[{"data":1,"prerenderedAt":800},["ShallowReactive",2],{"insight-insights_en-getting-cited-by-chatgpt-and-claude":3,"insight-related-insights_en-getting-cited-by-chatgpt-and-claude":169},{"id":4,"title":5,"author":6,"blobHue":10,"body":11,"category":144,"date":145,"description":17,"draft":146,"extension":147,"featured":146,"headline":148,"hue":149,"letter":150,"meta":151,"navigation":152,"path":153,"readMinutes":154,"related":155,"seo":159,"stem":160,"summary":161,"toc":162,"__hash__":168},"insights_en\u002Finsights\u002Fgetting-cited-by-chatgpt-and-claude.md","Getting cited by ChatGPT and Claude",{"name":7,"role":8,"bio":9},"André","Founder & CTO","André is the founder and CTO of WizardingCode. Eight years building the software companies run on, now putting agents into production.",null,{"type":12,"value":13,"toc":135},"minimark",[14,18,21,26,29,32,35,39,80,86,90,93,100,103,109,112,116,119,122,125,129,132],[15,16,17],"p",{},"More and more product research starts with a question typed into ChatGPT, Claude or another assistant, not a search box. “Which linen shirt holds up after washing?” “Is this blender loud?” The assistant answers in a paragraph, and sometimes it names a source.",[15,19,20],{},"Generative-engine optimisation is the work of making your pages the source it names. It is not a new trick. Most of it is writing product pages the way a careful shop assistant would talk: plainly, specifically, and without hiding the facts.",[22,23,25],"h2",{"id":24},"how-an-answer-engine-reads-your-page","How an answer engine reads your page",[15,27,28],{},"A search engine ranks pages. An answer engine assembles an answer from pieces of pages, and quotes the pieces it trusts. That changes what a good page looks like.",[15,30,31],{},"It prefers sentences that can be lifted out whole and still make sense. It prefers facts stated once, clearly, over facts implied by adjectives. It prefers sources that name things consistently, so it can be sure the product on your page is the same one a review talks about. And it reads structure: headings, lists, tables and structured data all help it find the piece it needs.",[15,33,34],{},"None of this replaces classic SEO. A page that is not indexed, fast and linked will not be read at all. Think of it as a floor and a ceiling: search engines are the floor, answer engines are the ceiling.",[22,36,38],{"id":37},"six-habits-of-pages-that-get-quoted","Six habits of pages that get quoted",[40,41,42,50,56,62,68,74],"ol",{},[43,44,45,49],"li",{},[46,47,48],"strong",{},"State facts plainly."," Material, dimensions, weight, care, compatibility, warranty, origin. In sentences, with units, not only in a spec table image.",[43,51,52,55],{},[46,53,54],{},"Write answer-shaped paragraphs."," Open a section with the direct answer in one sentence, then explain. The first sentence is the one most likely to be quoted.",[43,57,58,61],{},[46,59,60],{},"Add real questions and answers."," Use the questions customers actually ask, taken from support tickets and reviews, each with a short, complete answer.",[43,63,64,67],{},[46,65,66],{},"Build entity pages."," A page for the brand and a page for each category, saying what they are, who they are for and how they differ. Answer engines use them to understand where a product fits.",[43,69,70,73],{},[46,71,72],{},"Use structured data."," Product, offer, review and FAQ markup that matches the visible text exactly. Markup that contradicts the page does more harm than none.",[43,75,76,79],{},[46,77,78],{},"Keep names consistent."," The same product name, brand name and category name on the page, in the feed, in the markup and in every language.",[81,82,83],"blockquote",{},[15,84,85],{},"Write the sentence you would want an assistant to quote about you.",[22,87,89],{"id":88},"a-product-paragraph-before-and-after","A product paragraph, before and after",[15,91,92],{},"This is the kind of paragraph most catalogues are full of:",[15,94,95,99],{},[96,97,98],"em",{},"Before:"," “Discover effortless style with our iconic shirt. Crafted from premium fabrics for all-day comfort, it’s the perfect addition to any wardrobe. Elevate your look this season.”",[15,101,102],{},"There is nothing an assistant can quote. No material, no fit, no care, nothing that answers a question.",[15,104,105,108],{},[96,106,107],{},"After:"," “This shirt is made of 100% European linen, with a relaxed fit and a straight hem designed to be worn untucked. It is machine washable at 30 °C and softens with each wash. It runs true to size; if you are between sizes, choose the smaller one for a closer fit.”",[15,110,111],{},"The second version is not less appealing. It is more useful, and every sentence answers a question someone has asked. (The shirt and its details are an illustration, not a real listing.)",[22,113,115],{"id":114},"doing-it-for-thousands-of-products","Doing it for thousands of products",[15,117,118],{},"Writing like this by hand works for ten products. It does not work for a catalogue that changes every week.",[15,120,121],{},"In the marketplace system we built, a product studio writes product, category, brand and page copy with SEO and AI-search content, in four languages. It works from the supplier data and the catalogue, not from imagination. Brand and category entity pages are written the same way, so the names match everywhere.",[15,123,124],{},"A person still reviews what goes live. The studio’s job is to make the careful version the default, not to publish without anyone looking.",[22,126,128],{"id":127},"what-you-cant-control","What you can’t control",[15,130,131],{},"Nobody decides what an assistant cites. Models change, sources shift and the same question can get different answers on different days. Anyone who promises you a position in ChatGPT is promising something they do not control.",[15,133,134],{},"What you can control is whether your pages are worth quoting. Check your key products the simple way: ask the assistants the questions your customers ask, and see whose sentences come back. If they are not yours, the fix is usually on the page.",{"title":136,"searchDepth":137,"depth":137,"links":138},"",2,[139,140,141,142,143],{"id":24,"depth":137,"text":25},{"id":37,"depth":137,"text":38},{"id":88,"depth":137,"text":89},{"id":114,"depth":137,"text":115},{"id":127,"depth":137,"text":128},"engineering","2026-07-28",false,"md","Getting cited by ChatGPT and ==Claude.==","blue","G",{},true,"\u002Finsights\u002Fgetting-cited-by-chatgpt-and-claude",4,[156,157,158],"cart-recovery-with-agents-notes-from-our-own-store","build-buy-or-both","every-agent-needs-a-judge",{"title":5,"description":17},"insights\u002Fgetting-cited-by-chatgpt-and-claude","Generative-engine optimisation for product pages: facts, Q&A pairs and entity pages.",[163,164,165,166,167],{"id":24,"label":25},{"id":37,"label":38},{"id":88,"label":89},{"id":114,"label":115},{"id":127,"label":128},"_uD17NODHcZyxNIpzQ8lLewrVkFutpXnh-B7MI0qwuQ",[170,361,613],{"id":171,"title":172,"author":173,"blobHue":10,"body":174,"category":340,"date":341,"description":178,"draft":146,"extension":147,"featured":146,"headline":342,"hue":343,"letter":344,"meta":345,"navigation":152,"path":346,"readMinutes":154,"related":347,"seo":351,"stem":352,"summary":353,"toc":354,"__hash__":360},"insights_en\u002Finsights\u002Fcart-recovery-with-agents-notes-from-our-own-store.md","Cart recovery with agents: notes from our own store",{"name":7,"role":8,"bio":9},{"type":12,"value":175,"toc":333},[176,179,182,186,189,192,196,199,202,222,225,229,232,235,239,242,245,320,324,327,330],[15,177,178],{},"Our founder also runs a fashion marketplace. We built its Agentic OS the way we build one for any client, inside its own CRM: 90 agents in eight departments, each team checked by a judge. Cart recovery is one of the places where the data was already there and the problem was obvious. People fill a cart, and then life happens.",[15,180,181],{},"These are notes on what changed, what we measure, and what did not work. We have left out revenue and percentages on purpose. They depend on the season, the catalogue and the traffic, and a number without that context would tell you less than the method does.",[22,183,185],{"id":184},"where-we-started","Where we started",[15,187,188],{},"Before agents, cart recovery was like most stores’: a fixed sequence. One email after a few hours, another a day later, sometimes a discount in the third. The same words for everyone, in the language of the store, whatever the customer had been looking at.",[15,190,191],{},"It worked. It just spoke to nobody in particular, and got the results of that.",[22,193,195],{"id":194},"one-person-one-message","One person, one message",[15,197,198],{},"The first real change was writing each message for the person who left the cart. Every night, the system refreshes a profile for every buyer: a persona, a churn risk and a next best action. The recovery agent reads that profile, the cart and the customer’s history before it writes anything.",[15,200,201],{},"That changes three things.",[40,203,204,210,216],{},[43,205,206,209],{},[46,207,208],{},"Timing."," A customer who usually buys late in the evening is not reminded at nine in the morning. Someone who comes back within the hour is left alone for a while.",[43,211,212,215],{},[46,213,214],{},"Channel."," Email, SMS, push or WhatsApp, depending on what that person has agreed to and has actually responded to before. Not every channel, and not the loudest one.",[43,217,218,221],{},[46,219,220],{},"Tone."," A first-time visitor and a customer with years of orders do not get the same message. The first needs reassurance about sizes and returns. The second needs a short reminder, in their language, about the pieces they chose.",[15,223,224],{},"A judge checks the messages: the facts about the product and the policy must match the catalogue and the shared memory, and the tone must fit the brand.",[22,226,228],{"id":227},"vouchers-only-inside-limits","Vouchers, only inside limits",[15,230,231],{},"The easiest way to recover a cart is to give money away. That is also the easiest way to lose margin on a sale that might have happened anyway.",[15,233,234],{},"So the recovery agent can propose a voucher, but only inside limits the commerce team has set, and only after a margin guardrail has checked it. The guardrail is plain code, not a model. It works out the floor for each product after returns, fees and shipping, and anything below it is vetoed. No agent ever writes a price. A voucher is a tool for a specific person, not a default.",[22,236,238],{"id":237},"a-proposal-every-week-a-verdict-28-days-later","A proposal every week, a verdict 28 days later",[15,240,241],{},"Once a week, the system proposes how to recover more: change the timing for a segment, try a different channel, adjust the wording, tighten or loosen a voucher rule. A person reviews each proposal and decides whether to apply it.",[15,243,244],{},"Then comes the part we care about most. Every applied change is measured 28 days later, against what was happening before. Not the next morning, when a change always looks good or bad by chance. Twenty-eight days is long enough to see whether customers actually came back and bought, and whether the orders held once returns had come in.",[246,247,248,264],"table",{},[249,250,251],"thead",{},[252,253,254,258,261],"tr",{},[255,256,257],"th",{},"STEP",[255,259,260],{},"WHO",[255,262,263],{},"WHEN",[265,266,267,281,294,307],"tbody",{},[252,268,269,275,278],{},[270,271,272],"td",{},[46,273,274],{},"Propose a change",[270,276,277],{},"The recovery agent",[270,279,280],{},"Every week",[252,282,283,288,291],{},[270,284,285],{},[46,286,287],{},"Approve or reject",[270,289,290],{},"A person on the commerce team",[270,292,293],{},"Before anything changes",[252,295,296,301,304],{},[270,297,298],{},[46,299,300],{},"Measure the result",[270,302,303],{},"The system, against the baseline",[270,305,306],{},"28 days after it goes live",[252,308,309,314,317],{},[270,310,311],{},[46,312,313],{},"Keep, change or drop",[270,315,316],{},"The same person",[270,318,319],{},"After the measurement",[22,321,323],{"id":322},"what-didnt-work","What didn’t work",[15,325,326],{},"Not every proposal earns its place. Some changes made no measurable difference after 28 days and were dropped. That is the point of measuring: without it, they would still be running, and we would still believe in them.",[15,328,329],{},"A profile is only as good as the history behind it. For a first-time visitor there is little to go on, and the message is closer to the generic one than we would like. We treat that as a limit to be honest about, not something to paper over with guesses.",[15,331,332],{},"And a better message does not fix a bad reason for leaving. When a cart is abandoned because of shipping costs, a sold-out size or a confusing returns policy, the right move is to fix the store. The recovery agent’s most useful output is sometimes not a message at all, but a pattern for someone to look at.",{"title":136,"searchDepth":137,"depth":137,"links":334},[335,336,337,338,339],{"id":184,"depth":137,"text":185},{"id":194,"depth":137,"text":195},{"id":227,"depth":137,"text":228},{"id":237,"depth":137,"text":238},{"id":322,"depth":137,"text":323},"case-notes","2026-08-11","Cart recovery with agents: notes from our own ==store.==","sun","R",{},"\u002Finsights\u002Fcart-recovery-with-agents-notes-from-our-own-store",[348,349,350],"testing-a-campaign-on-customers-who-dont-exist","what-agents-should-never-do-alone","getting-cited-by-chatgpt-and-claude",{"title":172,"description":178},"insights\u002Fcart-recovery-with-agents-notes-from-our-own-store","Timing, channel and tone. What changed when the message was written for one person.",[355,356,357,358,359],{"id":184,"label":185},{"id":194,"label":195},{"id":227,"label":228},{"id":237,"label":238},{"id":322,"label":323},"KH3Tvoi1FzHdhgydQOS66KG25jDv5Q_dCBEpLT85Ztc",{"id":362,"title":363,"author":364,"blobHue":10,"body":365,"category":592,"date":593,"description":369,"draft":146,"extension":147,"featured":146,"headline":594,"hue":595,"letter":596,"meta":597,"navigation":152,"path":598,"readMinutes":154,"related":599,"seo":603,"stem":604,"summary":605,"toc":606,"__hash__":612},"insights_en\u002Finsights\u002Fbuild-buy-or-both.md","Build, buy or both?",{"name":7,"role":8,"bio":9},{"type":12,"value":366,"toc":585},[367,370,373,377,380,387,394,400,403,407,410,436,439,443,446,452,458,538,541,546,550,553,556,560,582],[15,368,369],{},"Every company we talk to already pays for some AI. A writing assistant here, a meeting summariser there, a chatbot on the website. The question is rarely whether to use AI. It is which work deserves more than a subscription.",[15,371,372],{},"The honest answer is that most of it doesn’t. And the part that does usually needs both: things you buy and things you build.",[22,374,376],{"id":375},"when-buying-is-enough","When buying is enough",[15,378,379],{},"An off-the-shelf tool is the right choice when three things are true.",[15,381,382,383,386],{},"The work is ",[46,384,385],{},"generic",". Summarising a meeting, rewriting an email, translating a document, drafting a first version of a job ad. Every company does it roughly the same way, so a product built for everyone fits you well.",[15,388,389,390,393],{},"It needs ",[46,391,392],{},"no live data",". The input is whatever the person pastes in, and the output goes back to that person. Nothing has to be read from your ERP or written into your CRM.",[15,395,396,399],{},[46,397,398],{},"Nobody owns an outcome."," It makes individuals faster, but no team has signed up to move a number with it. If it gets worse next month, someone switches tools and nothing breaks.",[15,401,402],{},"For this kind of work, building your own is a waste of money. Buy the tool, set some sensible rules on what data goes into it, and move on.",[22,404,406],{"id":405},"when-a-process-deserves-its-own-agent","When a process deserves its own agent",[15,408,409],{},"The picture changes when the work runs through your systems and your rules. A process deserves its own agent when:",[40,411,412,418,424,430],{},[43,413,414,417],{},[46,415,416],{},"It lives in your systems."," The agent has to read the order, the ticket history, the supplier file or the price list, and write back to them.",[43,419,420,423],{},[46,421,422],{},"There is a number."," First response time, hours spent on admin, stock left unsold. A team owns it and wants it to move.",[43,425,426,429],{},[46,427,428],{},"Some decisions need a person."," Refunds, promises, anything that cannot be undone. Those need approval rules that match how your company works, not how a vendor imagined companies work.",[43,431,432,435],{},[46,433,434],{},"It is specific to you."," Your policies, your exceptions, your tone, the thing only Maria knows. A generic tool does not know any of it and you cannot teach it properly.",[15,437,438],{},"No product sold to everyone can do all four for you. It doesn’t know your systems, it can’t own your number, and its rules are someone else’s.",[22,440,442],{"id":441},"the-both-pattern","The both pattern",[15,444,445],{},"In practice, the answer for most serious processes is both.",[15,447,448,451],{},[46,449,450],{},"You buy the commodity."," The language models themselves, from the big model providers. Hosting, email delivery, search, the plumbing. Nobody should build their own model to answer supplier emails.",[15,453,454,457],{},[46,455,456],{},"You build what is yours."," The agents, the approval rules and the shared memory. That is where your process, your policies and your judgement live, and it is the part that makes the difference between a demo and a system you can rely on.",[246,459,460,472],{},[249,461,462],{},[252,463,464,466,469],{},[255,465],{},[255,467,468],{},"BUY",[255,470,471],{},"BUILD",[265,473,474,487,499,512,525],{},[252,475,476,481,484],{},[270,477,478],{},[46,479,480],{},"Models",[270,482,483],{},"From the model providers",[270,485,486],{},"Never",[252,488,489,494,497],{},[270,490,491],{},[46,492,493],{},"Infrastructure",[270,495,496],{},"Hosting, email, storage",[270,498,486],{},[252,500,501,506,509],{},[270,502,503],{},[46,504,505],{},"Agents",[270,507,508],{},"Rarely a fit",[270,510,511],{},"Built around your process",[252,513,514,519,522],{},[270,515,516],{},[46,517,518],{},"Approval rules",[270,520,521],{},"Generic settings",[270,523,524],{},"Written with your team",[252,526,527,532,535],{},[270,528,529],{},[46,530,531],{},"Shared memory",[270,533,534],{},"Empty until you fill it",[270,536,537],{},"Your policies, history and decisions",[15,539,540],{},"This split also protects you. Models improve and prices change every few months. When the agents, rules and memory are yours, swapping the model underneath is an engineering task, not a migration. The marketplace system we built uses models from several providers at once, each chosen for the job, and the system does not depend on any single one.",[81,542,543],{},[15,544,545],{},"Buy the model. Build the part that knows your business.",[22,547,549],{"id":548},"who-owns-what","Who owns what",[15,551,552],{},"Whatever you build, ask who owns it at the end. In our projects the answer is simple. The agents run in your environment and the memory is in your systems. You own the code, the prompts and the data, so you can run them without us.",[15,554,555],{},"That matters more than it seems. A process that runs on an agent you do not own is a process someone else can reprice, change or switch off.",[22,557,559],{"id":558},"a-quick-test","A quick test",[561,562,564],"prose-checklist",{"title":563},"Before you buy another AI tool, ask",[565,566,567,570,573,576,579],"ul",{},[43,568,569],{},"Does it need to read or write your live systems?",[43,571,572],{},"Is there a number that a team is expected to move with it?",[43,574,575],{},"Are there decisions in it that must stay with a person?",[43,577,578],{},"Does it depend on policies, exceptions or tone that are specific to you?",[43,580,581],{},"If the vendor changed its price or product tomorrow, would a process stop?",[15,583,584],{},"If every answer is no, buy it. If two or more are yes, the process probably deserves its own agent, built on bought models, with rules and memory that belong to you.",{"title":136,"searchDepth":137,"depth":137,"links":586},[587,588,589,590,591],{"id":375,"depth":137,"text":376},{"id":405,"depth":137,"text":406},{"id":441,"depth":137,"text":442},{"id":548,"depth":137,"text":549},{"id":558,"depth":137,"text":559},"strategy","2026-08-04","Build, buy or ==both?==","violet","?",{},"\u002Finsights\u002Fbuild-buy-or-both",[600,601,602],"what-an-agentic-os-is-and-what-it-isnt","why-95-percent-of-ai-pilots-never-reach-production","the-30-day-playbook-week-by-week",{"title":363,"description":369},"insights\u002Fbuild-buy-or-both","When an off-the-shelf AI tool is enough, and when your process deserves its own agent.",[607,608,609,610,611],{"id":375,"label":376},{"id":405,"label":406},{"id":441,"label":442},{"id":548,"label":549},{"id":558,"label":559},"XoWrr_D4QkTedkkrnrgccPyPPGBK8-ZFnFJfkxogvLc",{"id":614,"title":615,"author":616,"blobHue":10,"body":617,"category":144,"date":784,"description":621,"draft":146,"extension":147,"featured":146,"headline":785,"hue":149,"letter":786,"meta":787,"navigation":152,"path":788,"readMinutes":154,"related":789,"seo":790,"stem":791,"summary":792,"toc":793,"__hash__":799},"insights_en\u002Finsights\u002Fevery-agent-needs-a-judge.md","Every agent needs a judge",{"name":7,"role":8,"bio":9},{"type":12,"value":618,"toc":777},[619,622,625,629,632,635,638,642,645,713,716,719,723,726,746,749,753,756,759,762,767,771,774],[15,620,621],{},"A language model is very good at sounding right. That is the problem. An answer with a wrong refund amount reads exactly as confidently as one with the right amount, and a person skimming a queue of drafts will not catch it every time.",[15,623,624],{},"So we don’t ask people to catch it. In the marketplace system we built, 90 agents work across eight departments, and every team has a judge: 9 judges in all. A judge is a second model whose only job is to check another agent’s answer before anyone relies on it.",[22,626,628],{"id":627},"why-a-second-model","Why a second model",[15,630,631],{},"Asking the same model to review its own work helps less than you would think. It tends to agree with itself, and it shares the blind spots that produced the mistake in the first place.",[15,633,634],{},"That is why our judges run on a different model family from the agents they check. Different training, different habits, different failure modes. When two unrelated models agree that an answer is grounded and within policy, that means much more than one model agreeing with itself twice.",[15,636,637],{},"A judge also sees the answer differently. The agent was trying to be helpful. The judge is only trying to find what is wrong. Giving it a narrow brief, and nothing else to do, is what makes it useful.",[22,639,641],{"id":640},"block-or-grade-later","Block or grade later",[15,643,644],{},"There are two ways to put a judge in the path, and choosing between them is the main design decision.",[246,646,647,659],{},[249,648,649],{},[252,650,651,653,656],{},[255,652],{},[255,654,655],{},"BLOCK AND REPAIR",[255,657,658],{},"SHIP, THEN GRADE",[265,660,661,674,687,700],{},[252,662,663,668,671],{},[270,664,665],{},[46,666,667],{},"When the judge runs",[270,669,670],{},"Before the answer leaves",[270,672,673],{},"After it has gone out",[252,675,676,681,684],{},[270,677,678],{},[46,679,680],{},"If it fails",[270,682,683],{},"Sent back once with the reasons, then shipped with reservations",[270,685,686],{},"Flagged, counted, fed into the next lessons",[252,688,689,694,697],{},[270,690,691],{},[46,692,693],{},"Cost to the user",[270,695,696],{},"Some extra seconds",[270,698,699],{},"None",[252,701,702,707,710],{},[270,703,704],{},[46,705,706],{},"Use it for",[270,708,709],{},"Anything a customer sees, anything with money",[270,711,712],{},"High volume, low risk, easy to correct",[15,714,715],{},"In the blocking path, a failed answer is not simply rejected. It goes back to the agent once, with the judge’s reasons, and the agent gets a chance to repair it. If the repaired answer still fails, it goes out marked with the judge’s reservations, or, where the rules require it, to a person. Nothing leaves without a verdict.",[15,717,718],{},"In the grading path, the answer goes out and the judge scores it afterwards. The scores show where an agent drifts, and they feed the lessons the system learns overnight. Those lessons are not applied on their own: a person approves each one before agents use it.",[22,720,722],{"id":721},"what-a-judge-checks","What a judge checks",[15,724,725],{},"A judge with a vague brief (“is this a good answer?”) is expensive noise. Ours check specific things, and each check can fail on its own:",[561,727,729],{"title":728},"What a judge checks, every time",[565,730,731,734,737,740,743],{},[43,732,733],{},"Can every number in the answer be traced to data the agent actually read in this run?",[43,735,736],{},"Does it contradict a policy in the shared memory?",[43,738,739],{},"Does it answer the question that was asked, completely?",[43,741,742],{},"Does it propose an action the approval rules do not allow?",[43,744,745],{},"Is the tone right for who will read it?",[15,747,748],{},"The first check is the one that matters most. A refund amount, a delivery date or a stock figure that does not appear in any tool result is treated as invented, however plausible it looks.",[22,750,752],{"id":751},"what-it-costs-and-when-to-skip-it","What it costs, and when to skip it",[15,754,755],{},"A judge is another model call on every answer. It adds tokens, and in the blocking path it adds seconds. For a customer reply that is cheap insurance. For some work, it is waste.",[15,757,758],{},"We skip the judge, or move it to grading later, when the output is low stakes, reversible and cheap to check. An internal tag suggestion that a person sees anyway, or a draft that someone always edits before sending, does not need a second model standing in front of it.",[15,760,761],{},"We also never ask a model to check what code can check. Whether a price is above its floor, whether a date is in the future or whether an order exists are deterministic questions. They get deterministic answers, which are faster and do not have opinions.",[81,763,764],{},[15,765,766],{},"A judge is for judgement. If a rule can be written as code, write it as code.",[22,768,770],{"id":769},"fail-loudly-not-silently","Fail loudly, not silently",[15,772,773],{},"Judges fail too. They time out, or their provider is down for a few minutes. The worst thing a system can do then is pretend the check happened.",[15,775,776],{},"When a judge cannot run, the answer is shown with a visible “not validated” mark, and the person reading it decides whether to rely on it. A silent pass is how a system loses the trust of the people who work with it, and that trust is much harder to rebuild than a timeout is to fix.",{"title":136,"searchDepth":137,"depth":137,"links":778},[779,780,781,782,783],{"id":627,"depth":137,"text":628},{"id":640,"depth":137,"text":641},{"id":721,"depth":137,"text":722},{"id":751,"depth":137,"text":752},{"id":769,"depth":137,"text":770},"2026-09-01","Every agent needs a ==judge.==","J",{},"\u002Finsights\u002Fevery-agent-needs-a-judge",[349,600,348],{"title":615,"description":621},"insights\u002Fevery-agent-needs-a-judge","Why we put a second model in front of every answer, and when it isn’t worth the cost.",[794,795,796,797,798],{"id":627,"label":628},{"id":640,"label":641},{"id":721,"label":722},{"id":751,"label":752},{"id":769,"label":770},"XzL5diXOtcDslArv3P74XXRkQd_4qZoVSP7GUPDaC8E",1790618465062]