[{"data":1,"prerenderedAt":544},["ShallowReactive",2],{"post-\u002Fhealthcare-ai-chatbot-triage-medical-device":3},{"page":4,"translations":398,"nav":406,"related":530,"random":536},{"id":5,"title":6,"body":7,"categories":368,"category":371,"changeHistory":371,"date":372,"description":373,"disclosures":374,"draft":377,"extension":378,"firstLiveAt":371,"image":379,"imageAlt":380,"kind":381,"lang":382,"meta":383,"navigation":384,"omitGermanLocalizationDisclosure":377,"path":385,"publishedAt":371,"readingTime":386,"rights":371,"seo":387,"seoTitle":388,"slug":389,"sources":371,"stem":389,"tags":390,"translationKey":389,"type":396,"updated":371,"__hash__":397},"posts\u002Fhealthcare-ai-chatbot-triage-medical-device.md","AI in Healthcare: The Difference Between Chatbots, Triage, and Medical Devices",{"type":8,"value":9,"toc":351},"minimark",[10,49,52,55,58,61,64,67,72,75,90,93,96,106,109,112,116,119,124,127,130,133,137,140,143,146,150,153,156,159,162,166,169,179,182,186,189,192,201,204,214,218,221,224,227,230,240,243,247,250,253,263,271,274,278,281,284,287,310,313,317,320,323,326,329,332,339,342,345],[11,12,13,21],"blockquote",{},[14,15,16,17],"p",{},"💡 ",[18,19,20],"strong",{},"Summary: Key Takeaways",[22,23,24,31,37,43],"ul",{},[25,26,27,30],"li",{},[18,28,29],{},"Using an AI system in a healthcare organisation does not automatically make it high risk or a medical device."," Its intended purpose and actual effect on decisions are what matter.",[25,32,33,36],{},[18,34,35],{},"An appointment chatbot, an emergency triage system, and software that recommends a diagnosis are not in the same category."," Emergency healthcare triage is listed as a high-risk use under the EU AI Act. Software intended for diagnosis or treatment may also fall under medical-device rules.",[25,38,39,42],{},[18,40,41],{},"Healthcare obligations overlap."," The EU AI Act, medical-device law, the GDPR or Türkiye's KVKK, professional rules, and the health regulations of the country where the service is delivered may all apply together.",[25,44,45,48],{},[18,46,47],{},"The high-risk timetable needs to be read separately."," A 2026 amendment postponed the relevant high-risk provisions for Annex III systems until 2 December 2027. For systems classified as high risk through Annex I products such as medical devices, the date is 2 August 2028.",[14,50,51],{},"Imagine three chatbots on the website of a medical-tourism company.",[14,53,54],{},"The first provides the clinic's address, opening hours, and available appointment times. The second asks about symptoms and determines how urgently the person should be seen. The third interprets an uploaded test result and suggests a possible diagnosis or treatment option.",[14,56,57],{},"All three could use the same chat interface. They could even run on the same language model behind the scenes.",[14,59,60],{},"But they are not the same system.",[14,62,63],{},"The first chatbot simplifies administrative work. The second affects how quickly a patient reaches care. The third enters the clinical decision itself. As the consequence of an incorrect answer grows, the legal, product, and operating responsibilities change with it.",[14,65,66],{},"That is why “we use a chatbot” is far too crude a description for assessing AI in healthcare. The real question is where the conversation begins to change a healthcare decision.",[68,69,71],"h2",{"id":70},"being-in-healthcare-is-not-a-classification-by-itself","Being in healthcare is not a classification by itself",[14,73,74],{},"Software that predicts meal demand in a hospital and software that interprets a radiology image may operate in the same organisation. One supports operational planning. The other affects a medical decision.",[14,76,77,78,89],{},"The EU AI Act does not classify a system by the name of its industry alone. Under ",[79,80,88],"a",{"href":81,"className":82,"rel":84,"target":87},"https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Freg\u002F2024\u002F1689\u002F2026-07-27\u002Feng",[83],"dofollow",[85,86],"nofollow","noopener","_blank","the current consolidated text of the Act",", an AI system can become high risk through two main routes.",[14,91,92],{},"The first is when AI is itself a product, or a safety component of a product, already regulated in the EU. Medical devices are one of those product groups. For the system to qualify as high risk through this route, the relevant product law must also require a third-party conformity assessment.",[14,94,95],{},"The second route covers sensitive uses listed directly in Annex III. These include systems used to assess emergency calls from natural persons, prioritise the dispatch of first-response services, or triage patients in emergency healthcare.",[14,97,98,99,105],{},"Classification and the implementation timetable need to be kept separate. ",[79,100,104],{"href":101,"className":102,"rel":103,"target":87},"https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Freg\u002F2026\u002F1744\u002Foj\u002Feng",[83],[85,86],"An amendment adopted in 2026"," postponed the relevant high-risk provisions for Annex III systems until 2 December 2027. For systems treated as high risk through Annex I products such as medical devices, the date is 2 August 2028. A system may already meet the Act's definition of high risk even though the binding application of risk management, technical documentation, and related high-risk duties begins later. Medical-device, data-protection, and healthcare law do not wait for those dates.",[14,107,108],{},"It is therefore wrong to say that every AI system used in healthcare is high risk. It is equally wrong to assume that a system cannot be high risk because it is called a chatbot.",[14,110,111],{},"Classification follows the job you give the system, not the product label.",[68,113,115],{"id":114},"four-uses-four-different-outcomes","Four uses, four different outcomes",[14,117,118],{},"The distinction becomes clearer through everyday uses.",[120,121,123],"h3",{"id":122},"_1-administrative-information-and-appointments","1. Administrative information and appointments",[14,125,126],{},"The chatbot provides an address, opening hours, a price range, or appointment availability. It does not make a clinical decision, rank patients by symptoms, or recommend treatment.",[14,128,129],{},"In most cases, this use is not a medical device and does not fall within the EU AI Act's high-risk healthcare category. The organisation may still need to tell users that they are interacting with AI, correct inaccurate information, and protect personal data.",[14,131,132],{},"An administrative chatbot does not necessarily remain administrative. If a user asks, “I have chest pain. Can I wait until tomorrow?” and the system continues to generate an answer, the designed boundary and real-world use begin to separate.",[120,134,136],{"id":135},"_2-patient-intake-and-information-collection","2. Patient intake and information collection",[14,138,139],{},"The system collects a patient's complaint, medical history, medication, and contact details, then sends a structured summary to a healthcare professional.",[14,141,142],{},"Collecting or transmitting information alone does not automatically make the system a medical device. But if the summary influences the patient's assessment, the risks of error, omission, and bias grow. A healthcare professional may treat an incomplete summary as a complete record if the system does not reveal what it left out.",[14,144,145],{},"The product boundary needs to be explicit. Is the system collecting information, or is it also making a clinical inference? Adding an output such as “possible emergency,” “most likely condition,” or “appropriate treatment” may change its purpose.",[120,147,149],{"id":148},"_3-emergency-healthcare-triage","3. Emergency healthcare triage",[14,151,152],{},"The system determines urgency, classifies calls, or affects which patient receives a first-response resource first.",[14,154,155],{},"This use is explicitly listed as high risk in Annex III of the EU AI Act. Requirements concerning risk management, data governance, technical documentation, logging, human oversight, accuracy, robustness, cybersecurity, conformity, and post-market monitoring will apply to this route from 2 December 2027. That date does not remove the need to classify the system correctly and prepare now.",[14,157,158],{},"Appearing in Annex III is not always the automatic end of the analysis. The Act allows a limited exception assessment for certain narrow or preparatory tasks that do not materially influence the outcome of a decision. A provider relying on that exception must document the assessment before placing the system on the market. Systems that profile natural persons cannot use the exception.",[14,160,161],{},"Transcribing an emergency call is not the same job as deciding who receives help first.",[120,163,165],{"id":164},"_4-diagnosis-or-treatment-decisions","4. Diagnosis or treatment decisions",[14,167,168],{},"Medical-device law may apply when software is offered for diagnosing, predicting, monitoring, or determining the prognosis of a disease, or for providing information used in a treatment decision.",[14,170,171,172,178],{},"The ",[79,173,177],{"href":174,"className":175,"rel":176,"target":87},"https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Freg\u002F2017\u002F745\u002Foj\u002Feng",[83],[85,86],"EU Medical Device Regulation"," includes software that the manufacturer intends to use for a specific medical purpose within its definition of a medical device. General-purpose software does not become a medical device merely because a healthcare organisation uses it.",[14,180,181],{},"This distinction extends into marketing claims. Instructions for use, technical documentation, sales materials, and product claims all help define the intended purpose. A “consult your doctor” warning may not turn a product back into an administrative aid if it is effectively providing diagnostic or treatment recommendations.",[68,183,185],{"id":184},"what-changes-when-software-becomes-a-medical-device","What changes when software becomes a medical device?",[14,187,188],{},"Medical-device software is not classified by the size of its AI model. Classification depends on which decision uses the information it produces and what could reasonably happen if that decision is wrong.",[14,190,191],{},"Rule 11 of the MDR generally starts software that provides information for diagnostic or therapeutic decisions in Class IIa. Class III may apply if an incorrect decision could cause death or an irreversible deterioration in health. Class IIb may apply when it could cause a serious deterioration or require surgical intervention. Some other software can remain in Class I.",[14,193,171,194,200],{},[79,195,199],{"href":196,"className":197,"rel":198,"target":87},"https:\u002F\u002Fhealth.ec.europa.eu\u002Fdocument\u002Fdownload\u002Fb45335c5-1679-4c71-a91c-fc7a4d37f12b_en?filename=mdcg_2019_11_en.pdf",[83],[85,86],"European Commission's guidance on medical-device software"," explains, with examples, how to determine first whether software qualifies as a medical device and then how to assess its risk class.",[14,202,203],{},"As the class changes, the work around quality management, clinical evaluation, technical files, conformity assessment, CE marking, surveillance, and incident reporting may become more demanding. These are not legal attachments to be added after the product is built. They affect model versions, data quality, performance thresholds, change management, and interface design.",[14,205,206,207,213],{},"The EU AI Act and medical-device law do not replace one another here. ",[79,208,212],{"href":209,"className":210,"rel":211,"target":87},"https:\u002F\u002Fhealth.ec.europa.eu\u002Fdocument\u002Fdownload\u002Fb78a17d7-e3cd-4943-851d-e02a2f22bbb4_en?filename=mdcg_2025-6_en.pdf",[83],[85,86],"Joint guidance from the European AI Board and the Medical Device Coordination Group"," explains that the two regimes apply in a complementary way when a medical-device AI system meets the high-risk conditions. The EU AI Act's high-risk provisions for this product route will begin to apply on 2 August 2028. Existing duties under medical-device law, including conformity, clinical evaluation, and surveillance, remain in force.",[68,215,217],{"id":216},"the-remote-service-boundary-in-medical-tourism","The remote-service boundary in medical tourism",[14,219,220],{},"A clinic operating from Türkiye may assume that its software is outside the picture because it has not been physically placed on the EU market. The position can change when that software forms part of an online diagnostic or therapeutic service delivered to a person established in the EU.",[14,222,223],{},"Article 6 of the EU Medical Device Regulation requires a device to comply with the Regulation when it is used to provide a diagnostic or therapeutic service, online or through another means of communication, to a person established in the EU, even if the device itself has not been placed on the EU market.",[14,225,226],{},"The same article also allows a Member State, on public-health grounds, to require a provider of information-society services to stop its activity. That does not mean every appointment chatbot can be blocked. It does create a direct intervention route when a system qualifies as a medical device and enters remote diagnosis or treatment.",[14,228,229],{},"If a system operating from Türkiye interprets a test result or recommends treatment for a patient established in the EU, “the server is in Türkiye” does not settle the question.",[14,231,232,233,239],{},"The field is not unregulated in Türkiye either. The ",[79,234,238],{"href":235,"className":236,"rel":237,"target":87},"https:\u002F\u002Fwww.resmigazete.gov.tr\u002Feskiler\u002F2021\u002F06\u002F20210602M1-1.htm",[83],[85,86],"Medical Device Regulation published on 2 June 2021"," establishes a product-safety and conformity framework aligned with the EU MDR. Whether software qualifies as a medical device in Türkiye also needs a separate assessment.",[14,241,242],{},"For a medical-tourism company, geography is therefore not a one-line answer. The assessment needs to consider where the company is established, where the patient is located, where the service is delivered, what the software promises, and where the clinical decision is made.",[68,244,246],{"id":245},"health-data-is-a-separate-and-sensitive-flow","Health data is a separate and sensitive flow",[14,248,249],{},"Data-protection duties continue when a system processes health data, even if it is neither high risk under the EU AI Act nor a medical device.",[14,251,252],{},"The GDPR treats health data as special-category data. Processing requires both a legal basis under the general rules and one of the specific conditions in Article 9. Healthcare, public health, scientific research, and explicit consent are different routes with different requirements.",[14,254,255,256,262],{},"Article 6 of Türkiye's KVKK also treats health data as special-category personal data. The Turkish authority's ",[79,257,261],{"href":258,"className":259,"rel":260,"target":87},"https:\u002F\u002Fwww.kvkk.gov.tr\u002FIcerik\u002F8184\u002FOzel-Nitelikli-Kisisel-Verilerin-Islenmesine-Iliskin-Rehber",[83],[85,86],"current guidance on processing special-category personal data"," explains the processing conditions amended in 2024 and the duties of data controllers.",[14,264,265,266,270],{},"The issue in a healthcare chatbot is not limited to the patient record held by the clinical system. If the user's message is sent to a third-party AI provider, questions about processors and subprocessors, retention, international transfers, and security follow. I examine that data path in detail in ",[79,267,269],{"href":268},".\u002Fchatbot-data-third-party-ai-provider","the previous article in this series",".",[14,272,273],{},"“The patient chose to write it” does not remove the organisation's responsibilities. Product design also determines whether the interface encourages free-text disclosure, prevents unnecessary data from being submitted, and moves a sensitive conversation to a safer channel.",[68,275,277],{"id":276},"what-does-human-oversight-mean-in-healthcare","What does human oversight mean in healthcare?",[14,279,280],{},"Giving a doctor or call-centre operator a final approval button is not sufficient oversight by itself.",[14,282,283],{},"The person needs to understand the system's limits, be able to reject its recommendation, inspect the underlying information, and move to a safe alternative path when necessary. If the interface shows only the model's conclusion while hiding uncertainty or missing data, the human may merely rubber-stamp the automation.",[14,285,286],{},"At a minimum, healthcare oversight design should answer these questions:",[22,288,289,292,295,298,301,304,307],{},[25,290,291],{},"Which decision does the system support, and which decision must it never make?",[25,293,294],{},"Which finding triggers urgent human intervention?",[25,296,297],{},"In which situations is the output excluded even when the model appears confident?",[25,299,300],{},"Can the healthcare professional see the input behind the recommendation and the information that is missing?",[25,302,303],{},"Is performance monitored separately across languages, ages, sexes, health conditions, and patient groups?",[25,305,306],{},"When an incorrect direction is detected, how is it recorded, reported, and corrected?",[25,308,309],{},"When the model or prompt changes, is the clinical evaluation reopened?",[14,311,312],{},"The answers cannot remain in a training document. They need to appear in authority design, interfaces, logs, staffing arrangements, and incident management.",[68,314,316],{"id":315},"a-better-starting-point-in-healthcare-define-the-purpose-not-the-feature","A better starting point in healthcare: define the purpose, not the feature",[14,318,319],{},"Companies often describe a healthcare chatbot through a feature list: natural language, multiple languages, 24\u002F7 answers, summarisation, recommendations, and automated routing.",[14,321,322],{},"For compliance and product safety, an intended-purpose statement is more useful.",[14,324,325],{},"Who will use this system, for which user, in which setting, with which inputs, and to support which decision? Which decision must it never make? What is the reasonable consequence of an incorrect output? When does a person take over?",[14,327,328],{},"Without that boundary, it is difficult to assess the dataset, test plan, medical-device class, or duties under the EU AI Act properly.",[14,330,331],{},"Not every problem requires AI. Appointment availability, document checks, and standard preparation instructions may be handled more reliably by a well-designed form and a rule-based flow. If AI's flexibility does not add clinical value, it may add only uncertainty and control cost.",[14,333,334,338],{},[79,335,337],{"href":336},".\u002Feu-ai-act-after-risk-classification","The assessment of company roles, data, and human oversight after risk classification"," applies in healthcare as well. The healthcare-specific difference is that incorrect outputs can have more serious consequences and several legal regimes can overlap.",[14,340,341],{},"The final question should therefore not be, “Can we use our chatbot in healthcare?”",[14,343,344],{},"Better questions are: How does this system affect a patient's appointment, priority, diagnosis, or treatment? Who is responsible for the direction it gives? When that direction is wrong, who will detect it and intervene?",[14,346,347],{},[348,349,350],"em",{},"This article provides general information and does not constitute legal or medical advice for a specific product, healthcare service, or data-processing activity.",{"title":352,"searchDepth":353,"depth":353,"links":354},"",2,[355,356,363,364,365,366,367],{"id":70,"depth":353,"text":71},{"id":114,"depth":353,"text":115,"children":357},[358,360,361,362],{"id":122,"depth":359,"text":123},3,{"id":135,"depth":359,"text":136},{"id":148,"depth":359,"text":149},{"id":164,"depth":359,"text":165},{"id":184,"depth":353,"text":185},{"id":216,"depth":353,"text":217},{"id":245,"depth":353,"text":246},{"id":276,"depth":353,"text":277},{"id":315,"depth":353,"text":316},[369,370],"ai","business",null,"2026-09-05","When is a healthcare chatbot an administrative tool, a high-risk triage system, or a medical device? A practical guide to the EU AI Act, medical-device rules, and data protection.",{"aiUse":375,"aiNote":376},"ai-assisted","The subject, approach, and interpretations in this article were determined by Evren Bal. AI-assisted tools were used for primary-source research and editorial development.",false,"md","\u002Fimages\u002Fhero\u002Fhealth-ai-interface-decision-depth.avif","Similar chat interfaces conceal increasingly consequential appointment, triage, and clinical decision mechanisms","Analysis","en",{},true,"\u002Fhealthcare-ai-chatbot-triage-medical-device",12,{"title":6,"description":373},"AI in Healthcare: Triage and Medical Device Rules","healthcare-ai-chatbot-triage-medical-device",[391,392,393,394,395],"healthcare-ai","eu-ai-act","medical-device","triage","health-data","post","41TG6Z5zgX29WjT__Y-OBBUJKET66Zm6Oa2H8zHZIM8",{"en":399,"tr":400,"de":403},{"path":385,"title":6},{"path":401,"title":402},"\u002Ftr\u002Fsaglikta-yapay-zeka-chatbot-triyaj-ve-tibbi-cihaz-arasindaki-fark","Sağlıkta yapay zekâ: Chatbot, triyaj ve tıbbi cihaz arasındaki fark",{"path":404,"title":405},"\u002Fde\u002Fki-im-gesundheitswesen-chatbots-triage-und-medizinprodukte","KI im Gesundheitswesen: Chatbots, Triage und Medizinprodukte",{"prev":407,"next":371,"others":410,"lucky":529,"readingTime":386},{"path":408,"title":409},"\u002Feu-ai-act-after-risk-classification","EU AI Act: What Should You Do Once You Know the Risk Level?",[411,414,417,420,423,426,429,432,435,438,441,444,447,450,453,456,459,462,465,468,471,474,477,480,483,486,489,492,495,498,501,504,505,508,511,514,517,520,523,526],{"path":412,"title":413},"\u002Fcomprehension-debt-the-bill-comes-due-alone","Comprehension Debt: The Bill Comes Due Alone",{"path":415,"title":416},"\u002Fwhen-does-enterprise-ai-need-rag","When Does Enterprise AI Actually Need RAG?",{"path":418,"title":419},"\u002Fwordpress-to-nuxt-ai-powered-content-pipeline","From WordPress to Nuxt: Building an AI-Powered Content Pipeline",{"path":421,"title":422},"\u002Fhow-llms-identify-experts","What Makes an LLM Recommend Someone as an Expert?",{"path":424,"title":425},"\u002Fproductlog-the-platform-i-built-for-myself-first","ProductLog: The Platform I Built for Myself First",{"path":427,"title":428},"\u002Fthe-ai-productivity-baseline-is-moving-faster-than-we-remember","AI Wasn’t Always This Good. We Just Got Used to It.",{"path":430,"title":431},"\u002Fstart-with-the-business-problem-not-the-ai-model","Start With the Business Problem, Not the AI Model",{"path":433,"title":434},"\u002Fhow-to-route-requests-across-multiple-ai-models","How Should You Route Requests Across Multiple AI Models?",{"path":436,"title":437},"\u002Fdoes-enterprise-ai-really-need-fine-tuning","Does Enterprise AI Really Need Fine-Tuning?",{"path":439,"title":440},"\u002Fis-your-ai-system-delivering-business-results","Is Your AI System Actually Delivering Business Results?",{"path":442,"title":443},"\u002Fone-victory-several-defeats","One Victory, Several Defeats",{"path":445,"title":446},"\u002Fgoogle-generative-ai-data-ai-citation-timing","AI Visibility Dropped Before Search: The Google Data That Changed My Theory",{"path":448,"title":449},"\u002Fhow-to-do-content-pruning-a-real-world-case-study","How to Do Content Pruning: A Real-World Case Study",{"path":451,"title":452},"\u002Fraising-children-in-the-age-of-artificial-intelligence","Raising Children in the Age of Artificial Intelligence",{"path":454,"title":455},"\u002Fredar-ai-powered-summaries-for-kap-disclosures-and-open-sources","Redar: AI-Powered Summaries for KAP Disclosures and Open Sources",{"path":457,"title":458},"\u002Fai-assisted-rest-api-development","Preserving API Quality in AI-Assisted Development",{"path":460,"title":461},"\u002Fwhen-do-you-actually-need-an-ai-agent","When Do You Actually Need an AI Agent?",{"path":463,"title":464},"\u002Fthe-era-of-the-previous-vibe-coder-begins","The Era of the \"Previous Vibe Coder\" Begins: The Invisibility of Clean Code and the Technical Debt Bill of AI",{"path":466,"title":467},"\u002Fopen-publishing-ai-monthly-seo-reporting-experiment","Open Publishing and an AI-Assisted Monthly SEO Reporting Experiment",{"path":469,"title":470},"\u002Ffrom-rules-to-decisions-the-real-time-sales-intelligence-platform-we-built-at-vanity","Medical Tourism Lead Management: How We Moved from Manual Routing to a Real-Time Sales System",{"path":472,"title":473},"\u002Fbeyond-the-bot-lessons-from-building-a-chat-system-for-global-patients","What We Learned Building a Healthcare Chatbot for International Patients",{"path":475,"title":476},"\u002Fdo-you-know-how-dependent-your-company-is-on-ai","Do You Know How Dependent Your Company Is on AI?",{"path":478,"title":479},"\u002Fwhen-is-one-ai-model-enough","When Is One AI Model Enough?",{"path":481,"title":482},"\u002Fwhy-sales-and-other-departments-keep-clashing","Why Sales and Other Departments Keep Clashing",{"path":484,"title":485},"\u002Ftesting-a-button-treating-an-entire-website-redesign-as-a-sure-thing","Testing a Button, Treating an Entire Website Redesign as a Sure Thing",{"path":487,"title":488},"\u002Fai-made-code-cheap-verification-is-still-expensive","AI Made Code Cheap. Verification Is Still Expensive.",{"path":490,"title":491},"\u002Fwhy-im-building-rankextension-making-google-search-console-actually-make-sense","What Happened to RankExtension?",{"path":493,"title":494},"\u002F1m-impressions-per-month-0-revenue-a-programmatic-seo-post-mortem","1M Impressions per Month, $0 Revenue: A Programmatic SEO Post-Mortem",{"path":496,"title":497},"\u002Fhow-can-companies-create-value-from-ai","How Can Companies Create Value From AI?",{"path":499,"title":500},"\u002Fchatbot-data-third-party-ai-provider","What Happens When Chatbot Data Is Sent to a Third-Party AI Provider?",{"path":502,"title":503},"\u002Fllms-txt-was-never-the-point","llms.txt Was Never the Point",{"path":408,"title":409},{"path":506,"title":507},"\u002Fpesintaksit-cash-vs-installments-a-turkish-inflation-aware-payment-comparison-tool","Cash or Installments? – The Story Behind PeşinTaksit",{"path":509,"title":510},"\u002Fan-seo-experiment-in-a-low-competition-serp-with-google-maps-and-openai","Building camiler.org: A Programmatic SEO Experiment with Google Maps and OpenAI",{"path":512,"title":513},"\u002Fis-your-data-ready-for-ai","Is Your Data Ready for AI? Start With the Decision It Must Support",{"path":515,"title":516},"\u002Fwho-captures-ai-productivity-gains","Who Captures the Productivity Gains from AI?",{"path":518,"title":519},"\u002Fai-turned-a-refactor-i-wouldnt-do-into-a-one-hour-job","With AI, a Refactor I Wouldn't Bother With Took One Hour",{"path":521,"title":522},"\u002Fthe-threshold-collapsed-to-zero","The Threshold Collapsed: What ProductLog Taught Me About Building in Public",{"path":524,"title":525},"\u002Fhow-i-built-a-modern-infrastructure-using-open-source-tools-and-the-power-of-cloudflare","Why I Run camiler.org on a Single VPS",{"path":527,"title":528},"\u002Fthe-end-of-coding-or-a-new-renaissance-the-invisible-crisis-of-ai","The End of Coding or a New Renaissance? The Invisible Crisis of AI",{"path":445,"title":446},[531,532],{"path":408,"title":409,"date":372},{"path":533,"title":534,"date":535},"\u002Fwhat-does-the-eu-ai-act-actually-regulate","What Is the EU AI Act, and What Does It Regulate?","2026-09-03",[537,538,542],{"path":499,"title":500,"date":372},{"path":539,"title":540,"date":541},"\u002Fbank-account-api-integration","Integrating One Bank Is Easy. Keeping Forty Running Is Not.","2026-08-27",{"path":418,"title":419,"date":543},"2026-06-22",1788559980504]