-through-the-crew)\n- [Reason 2: Big Contexts Get Worse, Not Better](#reason-2-big-contexts-get-worse-not-better)\n- [Reason 3: Merch Has Constraints You Can't Imagine Away](#reason-3-merch-has-constraints-you-cant-imagine-away)\n- [Reason 4: You Can Replay a Crew](#reason-4-you-can-replay-a-crew)\n- [One Model vs a Crew, Side by Side](#one-model-vs-a-crew-side-by-side)\n- [The Honest Part: When One Model Wins](#the-honest-part-when-one-model-wins)\n- [How the Crew Is Built Against Those Failures](#how-the-crew-is-built-against-those-failures)\n- [What This Means for You](#what-this-means-for-you)\n- [FAQ](#faq)\n\n## The Sentence That Hides Nine Jobs\n\nSay this to Custyle.ai, the AI Merch Agent platform: *\"Make me three retro cat shirts for summer, and add the good one to my cart.\"*\n\nIt sounds like one request. It isn't.\n\nRead it again slowly. Someone has to work out what \"retro\" means to you. Someone has to turn a loose idea into a real creative direction. Someone has to draw the cats. Someone has to know that a summer tee prints differently from a heavyweight hoodie. Someone has to place the art so it sits right on the chest. Someone has to pick the actual product. Someone has to show you how it looks on a person. And at the end, someone has to remember which one \"the good one\" was.\n\nThat is nine jobs hiding inside one sentence. Custyle handles them with nine agents — a group the Custyle Lab calls the Crew.\n\nThe framing the Lab uses internally is blunt: **Not one AI. A whole crew built to get your merch right.**\n\n## So Why Not Just Use One Big Model?\n\nIt is the fair question, and it gets fairer every year. Frontier models keep growing. Context windows now reach two million tokens. If one model can hold a novel, why can't it hold a T-shirt order?\n\nThe short answer: it can hold it. It just gets worse at it as the pile grows.\n\nThe longer answer is the rest of this article. Four reasons drove the Lab's call, and one honest counterweight that deserves its own section. None of them are about the Crew being cute. They are about what breaks.\n\n## Reason 1: Nine Jobs, Nine Different Skills\n\nReading someone's taste is not the same skill as drawing. Drawing is not the same skill as knowing how ink behaves on cotton. And none of that is the same skill as photography.\n\nA generalist model does all four passably. A specialist does one of them well. The gap is measurable: a research agent given curated tools and a focused prompt outperforms a generalist by **23–31%** on quality benchmarks. Microsoft has shown a coordinated network of **100+ smaller specialized models** beating a single large frontier model on industry benchmarks.\n\nIt helps to be concrete about how different these skills really are.\n\nReading taste is an inference problem. You said \"retro,\" and that word means nine different decades depending on who you are. The job is to narrow it using your references and the things you did not say.\n\nDrawing is a generation problem. Once the direction is set, the work is craft — line weight, colour relationships, whether the cat reads as charming or cursed at small sizes.\n\nProduction is a physics problem. Ink sits differently on cotton than on a blend. A design with fine hairlines survives on paper and disappears on fabric. Nothing about this is a matter of opinion.\n\nLayout is a geometry problem. Chest print areas are fixed. Hierarchy has rules. A composition that works in a square crop can break the moment it becomes a garment.\n\nPhotography is a lighting problem. Showing how something lands on a real person needs different instincts from drawing it in the first place.\n\nAsk one model to hold all five mindsets at once and it will average them. Averaging is exactly what you do not want from a design tool.\n\nSo the Lab split the work nine ways.\n\n| # | Crew Member | Role | What they own |\n|---|---|---|---|\n| 1 | **Vibbi** | Design Lead | Plans the turn, sends work out, keeps track. Dispatches — never executes. |\n| 2 | **Pia** | Preference Reader | Your taste, your references, the leanings you never said out loud |\n| 3 | **Nova** | Concept Shaper | Loose prompt into a sharper creative direction |\n| 4 | **Ink** | Artwork Maker | The visual language and detail that make it feel like a real piece |\n| 5 | **Bolt** | Production Brain | Right process, right material, right technique |\n| 6 | **Grid** | Layout Specialist | Composition, hierarchy, spacing, placement |\n| 7 | **Axis** | Product Architect | Finds the right product form for the idea |\n| 8 | **Moxy** | Try-On Director | How it lands on a real person |\n| 9 | **Lumi** | Scene Stylist | Mood, context, and atmosphere around the finished thing |\n\nThe order they work in is fixed:\n\n```\nUser Input\n  → Vibbi (orchestration)\n    → Pia (preference reading)\n      → Nova (concept shaping)\n        → Ink (artwork creation)\n          → Bolt (production decision)\n            → Grid (layout)\n              → Axis (product form selection)\n                → Moxy (try-on preview)\n                  → Lumi (scene styling)\n                    → Final Output\n```\n\nCustyle is not alone here. The 2026 agentic-design pattern across studios looks almost identical. An orchestrator handles creative direction. Generation agents make the assets. Review agents check structure. A refinement agent finishes the job. More than **72% of studios** now run multi-agent pipelines for consistent visual output.\n\n[→ Related: Meet the Custyle Crew](\u002Fblog\u002Fmeet-the-custyle-crew)\n\n## How One Sentence Moves Through the Crew\n\nSplitting work nine ways only helps if something holds the pieces together. That something is Vibbi, and the job runs as a fixed loop on every turn you take.\n\n```\nPolicy check (safety, limits)\n  → Load memory (design state, shop results, research, cart, conversation)\n    → Analyze intent (which domains, what actions, what references)\n      → Resolve references (\"the second one\" → specific artifact)\n        → Build execution plan (single step or multi-step)\n          → Validate risk (read \u002F credit \u002F cart change \u002F payment)\n            → Execute steps (send work to specialists)\n              → Compose response (one stream back to you)\n```\n\nNotice what happens before any art gets made. The system checks limits. It loads what it already knows about you. It works out what you actually asked for, then pins down what your words point at. Only then does it plan.\n\nThat loop runs across five areas of capability: **Create, Shop, Inspire, Transact, and Converse.** The retro cat shirt sentence touches four of them. It asks for creation, it asks for a product decision, it changes your cart, and it holds a conversation while doing all three.\n\nThis is the part that breaks flat prompting. \"Add the good one to my cart\" is not a design instruction. It is a cart change resting on a judgement made two steps earlier. One model doing this end to end must keep that thread alive mid-context. As the next section shows, mid-context is exactly where models get unreliable.\n\nVibbi handles it differently. Every result becomes a named thing with a position number. \"The second one\" is not re-derived from memory each time it comes up. It points at an object. And when the wording genuinely could mean two things, Vibbi asks rather than picking.\n\n## Reason 2: Big Contexts Get Worse, Not Better\n\nThis is the reason that settles the argument, and it comes from measurement rather than taste.\n\nChroma ran a study on **18 frontier models** — Claude Opus 4, GPT-4.1, o3, Gemini 2.5 Pro, Qwen3-235B, and more. Every single one degraded as input length grew, at every increment tested. Their conclusion is worth quoting exactly:\n\n> \"Models do not use their context uniformly; instead, their performance grows increasingly unreliable as input length grows.\"\n\nThree findings matter for merch:\n\n- **Things get lost in the middle.** Accuracy is highest when the important detail sits at the start or the end. Bury it in the middle and accuracy drops **by more than 30%**.\n- **Focused beats big.** On the LongMemEval benchmark, tight **~300-token** prompts beat full **~113,000-token** prompts across every model family tested.\n- **The marketed window is not the working window.** Practical safe budgets for two-million-token models land around **150,000–400,000 tokens** for high-accuracy work.\n\nDegradation also gets worse on tasks that need several reasoning hops. A merch turn is exactly that kind of task.\n\nNow picture one model holding all of it at once. Your taste notes. Three concept directions. The artwork. The print constraints. The layout maths. The product catalogue. The try-on render. The scene styling. That is the precise condition where measured performance falls apart. **Nine focused contexts beat one crowded one — not as a preference, as a result.**\n\n## Reason 3: Merch Has Constraints You Can't Imagine Away\n\nA model can imagine a hoodie. It cannot imagine what the hoodie costs today, whether it is in stock, or how wide the printable area actually is.\n\nThis is where the split earns its keep. Bolt exists because materials behave in specific ways. Axis exists because a design that sings on a tote can die on a cap. Neither of them guesses — they read real data.\n\nThe Lab holds four rules that shape the whole architecture:\n\n1. Every result is a **named artifact with a position number**. \"The second one\" resolves to a specific thing. If it is ambiguous, Vibbi asks instead of assuming.\n2. **Transactions never hallucinate.** Prices, stock, and order status come from real systems, never from a model's best guess.\n3. **High-risk actions need your explicit go-ahead.** Checkout, payment, refund.\n4. **Every run is replayable** — what ran, what went in, which models, what it cost.\n\nHere is the thing worth sitting with: you cannot guarantee any of these inside one forward pass. You cannot put a confirmation gate halfway through a single generation. You cannot swap the pricing step for a database lookup when there is no step to swap. **The architecture is a consequence of the guarantees, not a style choice.**\n\n[→ Related: How Does AI Merch Actually Work?](\u002Fblog\u002Fhow-does-ai-merch-actually-work)\n\n## Reason 4: You Can Replay a Crew\n\nWhen a design comes out wrong, the useful question is *which part went wrong*.\n\nWith nine agents, that question has an answer. You can see that Pia read the brief as minimalist when you meant maximalist, and that everything downstream followed a bad read. You fix one step and rerun.\n\nWith one model, there is one output and one shrug. The reasoning is not separable, so the fix is \"try the prompt again.\"\n\nAnthropic's own research system leans on the same idea. An orchestrator plans. **3 to 5 specialist subagents** run in parallel. A separate pass handles citations. They report a **90.2% improvement over single-agent** on internal evaluations, and up to **90% less time** on complex work.\n\nStructure also contains damage. Centralized systems with an orchestrator that breaks work down and reassembles it hold error amplification to **4.4x**. Independent agents with no coordinator hit **17.2x**. The captain is not overhead. The captain is the thing stopping one bad read from becoming nine bad reads.\n\n## One Model vs a Crew, Side by Side\n\nStripped of the argument, the trade reads like this.\n\n| Dimension | One giant model | The 9-agent Crew |\n|---|---|---|\n| **Speed per task** | Faster — roughly 2–4 seconds | Slower — roughly 8–15 seconds |\n| **Simple sequential work** | Usually wins; coordination adds nothing | Coordination can cost up to 70% on strictly linear tasks |\n| **Context handling** | One growing context; degrades as it fills | Nine focused contexts, each kept small |\n| **Specialist quality** | Competent across the board | 23–31% better on the specialist's own task |\n| **Real-world data** | Must be handed everything up front | Each specialist reads live production data |\n| **Error containment** | One output, one failure surface | Orchestrated: 4.4x amplification vs 17.2x uncoordinated |\n| **Debugging** | Reroll the prompt and hope | Replay the run, find the step, fix that step |\n| **Build cost** | Low — one prompt, one call | High — nine roles, handoffs, verification |\n\nRead the right-hand column honestly and it is not a clean win. It is a trade: more machinery and more seconds, bought in exchange for correctness you can check.\n\nFor a chatbot answering a question, that trade is bad. For something that spends your money and ships a physical object to your door, it is the right way round.\n\n## The Honest Part: When One Model Wins\n\nA page that only argued one side would not be worth citing. So here is the case against the Crew, stated properly.\n\n**Cognition published a post called \"Don't Build Multi-Agents.\"** Their argument is sharp. Actions carry implicit decisions, and when agents work in parallel on partial views, those decisions conflict. Their example: one subagent builds a Super Mario Bros. background while another builds the bird, and the final agent has to reconcile two incompatible ideas. Their diagnosis — *\"the decision-making ends up being too dispersed and context isn't able to be shared thoroughly enough.\"*\n\n**Multi-agent systems fail in patterned ways.** Cemri et al. analysed **1,600+ annotated traces** across **7 frameworks** and catalogued **14 distinct failure modes**, with six expert annotators reaching a Cohen's Kappa of **0.88**.\n\n| Failure category | Share | What it looks like |\n|---|---|---|\n| Specification & system design | **41.8%** | Fuzzy roles, bad task breakdown, duplicate agents, no stop condition |\n| Inter-agent misalignment | **36.9%** | Context lost on handoff, conflicting outputs, format mismatches |\n| Verification & termination | **21.3%** | Stopping early, checking badly, not checking at all |\n\n**And single agents often just win.** Princeton's NLP group found single-agent systems matched or beat multi-agent on **64% of tasks** when given the same tools and context. On strictly sequential tasks, naive coordination degraded results by up to **70%**. Multi-agent also costs time: roughly **8–15 seconds** per task versus **2–4 seconds** single-agent.\n\nThe fairest summary of the 2026 evidence is this: **the burden of proof sits on multi-agent, not on the simpler design.** Coordination has to earn its place.\n\n## How the Crew Is Built Against Those Failures\n\nCustyle pays that coordination tax knowingly. Merch creation happens to be the shape of problem where it pays back. The skills are genuinely different. Several steps run in parallel. And the constraints are real, so they have to come from systems rather than imagination.\n\nBut paying the tax only works if you build against the known failure modes. Look at that table again — roughly **79%** of multi-agent failures are role-definition and handoff problems. Not model problems. So:\n\n- **Fuzzy roles cause 41.8% of failures.** The Crew has nine roles with hard edges. Vibbi dispatches and never executes. Ink draws and never picks the product. No overlap, no duplicate agents.\n- **Handoff context loss causes 36.9%.** Named artifacts with position numbers close that gap. \"The second one\" is never re-derived from vibes — it points at a specific object.\n- **Weak verification causes 21.3%.** Replayable runs mean every step is inspectable after the fact.\n- **Cognition's strongest point still stands**, and the Lab took it: extra agents are fine when they read and analyse, but state changes should stay single-threaded. In the Crew, specialists generate — **Vibbi alone owns the plan**, and anything touching your cart or payment waits for you to say yes.\n\nThat last one is the direct answer to the Flappy Bird problem. Nine agents propose. One agent decides. You confirm the parts that cost money.\n\n[→ Related: Inside Vibbi — How Custyle's Design Lead Orchestrates 8 Specialists](\u002Fblog\u002Finside-vibbi-design-lead)\n\n## What This Means for You\n\nNone of this should be visible while you make a T-shirt.\n\nWhat you should feel is that the thing understood your vibe instead of averaging it. That the artwork looks intentional rather than generated. That the product it picked makes sense for the design. That \"the second one\" meant the second one. And that when it shows up at your door, the quality is something you can feel.\n\nThe Crew is how that gets delivered. Nine specialists, one captain, and a set of rules about what is allowed to happen without asking you first.\n\nNot one AI. A whole crew built to get your merch right.\n\n[→ Related: What Is an AI Merch Agent?](\u002Fblog\u002Fwhat-is-an-ai-merch-agent)\n\n## FAQ\n\n### How many AI agents does Custyle use?\n\nNine. One orchestrator called Vibbi, plus eight specialists: Pia, Nova, Ink, Bolt, Grid, Axis, Moxy, and Lumi. Each owns one part of the journey from your first sentence to the finished product, and they always run in the same fixed order.\n\n### Is multi-agent AI better than one big model?\n\nNot always, and the honest answer depends on the task. Princeton found single agents matched or beat multi-agent on 64% of tasks with equal tools. Multi-agent wins on work that splits into genuinely different skills with parallel steps — which is what merch creation looks like.\n\n### Why not just use ChatGPT or Claude directly for merch?\n\nA general model can draw a design. It cannot check live stock, read a real print area, or promise the price it quoted you. Custyle's Bolt and Axis pull from real production data, and the rule is strict: transactions never hallucinate.\n\n### Does a 9-agent system make it slower?\n\nSomewhat. Multi-agent work typically runs 8–15 seconds against 2–4 seconds for a single agent. Custyle accepts that cost for two reasons. Parallel steps recover much of it. And a wrong hoodie costs far more than a few extra seconds.\n\n### What happens if one agent gets it wrong?\n\nYou can see which one. Every run is replayable — what ran, what went in, which models, what it cost. That means a bad read gets fixed at its source rather than by rerolling the whole thing. An orchestrator also holds error amplification to 4.4x versus 17.2x for uncoordinated agents.","Why 9 AI Agents Beat One Giant Model","Models keep getting bigger. Custyle still runs nine specialist agents. Here is the evidence behind that call - including the case against it.",[32,35,38,41,44],{"answer":33,"question":34},"Nine. One orchestrator called Vibbi, plus eight specialists: Pia, Nova, Ink, Bolt, Grid, Axis, Moxy, and Lumi. Each owns one part of the journey from your first sentence to the finished product, and they always run in the same fixed order.","How many AI agents does Custyle use?",{"answer":36,"question":37},"Not always, and the honest answer depends on the task. Princeton found single agents matched or beat multi-agent on 64% of tasks with equal tools. Multi-agent wins on work that splits into genuinely different skills with parallel steps — which is what merch creation looks like.","Is multi-agent AI better than one big model?",{"answer":39,"question":40},"A general model can draw a design. It cannot check live stock, read a real print area, or promise the price it quoted you. Custyle's Bolt and Axis pull from real production data, and the rule is strict: transactions never hallucinate.","Why not just use ChatGPT or Claude directly for merch?",{"answer":42,"question":43},"Somewhat. Multi-agent work typically runs 8-15 seconds against 2-4 seconds for a single agent. Custyle accepts that cost because parallel steps recover much of it, and because a wrong hoodie is far more expensive than a few extra seconds of waiting.","Does a 9-agent system make it slower?",{"answer":45,"question":46},"You can see which one. Every run is replayable — what ran, what went in, which models, what it cost. That means a bad read gets fixed at its source rather than by rerolling the whole thing. An orchestrator also holds error amplification to 4.4x versus 17.2x for uncoordinated agents.","What happens if one agent gets it wrong?",[48,53,57,61,65],{"id":49,"name":50,"slug":51,"color":52},"7ffa5389-f3c5-4c8f-8033-a52190149d24","Custyle Crew","custyle-crew",null,{"id":54,"name":55,"slug":56,"color":52},"3400900b-d276-4d58-a0cb-171eb837ddbc","Multi-Agent AI","multi-agent-ai",{"id":58,"name":59,"slug":60,"color":52},"e3937241-eef5-48e4-9b19-bc7f76b309e7","AI agent architecture","ai-agent-architecture",{"id":62,"name":63,"slug":64,"color":52},"f59b1fa7-9ea9-4885-8704-a31bcd080b20","ai merch agent","ai-merch-agent",{"id":66,"name":24,"slug":67,"color":52},"51de8512-7a52-4b4d-af12-30f78487ea26","custyle-lab",[69,101,137,161],{"id":70,"slug":71,"title":72,"excerpt":73,"coverImage":74,"status":11,"isPinned":12,"isFeatured":12,"publishedAt":75,"viewCount":76,"readingTime":77,"createdAt":78,"updatedAt":79,"category":80,"author":81,"tags":82},"48795332-c084-4ab1-a913-50004014efca","inside-moxy-lumi-presentation","Inside Moxy & Lumi: How Custyle Stages Your Product Before You See It","Moxy and Lumi are the presentation layer of the Custyle Crew — Moxy shows how your merch lands on a real person; Lumi builds the mood, context, and atmosphere around what you made.","https:\u002F\u002Fpub-7cfffa54b5844016944f48b05a9f2282.r2.dev\u002Fblog\u002Finside-moxy-lumi-presentation\u002Fcover.png","2026-09-08T10:46:03.315Z",39,10,"2026-09-08T10:46:03.316Z","2026-09-21T23:22:53.930Z",{"id":19,"name":20,"slug":21},{"id":23,"name":24,"avatarUrl":25,"authorBio":26,"authorTitle":27},[83,84,89,93,97],{"id":49,"name":50,"slug":51,"color":52},{"id":85,"name":86,"slug":87,"color":88},"d24c3c71-dd9e-4444-85a0-7ba902c2436e","AI Merch","ai-merch","#7B61FF",{"id":90,"name":91,"slug":92,"color":52},"2a8d3670-e3e4-4335-a5f0-fb01ec9a1590","Moxy","moxy",{"id":94,"name":95,"slug":96,"color":52},"37dcb5fc-c466-41c4-8a67-434818093b4e","Lumi","lumi",{"id":98,"name":99,"slug":100,"color":52},"1768010a-3492-4022-80a3-6d49d9c0bbed","Agent System","agent-system",{"id":102,"slug":103,"title":104,"excerpt":105,"coverImage":106,"status":11,"isPinned":12,"isFeatured":12,"publishedAt":107,"viewCount":108,"readingTime":109,"createdAt":110,"updatedAt":111,"category":112,"author":113,"tags":119},"0e48cade-3dc8-40b1-9a63-74190bc5b162","creator-case-study-template","Creator Merch Case Study: How to Build a Merch Line in One Afternoon","A creator merch case study you can copy: real benchmarks (0.5–2% conversion, $68 AOV), an hour-by-hour afternoon build, and a fill-in template for your first drop.","https:\u002F\u002Fpub-7cfffa54b5844016944f48b05a9f2282.r2.dev\u002Fblog\u002Fcreator-case-study-template\u002Fcover.png","2026-09-04T09:29:51.231Z",77,11,"2026-09-04T09:29:51.233Z","2026-09-21T19:50:33.345Z",{"id":19,"name":20,"slug":21},{"id":114,"name":115,"avatarUrl":116,"authorBio":117,"authorTitle":118},"8958e79f-73d4-4b71-92de-1ae8c41e61da","Custyle Picks","https:\u002F\u002Fcustyle-s3.s3.us-east-1.amazonaws.com\u002Fcommon-dev\u002Fhomepage\u002Fimages\u002Fauthors\u002Fcustyle-picks.png","Custyle Picks breaks down the tools, platforms, and options so you can make better decisions about merch, design, and AI creation.","Reviews & Comparisons",[120,124,128,132,136],{"id":121,"name":122,"slug":123,"color":52},"28422699-60a5-4d8c-b2b6-55c4922505cb","Creator Merch","creator-merch",{"id":125,"name":126,"slug":127,"color":52},"7ac60104-2ec4-485b-9b70-11c126f1dd18","merch case study","merch-case-study",{"id":129,"name":130,"slug":131,"color":52},"6960efe7-5942-4b16-8f2c-098e3cfa1b98","creator economy","creator-economy",{"id":133,"name":134,"slug":135,"color":52},"22283c32-0d0f-4151-ac86-5a8b291795f2","merch drop","merch-drop",{"id":85,"name":86,"slug":87,"color":88},{"id":138,"slug":139,"title":140,"excerpt":141,"coverImage":142,"status":11,"isPinned":12,"isFeatured":12,"publishedAt":143,"viewCount":144,"readingTime":109,"createdAt":145,"updatedAt":146,"category":147,"author":148,"tags":149},"34b93e69-7224-472b-8cda-15b53a390137","inside-axis-grid-structure","Inside Axis & Grid: Custyle's Form-and-Composition Crew","Axis and Grid are the structure layer of the Custyle Crew — Grid gets composition, spacing, and placement right; Axis finds the right product form your idea should live on.","https:\u002F\u002Fpub-7cfffa54b5844016944f48b05a9f2282.r2.dev\u002Fblog\u002Finside-axis-grid-structure\u002Fcover.png","2026-09-02T12:47:28.635Z",44,"2026-09-02T12:47:28.636Z","2026-09-21T23:22:54.341Z",{"id":19,"name":20,"slug":21},{"id":23,"name":24,"avatarUrl":25,"authorBio":26,"authorTitle":27},[150,151,152,156,160],{"id":49,"name":50,"slug":51,"color":52},{"id":85,"name":86,"slug":87,"color":88},{"id":153,"name":154,"slug":155,"color":52},"ff56d8f6-f419-47d7-a5e4-c4eeef87cf90","Axis","axis",{"id":157,"name":158,"slug":159,"color":52},"65f85316-b0be-40e0-a3ce-ccd860fa8150","Grid","grid",{"id":98,"name":99,"slug":100,"color":52},{"id":162,"slug":163,"title":164,"excerpt":165,"coverImage":166,"status":11,"isPinned":12,"isFeatured":12,"publishedAt":167,"viewCount":168,"readingTime":77,"createdAt":169,"updatedAt":170,"category":171,"author":172,"tags":173},"4dc0a37c-722a-4668-997d-e93a976db233","inside-bolt-production-brain","Inside Bolt: The Production Brain That Picks Materials Without Asking","Bolt is the AI production brain of the Custyle Crew — the agent that picks the process, material, and build quality for every design. Here is how.","https:\u002F\u002Fpub-7cfffa54b5844016944f48b05a9f2282.r2.dev\u002Fblog\u002Finside-bolt-production-brain\u002Fcover.png","2026-08-28T11:05:58.734Z",50,"2026-08-28T11:05:58.735Z","2026-09-21T23:22:53.624Z",{"id":19,"name":20,"slug":21},{"id":23,"name":24,"avatarUrl":25,"authorBio":26,"authorTitle":27},[174,175,176,180,184],{"id":49,"name":50,"slug":51,"color":52},{"id":85,"name":86,"slug":87,"color":88},{"id":177,"name":178,"slug":179,"color":52},"77dcaab7-7b0c-4897-8b18-c1741e35c30b","Production AI","production-ai",{"id":181,"name":182,"slug":183,"color":52},"f4cd4a76-2830-45de-a53d-d5165be3a4ba","Bolt","bolt",{"id":98,"name":99,"slug":100,"color":52}]