<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Jérôme Chambard on Medium]]></title>
        <description><![CDATA[Stories by Jérôme Chambard on Medium]]></description>
        <link>https://medium.com/@jerome.chambard?source=rss-dd84dd98a597------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*ieVUv6O3ou1I72HB.</url>
            <title>Stories by Jérôme Chambard on Medium</title>
            <link>https://medium.com/@jerome.chambard?source=rss-dd84dd98a597------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 03:25:09 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@jerome.chambard/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[AI Overview has landed in France. And it’s great news for marketers who actually work.]]></title>
            <link>https://medium.com/@jerome.chambard/ai-overview-has-landed-in-france-and-its-great-news-for-marketers-who-actually-work-547f4596c742?source=rss-dd84dd98a597------2</link>
            <guid isPermaLink="false">https://medium.com/p/547f4596c742</guid>
            <category><![CDATA[seo]]></category>
            <category><![CDATA[google-ai-overview]]></category>
            <dc:creator><![CDATA[Jérôme Chambard]]></dc:creator>
            <pubDate>Fri, 31 Jul 2026 13:34:27 GMT</pubDate>
            <atom:updated>2026-07-31T13:34:27.880Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="What AI overview think of Taqt" src="https://cdn-images-1.medium.com/max/1024/1*8yQUHQgpAMISPG0EoEvFfg.png" /></figure><p>Everyone knows the tune by heart: to rank first on Google, you need “quality content.” And the magic of backlinks will do the rest. Someone will stumble on your laser peanut-slicing product, rave about it spontaneously on TikTok or YouTube, and boom, the whole planet discovers your amazing product.</p><p>I’ve heard this fairy tale a million times in my career, and if I’ve always performed well in my field, it’s precisely because I never believed it.</p><p>Because, obviously, most of it is nonsense. “Natural” backlinks are rare, the influencers worth their salt are all paid, and editorial sites have already colonized every niche there is, just to resell ad space.</p><p>Even if you’re a plumber in the middle of nowhere, you’ll have to pay up to be visible to your customers online (thankfully, the good tradespeople couldn’t care less, they don’t need new customers).</p><h3>Enter AI Overview</h3><p>With AI Overview, things are shifting in an interesting direction for niche B2B companies like Taqt.</p><p>For unbranded queries, nothing really changes: you still need to get cited by so called “trusted” sources. Naturally, those sources look a bit different these days. Reddit, once a stronghold of actual humans, is now blooming with “suspicious” posts trying to influence LLMs (I too am dying to discuss the cleanliness of public restrooms on Reddit).</p><h3>The bonus for marketers who did the real work</h3><p>On the other hand, there’s a real bonus for marketers who put in the groundwork. The moment a prospect or customer wants to dig deeper into your brand, AI Overview turns out to be a powerful tool. Sure, traffic won’t necessarily flow to your site anymore. Customers won’t read all the prose you carefully wrote on your beautiful website. But AI Overview will summarize it far more effectively than a plain old link ever could.</p><p>At Taqt, for instance, Google’s AI is able to pull details from our product documentation, which is dense, as well as from our partner portal. And what better way to prove a feature exists than to actually document it? Review platforms are also put to good use, judging the brand on the criteria that actually matter.</p><p>The competitor comparisons are strikingly sharp too.</p><p>But of course, winning that match requires having taken the time to write hundreds, if not thousands, of pages to feed AI Overview with genuinely good content.</p><p>And this time (well, for now), it actually works.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=547f4596c742" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Strategy Is Not a Dirty Word. But It’s Dying in Most Organizations]]></title>
            <link>https://medium.com/@jerome.chambard/strategy-is-not-a-dirty-word-but-its-dying-in-most-organizations-6b950e2badef?source=rss-dd84dd98a597------2</link>
            <guid isPermaLink="false">https://medium.com/p/6b950e2badef</guid>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[strategic-marketing]]></category>
            <dc:creator><![CDATA[Jérôme Chambard]]></dc:creator>
            <pubDate>Wed, 17 Jun 2026 06:44:05 GMT</pubDate>
            <atom:updated>2026-06-17T06:44:05.219Z</atom:updated>
            <content:encoded><![CDATA[<p>My grandfather had a saying he repeated often: <em>“Better one fool who acts than ten intellectuals who sit.”</em></p><p>For a long time, I believed him. I’d seen enough meetings where strategy was used as a smoke screen, dressed up in frameworks and decks that went nowhere. So I joined the doers. I moved fast, I reacted, I shipped. And for a while, it felt like winning.</p><p>Then I got burned. Several times.</p><h3>The quiet contempt for strategic thinking</h3><p>In most organizations, especially fast-growing ones, anything that smells like strategic thinking gets treated with mild condescension. Not always openly. More of a slightly weary tone when someone suggests stepping back. An unspoken hierarchy that rewards whoever had the most calls, the most Slack messages, the most output — regardless of whether any of it pointed somewhere meaningful.</p><p>I’ve been on both sides. Pure executor, head down, proud of my velocity. And strategist, trying to slow the machine down long enough to aim it. The difference in outcomes is not subtle. It’s the difference between a company that builds something durable and one that exhausts itself producing impressive activity reports.</p><h3>Opportunism is not a strategy. It’s a hard drug.</h3><p>Euphoric at first. The adrenaline of reacting, the clarity of having an obvious next thing to do, the social reward of being seen as someone who moves. But it slowly drains the organization’s capacity for coherent thought. Each reactive sprint leaves the team slightly more exhausted, slightly less able to ask the question that actually matters: <em>where are we trying to go?</em></p><p>Twelve months in, nobody can figure out why the company is running around like a headless chicken. The answer is always the same: because nobody protected the space where the head does its work.</p><h3>A strategist alone can do nothing</h3><p>Strategy doesn’t fail because it’s wrong. It fails because it gets experienced as noise by the people who need to execute it. If execution-focused teams believe that reacting to every market signal is what great strategy looks like, any serious direction will be drowned out before it takes root. Not experienced as guidance. Experienced as interference.</p><p>This isn’t a caricature. In most organizations I know, this is exactly how it plays out. And the failure isn’t strategic. It’s cultural.</p><h3>So how do you hold the line?</h3><p>Make the cost of opportunism visible. Most companies measure what they shipped. Very few measure what they abandoned, reversed, or started and quietly dropped. When you track that, the cost of reactive pivoting becomes impossible to ignore.</p><p>Give strategy a seat at the table with actual teeth. A CMO who can say “this is not where we’re going, and here’s why we shouldn’t do it” needs organizational cover to do so. Without it, they just validate decisions already made.</p><p>And stop treating strategy as an offsite activity. If it only lives in annual planning cycles, it dies between them.</p><p>The most dangerous person in any organization isn’t the strategist who thinks too much. It’s the leader who mistakes activity for progress, and then wonders why nothing is sticking.</p><p>Speed is not the problem. Directionless speed is.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6b950e2badef" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Conversational CRM Is Finally Here (And It’s Not Vibe Coding)]]></title>
            <link>https://medium.com/@jerome.chambard/the-conversational-crm-is-finally-here-and-its-not-vibe-coding-f3c25b78a04a?source=rss-dd84dd98a597------2</link>
            <guid isPermaLink="false">https://medium.com/p/f3c25b78a04a</guid>
            <category><![CDATA[crm-software]]></category>
            <category><![CDATA[crm]]></category>
            <dc:creator><![CDATA[Jérôme Chambard]]></dc:creator>
            <pubDate>Sun, 19 Apr 2026 16:18:18 GMT</pubDate>
            <atom:updated>2026-04-19T16:18:18.054Z</atom:updated>
            <content:encoded><![CDATA[<p><em>After 20 years of watching CRMs fail salespeople, something fundamental just shifted.</em></p><p>Back when I started my career at Schneider Electric, one of my first assignments was leading a project to build a quoting tool that could generate proposals in under 3 minutes. We called it Express O.</p><p>At the time, the company’s in-house CRM, owned by Marketing, was a nightmare. Sales leadership had spent years trying to kill it and replace it with tools that actually fit the way reps worked.</p><p>We ended up hiring a specialized vendor and building Express O from scratch. It took months. It cost tens of thousands of dollars. It generated millions in quotes. And its maintenance was a never-ending battle, because by definition, it wasn’t connected to anything else in the stack.</p><p>That was 20 years ago. Honestly? Not much has changed.</p><h3>CRMs still don’t fit the way salespeople actually work</h3><p>In every company I’ve worked with since, CRMs have never quite lived up to what sales teams needed. We’ve just learned to live with the compromises. To do “a little bit of everything,” vendors have built bloated platforms where the average user touches maybe 10% of the features. And the pricing is completely disconnected from reality. Paying $50 per user per month for software that manages accounts, contacts, and quotes is a structural anomaly.</p><p>That’s why so many small companies fall back on Google Sheets or Excel. But spreadsheets aren’t designed to link objects together. Something as basic as “tasks tied to a contact” becomes a mess, and you end up with endless comment threads stuffed into “Notes” cells.<br>No-code tools like Airtable have tried to fill the gap. They get you closer, but never all the way. And the closer you push them, the more brittle and complex the setup becomes.</p><h3>A concrete example</h3><p>Say you have different types of customers. Take churches and nonprofits. Each has a completely different profile. For churches, the type of worship matters. For nonprofits, it’s the mission.</p><p>In most CRMs, you handle this by adding conditional fields that apply to one type of record or another. You end up with dozens of fields that confuse everyone, and data quality tanks under the weight of the complexity.</p><p>This is exactly the kind of problem the new generation of AI-native tools solves effortlessly. And to prove it, I spent a Saturday morning building a CRM tailored to this exact scenario. It took two hours. With my wife. Start to finish.</p><h3>This is not vibe coding</h3><p>I need to be very clear about this, because LinkedIn and Medium are flooded right now with “I built an app in 3 hours with AI, look how magical it is” posts. Most of those projects are fragile prototypes built on mangled data models. They end up in the trash two weeks later.</p><p>What we did was the opposite.</p><p>I didn’t ask an AI to “build me a CRM” and hope for the best. The system was specified table by table, relationship by relationship, drawing on 20 years of watching what breaks in commercial CRMs. The LLM and the database were the executors. The brain was the domain expertise.</p><p>That’s exactly why the result holds up. The database runs on Postgres, a battle-tested engine with 30 years of production history. The schema is clean, normalized, properly indexed. The data is queryable by any tool that speaks SQL. This isn’t a toy. It’s a real foundation you can build on.</p><h3>What changes when you pair an LLM with Supabase</h3><p>Supabase is a service that lets you interact with a relational database via API, and thanks to the MCP protocol, with AI. It’s a deeply technical layer, but it unlocks massive possibilities when paired with an LLM like Claude.</p><p>Until now, designing a database was specialist work. You needed to know UML modeling, SQL, and core concepts like indexing and foreign keys.</p><p>Today, an LLM can translate business requirements into a coherent schema, then query and update that database in natural language.</p><p>And once the database is in place, interacting with the CRM becomes trivial. No more opening forms, clicking through three tabs, filling in twelve fields. One sentence is enough.</p><p>The CRM becomes conversational and steps out of the way, so what matters, the customer relationship, takes center stage.</p><h3>Is it a silver bullet? Obviously not.</h3><p>Traditional CRMs still win on a few fronts.</p><p>Speed first. In Salesforce or HubSpot, opening a contact record is one click, in an interface purpose-built for it. Going through an LLM introduces a few seconds of latency, which gets painful on high-frequency, repetitive actions.</p><p>Visual ergonomics second. Established CRMs offer kanban views, dashboards, and advanced filters you don’t rebuild in two hours on a Saturday morning.</p><p>And finally, the maturity of a shipped product: fine-grained permissions, audit logs, compliance, native integrations with hundreds of tools. All of that exists in the Supabase ecosystem, but you have to assemble it yourself.</p><p>In other words: this new model doesn’t replace an enterprise CRM across the board. It makes a level of customization accessible that used to be reserved for Fortune 500 budgets.</p><h3>The economics are a different story</h3><p>Where a traditional CRM runs $50 to $150 per user per month, Supabase offers a free tier that’s more than enough to start, and a Pro plan at $25 per month for the entire organization. You do need to add the cost of LLM usage, which stays marginal as long as you’re not hammering it nonstop, but which can climb if a full sales team is chatting with the CRM all day.</p><p>Even so, you’re still miles below what established CRMs charge. Savings of 90%+ are realistic.</p><p>And the implications go far beyond CRM. Because the database is built to your exact shape, it plugs naturally into your production systems, your marketing stack, your finance tools. Data flows between the pieces without expensive connectors or fragile sync jobs. Auto-enrichment of customer records, opportunity updates triggered by completed tasks, context-aware follow-ups, all the automation dreams suddenly become reachable.</p><h3>What actually changes</h3><p>Technical skill is no longer the limiting factor. What matters now is the ability to articulate what you need, to understand the logic of your own business, and to structure your thinking.</p><p>Could my wife have gotten the same result on her own? Not a chance. Because what made it work wasn’t the tools, it was 20 years of watching CRMs succeed and fail, which made it possible to specify the system while sidestepping the classic pitfalls.</p><p>And that’s the real beauty of this new IT world we’re stepping into: from now on, the differentiator is your ability to define your needs clearly.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f3c25b78a04a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Don’t Ask Your Sales Team What Clients Think]]></title>
            <link>https://medium.com/@jerome.chambard/dont-ask-your-sales-team-what-clients-think-05eb10cbc96e?source=rss-dd84dd98a597------2</link>
            <guid isPermaLink="false">https://medium.com/p/05eb10cbc96e</guid>
            <category><![CDATA[sales]]></category>
            <category><![CDATA[newsletter]]></category>
            <category><![CDATA[marketing]]></category>
            <dc:creator><![CDATA[Jérôme Chambard]]></dc:creator>
            <pubDate>Mon, 13 Apr 2026 06:56:01 GMT</pubDate>
            <atom:updated>2026-04-13T06:56:01.304Z</atom:updated>
            <content:encoded><![CDATA[<p><em>And why a newsletter won’t fix what you think it will.</em></p><p>Last week, a colleague walked into my office with the kind of energy that usually means someone just read a LinkedIn post and decided it was strategy.</p><blockquote>“Our competitors are more visible with clients because they send newsletters. We should do one.”</blockquote><p>I didn’t argue. I asked where that came from.</p><blockquote>“The sales team says clients mention it.”</blockquote><p>There it is. <strong>The sales team says.</strong></p><p>I’ve been in B2B marketing long enough to know that when the evidence chain starts and ends with the sales team, something’s off. Not because they’re dishonest. They’re not. But because they are, structurally, the worst possible source of unbiased market intelligence. And yet, in most tech companies, they’re the only antenna we bother installing.</p><h3>The Filter You Never Agreed to Install</h3><p>Here’s what actually happens when you ask your sales reps what clients are saying:</p><p>You get a sample of clients who talk to reps, which excludes the majority who don’t. You get the clients who mention competitors, which skews toward the ones currently considering switching. You get the rep’s interpretation of what the client said, filtered through whatever narrative that rep is currently pushing internally: more budget, more marketing support, more product features.</p><p>This isn’t cynicism. It’s epistemology.</p><p>A Dentsu study conducted across more than 14,000 interviews between 2021 and 2024 found that 71% of B2B marketers believe their communications express a distinct brand position, but 68% of buyers say that’s simply not true. That gap doesn’t come from bad strategy. It comes from marketing teams building their understanding of the market through intermediaries: sales debriefs, internal anecdotes, secondhand impressions, instead of direct contact with the people they’re trying to reach. When you don’t talk to customers yourself, you end up optimizing for a version of reality that only exists inside your own building.</p><h3>Meanwhile, the Buyer Has Already Moved On</h3><p>Here’s the uncomfortable truth that makes the newsletter debate almost irrelevant.</p><p>According to Forrester’s 2025 B2B Buying Study, B2B buyers complete 70 to 80% of their research before they ever contact a sales rep. The majority of that research happens in channels that are completely invisible to your CRM, your sales team, and your attribution model.</p><p>According to 6sense, buyers now control 61% of their journey before any vendor contact, and 95% of the time, the winning vendor is already on the buyer’s shortlist before first contact occurs.</p><p>Let that land. By the time a prospect talks to your sales rep, the decision is nearly made. The newsletter your competitor sends? It probably didn’t make that list. Your LinkedIn presence, your thought leadership, your review site profile, the podcast episode someone caught on the way to work: that’s what shaped the shortlist.</p><p>And none of that shows up in what your reps “heard from clients.”</p><h3>The Newsletter Fallacy — and the Newsletter Reality</h3><p>Now here’s where it gets interesting. Because we <em>do</em> send newsletters. To our existing customers. And they read them. The open rates say so.</p><p>That distinction matters enormously, and it’s exactly the kind of nuance that gets lost in a hallway conversation filtered through a sales rep.</p><p>A newsletter to an existing customer base is a retention tool. It sustains a relationship that’s already been earned. McKinsey research shows that 80% of long-term B2B value comes from existing customers, not new prospects, and the probability of selling to an existing customer is 14 times higher than acquiring a new one. Sending relevant content to people who already trust you, in a format they’ve opted into, is one of the highest-ROI things marketing can do quietly.</p><p>A newsletter to prospects who’ve never heard of you is something else entirely. It’s an acquisition play pretending to be a relationship. 71% of decision-makers ignore emails because they’re simply not relevant to them. Context is everything.</p><p>The fact that our competitor sends a newsletter tells you almost nothing useful. Who they send it to, what engagement they actually get, and whether it drives pipeline: those are the questions worth asking. We don’t know any of that. The sales rep definitely doesn’t know any of that.</p><h3>We Already Do the Right Thing (We Just Don’t Say So)</h3><p>There’s another layer here that the sales feedback completely misses.</p><p>We don’t blast the same message to every contact in the database. We do segmented communications: the right message to the right audience at the right stage. A feature update for power users doesn’t land in a C-suite inbox. A strategic benchmark doesn’t get sent to someone who’s still in onboarding. This is basic hygiene, but it’s genuinely rare. Segment-based messaging drives up to three times higher engagement in email nurture sequences, and it protects your sender reputation and your relationship with each audience segment.</p><p>The irony is that by <em>not</em> sending generic newsletters to everyone, we’re doing something more sophisticated than most. We’re just not loud about it.</p><h3>What Good Market Intelligence Actually Looks Like</h3><p>When I want to understand what’s happening in our market, I don’t start with a sales debrief. I look at:</p><p><strong>Direct signals from buyers who aren’t talking to us.</strong> Win/loss interviews, G2 reviews, Capterra comments, LinkedIn conversations in communities we’re not part of. 73% of B2B marketing executives rank word-of-mouth and peer recommendations as the most influential factor in deciding which vendors to consider. That conversation is happening whether we’re in the room or not.</p><p><strong>Our own data, unmediated.</strong> Website behavior, content performance, search intent, brand search volume. Noisy, imperfect signals, but not filtered through someone else’s agenda.</p><p><strong>The questions buyers ask before they’re ready to buy.</strong> 89% of B2B buyers have now adopted generative AI as one of their top sources of self-guided information at every phase of buying. If you’re not thinking about what ChatGPT or Perplexity says about your category, your competitors, or your product, you’re blind to a huge share of the consideration stage.</p><h3>The Actual Problem to Solve</h3><p>The real issue isn’t newsletter vs. no newsletter. It’s that marketing is too often flying on input that was never designed to produce useful insight.</p><p>Sales teams are wired to close, not to research. The feedback they bring back from the field is real, but it’s a sample of convenience, interpreted through a lens of self-interest, reported in service of whatever they’re currently trying to get from you. That’s not a character flaw. That’s the job.</p><p>Our job, the CMO’s job, is to see the market as it is, not as it gets reported to us. Which means building intelligence infrastructure that doesn’t depend on the sales team’s memory of last Tuesday’s client dinner.</p><p>The next time someone tells you a competitor is winning because of their newsletter, ask one question: <strong>how do you know?</strong></p><p>If the answer is “the reps said so,” you don’t know yet. You have a hypothesis. Go test it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=05eb10cbc96e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[“Trust me bro” marketing — and the art of nodding politely while dying inside]]></title>
            <link>https://medium.com/@jerome.chambard/trust-me-bro-marketing-and-the-art-of-nodding-politely-while-dying-inside-4ced90f4fdd5?source=rss-dd84dd98a597------2</link>
            <guid isPermaLink="false">https://medium.com/p/4ced90f4fdd5</guid>
            <category><![CDATA[seo]]></category>
            <category><![CDATA[cmo]]></category>
            <dc:creator><![CDATA[Jérôme Chambard]]></dc:creator>
            <pubDate>Mon, 06 Apr 2026 06:55:30 GMT</pubDate>
            <atom:updated>2026-04-06T06:56:50.810Z</atom:updated>
            <content:encoded><![CDATA[<p>You spend your days buried in Semrush data. They spend theirs telling you you’re not ranked on “that one keyword.” Here’s to surviving both.</p><figure><img alt="Everyone’s a CMO on Google search day" src="https://cdn-images-1.medium.com/max/1024/1*p1Ba77wwT0v5e8anHhOmMg.jpeg" /></figure><p>There’s a moment every CMO knows. You’re mid-flow — reviewing keyword clusters, cross-referencing search volume with conversion intent, building a 90-day content roadmap that actually makes sense — when a Slack message arrives.</p><blockquote>Dave from Sales: “Hey, quick one — I Googled ‘<em>best [thing we sell] for [random use case]</em>’ and we don’t show up? Seems like a miss 👀”</blockquote><p>Welcome to <strong>TrustMeBro Marketing</strong> — the discipline of receiving unsolicited SEO feedback from people whose last interaction with Google Analytics was accidentally opening it.</p><h3>The anatomy of a TrustMeBro insight</h3><p>The TrustMeBro insight always follows the same structure. It’s almost beautiful in its consistency.</p><blockquote>Step 1 — The observation: “We’re not ranking for [keyword they just thought of].”</blockquote><blockquote>Step 2 — The implied accusation: A pause. The kind that says “I’m not saying you’re bad at your job, but…”</blockquote><blockquote>Step 3 — The non-ask ask: “Just wanted to flag it. Probably nothing!” (It is, in fact, nothing.)</blockquote><p>What they don’t know — and bless them for it — is that you pulled this exact keyword three weeks ago. Search volume: 40/month. Keyword difficulty: 78. Top results: three enterprise players with domain authorities above 90, all publishing since 2018. You made a conscious, data-backed decision to not go after it.</p><h3>The four types of TrustMeBro senders</h3><p>Because not all TrustMeBros are created equal.</p><p>1. The Curious Exec</p><blockquote><em>“I was just browsing and noticed we don’t come up for [thing]. Is that intentional?”</em></blockquote><p>Genuinely interested, zero context. Saw a competitor’s ad, spiraled. Means well. Needs a 2-minute explainer and will immediately become your biggest internal advocate once they understand how you think.</p><p>2. The Field Reporter</p><blockquote><em>“So I was on a call with a prospect this morning and they said they couldn’t find us when they searched for [very specific query no one ever types]. That’s how they found our competitor instead. Just thought you should know.”</em></blockquote><p>One prospect. One sentence. On one call. Now it’s a strategic crisis. This is anecdotal evidence wearing a trench coat pretending to be market research. The data point is real — the conclusion absolutely is not. They mean well too. The problem is it <em>sounds</em> like signal, so you feel the pull to act. You won’t. But you’ll spend 20 minutes justifying why.</p><p>3. The Competitive Panicker</p><blockquote><em>“Have you seen what [Competitor] is doing with their content? They’re everywhere. We need to do something.”</em></blockquote><p>Just stumbled across a competitor ranking for something. No volume check, no intent analysis, no idea if it converts. Just vibes and mild panic. Give them one slide. One number. Make it big.</p><p>4. The Eternal Dave</p><blockquote><em>“Hey! Not sure if you’ve seen this but…” (every quarter, different keyword, identical energy)</em></blockquote><p>Sends the same type of message quarterly, about different keywords, with the same “just flagging!” tone. Requires a standing 15-min monthly SEO update that pre-answers all his questions. You’ll set this up. It will change your life.</p><h3>Why this keeps happening (it’s not Dave’s fault)</h3><p>Here’s the uncomfortable truth: TrustMeBro insights exist because SEO feels intuitive. Everyone uses Google. Everyone has opinions about what should rank. The gap between “I Googled something” and “I understand keyword strategy” is invisible to the naked eye.</p><p>Meanwhile, you’re operating with an entirely different mental model. You’re not asking “do we rank for this?” You’re asking:</p><blockquote>What’s the actual monthly search volume — and is it growing or decaying?<br>What’s the intent behind this query — informational, transactional, navigational?<br>Can we realistically rank given our current domain authority?<br>If we rank, what does the conversion path look like?<br>What’s the opportunity cost vs. the 12 other keywords on the roadmap?</blockquote><p>That’s not a vibe. That’s a framework. And frameworks don’t fit in a Slack message.</p><h3>The line you’ll learn to love</h3><p>After years of “great catch, let me look into it!” (Reader: you had already looked into it), most CMOs land on some version of this response:</p><blockquote><em>“Good instinct — we actually evaluated that one. Here’s where it sits in our current prioritization and why we’re focusing elsewhere right now. Happy to walk through the logic if useful.”</em></blockquote><p>It’s not defensive. It’s not dismissive. It signals: <em>there is a system, you are in it, and I am its custodian.</em></p><h3>The real flex</h3><p>The goal isn’t to prove Dave wrong. It’s to build enough internal trust that when you do say “we’re not going after that keyword,” people assume it’s because you have a better reason — not because you missed it.</p><p>That trust is built slowly, through consistent reporting, visible wins, and the occasional “remember that keyword cluster we bet on six months ago? Here’s what happened.” It’s the long game. It’s the only game.</p><p>Until then: deep breath, open Semrush, and remember you already knew.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4ced90f4fdd5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Turn Your LoRaWAN Sensors Into a Smart Alarm System with n8n (100% Free & Open Source)]]></title>
            <link>https://medium.com/lorawan-experience/turn-your-lorawan-sensors-into-a-smart-alarm-system-with-n8n-100-free-open-source-a8c30db4f03d?source=rss-dd84dd98a597------2</link>
            <guid isPermaLink="false">https://medium.com/p/a8c30db4f03d</guid>
            <category><![CDATA[n8n-workflow]]></category>
            <category><![CDATA[lorawan]]></category>
            <category><![CDATA[chirpstack]]></category>
            <category><![CDATA[iot]]></category>
            <category><![CDATA[n8n]]></category>
            <dc:creator><![CDATA[Jérôme Chambard]]></dc:creator>
            <pubDate>Sun, 18 Jan 2026 10:25:45 GMT</pubDate>
            <atom:updated>2026-01-18T10:26:26.841Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Simple workflow to trigger alerts from LoRaWAN sensors" src="https://cdn-images-1.medium.com/max/1024/1*OZaejIPL7AQZ7l0CYk5DVA.png" /></figure><p>The world of connected objects is vast, with countless use cases and possibilities when connecting the real world to computerized systems.</p><p>But whatever you want to do with your LoRaWAN sensors, you’ll always be waiting for one or both of these things: a visualization of historical data, and/or “alerts” that allow you to react based on thresholds: a temperature that’s too low requiring the heating to start, a window left open in winter when no one is in the building, or a machine whose vibration signature indicates imminent failure.</p><p>With hindsight and experience, this information is often the most useful, provided it’s properly calibrated (too many alerts become counterproductive).</p><p>Today, I’d like to show you how to create an IoT system capable of configuring and sending you any alert based on data from your sensors, using only free and open-source software that you can host in the cloud or on-premise, that is, on your own machines, behind your firewall.<br>And of course, we’ll use <a href="https://www.chirpstack.io/">Chirpstack</a> as our LoRaWAN Network Server, though any Network Server can be used as long as you can send your data through a webhook or MQTT.</p><p>To manage our alerts, we’ll use software that’s gaining momentum, particularly in AI-driven automation workflows: <a href="https://n8n.io/">n8n</a>.</p><p>n8n is open-source software with a free Community version. It’s the equivalent of Zapier, Make, or Power Automate, with the key difference being that you can install it anywhere, including on your PC. Compared to similar systems like Node-RED, n8n offers a more user-friendly interface that’s “no-code ready.”</p><p>The principle is simple: when a set of sensors from a Chirpstack application sends data, we’ll forward this data to n8n where it will be filtered and analyzed through a “workflow”, a series of processing steps (called steps in n8n terminology). For example, for an alarm system, we’ll send an email if a door opening signal is detected at unusual hours. This is <a href="https://medium.com/lorawan-experience/detecting-intrusions-before-they-happen-with-vibration-sensors-and-lorawan-b596589bcca3">the use case we’ll implement</a> in the rest of this article.</p><h3>Installing n8n</h3><p>If you’re familiar with Docker (and if not, I highly recommend looking into it), you can <a href="https://docs.n8n.io/hosting/installation/docker/">install your own instance in just a few clicks</a>, including on your Synology NAS for example. In my case, I added n8n to my personal mini server with Portainer to benefit from a graphical interface for container management.</p><p>Some important parameters must be added to the default n8n installation to implement our use case (to be set in the container’s environment variables), particularly:</p><p>Parameters for configuring email sending with an external SMTP server:</p><ul><li>N8N_SMTP_HOST, N8N_SMTP_PASS, N8N_SMTP_PORT, N8N_SMTP_SSL, N8N_SMTP_USER</li><li>Your SSL certificate parameters: N8N_SSL_CERT, N8N_SSL_KEY, in combination with N8N_PROTOCOL set to true</li><li>If you’re behind your firewall and using an IP address (not a domain name): NODE_TLS_REJECT_UNAUTHORIZED set to 0</li><li>The timezone that will apply to the instance: GENERIC_TIMEZONE</li></ul><p>Once installed and running, you can access n8n on your local network or remote server.</p><h3>Creating the workflow to detect abnormal door openings</h3><h4>Step 1: Receiving data</h4><p>Once the workflow is created, the first step is to configure an input step that will retrieve data from the LoRaWAN network managed by Chirpstack.</p><p>To do this, we’ll add an MQTT trigger that will “consume” all data from the Chirpstack application where our door opening sensors are grouped.</p><p>The first step is to create an MQTT connection to your Chirpstack instance:<br>If you’re behind your firewall on your home local network, like me, you can use an unsecured MQTT connection on port 1883 of your Chirpstack instance:</p><figure><img alt="MQTT connection with Chirpstack" src="https://cdn-images-1.medium.com/max/1024/1*j0Z0TA7UQ8IJpop-npd2oQ.png" /></figure><p>DO NOT DO THIS ON A CLOUD INSTANCE OR FOR A PROFESSIONAL PROJECT. Instead use MQTTS on port 8883 with TLS certificates (see Chirpstack documentation).</p><p>Once this connection is established in n8n, you can specify the MQTT topic for your Chirpstack application, visible in the Network Server interface, which follows this format: application/xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx/device/+/event/up</p><figure><img alt="MQTT configuration with n8n" src="https://cdn-images-1.medium.com/max/605/1*Y_pVK-uOWh30FYUV8SCWRA.png" /></figure><h4>Step 2: Testing the sensor state</h4><p>Step 2 involves adding an “If” step, which will determine, based on received data, whether or not to trigger the alert:</p><figure><img alt="Detect if the door is open" src="https://cdn-images-1.medium.com/max/657/1*WGbWhgZZTZSm95pMl_QGVw.png" /></figure><p>In our case, we’re using a Dragino vibration sensor that detects door openings: when the door opens, the vibration generated triggers the sensor’s alarm. We simply need to test the variable: {{ $json.message.object.Alarm }}</p><p>If it’s TRUE, we continue to the next step; if not, we stop there.</p><h4>Step 3: Sending the email</h4><p>The final step is to send an email to the user if an opening has been detected. To do this, we’ll use the “Send Email” step, using dynamic subject and content, for example using the variable containing the sensor’s location to know which door was opened (you can use the sensor label in Chirpstack to store this information).</p><h3>Scheduling activations/deactivations</h3><p>Of course, during the day, it’s normal for doors to open. We’ll therefore need to set up a scheduling system to only activate monitoring at times and days when an opening is abnormal. Here too, n8n offers a very simple solution, which involves creating a separate workflow that will activate monitoring, then a workflow that will deactivate it, using a flexible calendar thanks to a “scheduler” step:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/561/1*txPniK4GID98Z6oiyHLtUA.png" /></figure><h3>What’s next? Action!</h3><p>Sending an email in case of intrusion is good. But taking immediate action to deter the intruder is even better!</p><p>In an upcoming article, we’ll see how to use n8n to automatically trigger connected lights via a LoRaWAN Class C smart plug. Imagine this: a door opens at 3 AM, and instantly all the lights in your house turn on, creating the impression of occupancy and scaring off intruders before you even need to leave your bed. IoT serving your security, in real-time.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a8c30db4f03d" width="1" height="1" alt=""><hr><p><a href="https://medium.com/lorawan-experience/turn-your-lorawan-sensors-into-a-smart-alarm-system-with-n8n-100-free-open-source-a8c30db4f03d">Turn Your LoRaWAN Sensors Into a Smart Alarm System with n8n (100% Free &amp; Open Source)</a> was originally published in <a href="https://medium.com/lorawan-experience">LoRaWAN Experience</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Detecting Intrusions Before They Happen with Vibration Sensors and LoRaWAN]]></title>
            <link>https://medium.com/lorawan-experience/detecting-intrusions-before-they-happen-with-vibration-sensors-and-lorawan-b596589bcca3?source=rss-dd84dd98a597------2</link>
            <guid isPermaLink="false">https://medium.com/p/b596589bcca3</guid>
            <category><![CDATA[iot]]></category>
            <category><![CDATA[lorawan]]></category>
            <category><![CDATA[alarm-systems]]></category>
            <dc:creator><![CDATA[Jérôme Chambard]]></dc:creator>
            <pubDate>Wed, 31 Dec 2025 13:29:38 GMT</pubDate>
            <atom:updated>2026-01-02T05:59:05.454Z</atom:updated>
            <content:encoded><![CDATA[<p>Last week, several shops on our street had an unwelcome visitor. Someone went door to door, forcing locks at random, hoping to find cash left in the registers overnight. Most local merchants never leave money on the premises, so our would-be thief walked away empty-handed, but not before damaging doors and locks along the way.</p><p>The city has surveillance cameras, of course. What it doesn’t have is the budget to monitor them in real time. A very useful system indeed: you get to watch the burglar in action the next morning.</p><p>Some shops are equipped with “high-security” alarms that detect when a door opens. The problem? By the time the door opens, it’s already too late.</p><p>As with 93% of burglaries in France, the perpetrator will never be caught. The merchants and their insurers are left to foot the bill.</p><p>My wife’s shop was spared this time, but there had already been a previous attempt. The screwdriver marks on the door frame are still visible, even though there’s literally nothing worth stealing inside.</p><h3>Rethinking intrusion detection</h3><p>This got me thinking about ways to prevent this kind of incident. I remembered something a German partner once told me: vibration sensors are vastly underrated and can do incredible things.</p><p>What if we could detect someone tampering with a lock or forcing a door <em>before</em> it opens? Unlike traditional door-open sensors (the kind I use to <a href="https://medium.com/lorawan-experience/unlocking-potential-monitoring-door-and-window-openings-with-lorawan-8bbc7e05f538">make sure my kids’ windows are closed</a> on the third floor) a vibration-based system could trigger a response while the intrusion is still in progress.</p><p>Imagine automatically turning on lights, sounding an alarm, or activating something more dissuasive, all <em>before</em> any damage occurs.</p><h3>Setting up the test</h3><p>Our building runs a private LoRaWAN network powered by Chirpstack, so I ordered a <a href="https://dragino.com/products/lora-lorawan-end-node/item/350-lht65n-vib.html">Dragino LHT65N-VIB sensor</a> to evaluate how well this concept works in practice.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z3BTCpbrVDkZghcOoAHp6A.jpeg" /></figure><p>I’ll skip over the basics: retrieving the codec, creating the device profile, provisioning via the QR code. That part is classic LoRaWAN: simple and frictionless.</p><p>The real challenge was calibrating the sensor to detect what I wanted without triggering false alarms. That took some time.</p><h3>Understanding the parameters</h3><p>At its core, this sensor is an accelerometer. Getting it right requires understanding two key parameters:</p><p><strong>1. Event duration threshold</strong> How long does the vibration need to last to count as an event? In other words: how long will our burglar be shaking that door? I set this to 1 second, the minimum this sensor allows.</p><p><strong>2. Sensitivity settings</strong> What type and amplitude of movement should trigger detection? After testing, I found this combination works well:</p><ul><li>Frequency: 200 Hz</li><li>ODR (Output Data Rate): 400 Hz</li></ul><h3>The verdict</h3><p>After several rounds of testing, I can say this without hesitation: <em>it works</em>. And it’s impressive.</p><p>A light shoulder bump against the door is enough to trigger an alert, yet the system generates zero false positives during normal operation no phantom alarms when someone simply walks by or when the door remains untouched.</p><h3>What’s next</h3><p>The lab test is complete. Now it’s time to see how it performs in real-world conditions, mounted on the glass door of the shop. I’ll report back in a few days.</p><p>If it works as expected, the next step will be designing and implementing the “response scenario” once an intrusion attempt is detected.</p><p><em>Things are about to get interesting.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b596589bcca3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/lorawan-experience/detecting-intrusions-before-they-happen-with-vibration-sensors-and-lorawan-b596589bcca3">Detecting Intrusions Before They Happen with Vibration Sensors and LoRaWAN</a> was originally published in <a href="https://medium.com/lorawan-experience">LoRaWAN Experience</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Will LoRaWAN Become the Standard for Facility Management?]]></title>
            <link>https://medium.com/@jerome.chambard/will-lorawan-become-the-standard-for-facility-management-28ccae7cfdc7?source=rss-dd84dd98a597------2</link>
            <guid isPermaLink="false">https://medium.com/p/28ccae7cfdc7</guid>
            <category><![CDATA[facility-management]]></category>
            <category><![CDATA[smart-buildings]]></category>
            <category><![CDATA[lorawan]]></category>
            <category><![CDATA[iot]]></category>
            <dc:creator><![CDATA[Jérôme Chambard]]></dc:creator>
            <pubDate>Thu, 25 Dec 2025 15:56:11 GMT</pubDate>
            <atom:updated>2026-01-06T07:30:51.193Z</atom:updated>
            <content:encoded><![CDATA[<h4>As 2025 wraps up, it’s a good time to look back at what happened in the smart building space this year. And one trend stands out: LoRaWAN has moved from “promising technology” to “default choice” for facility management IoT.</h4><figure><img alt="The wireless sensor showdown: who wins in facility management?" src="https://cdn-images-1.medium.com/max/1024/1*sdoBQv3CJPtU0A86oB3Sow.jpeg" /></figure><p>I’ve been writing about LoRaWAN for a while now, and this year felt different. The LoRa Alliance reported smart buildings as their fastest-growing vertical, with double-digit deployment growth. The ecosystem consolidated. BMS integration went from “possible with effort” to “plug and play.”</p><p>So where do we stand heading into 2026? Here’s my take.</p><h3>The real problem with alternatives</h3><p>Let’s start with what doesn’t work, because that’s where most facility managers get stuck.</p><h4>WiFi: a security nightmare for IoT</h4><p>The battery argument against WiFi is getting old. Yes, WiFi sensors used to drain in weeks. But today you can find sensors that last 6 months on a charge. That’s not the real issue anymore.</p><p>The real problem is operational. Every WiFi device needs to be configured with your network credentials. That means someone has to enter the SSID and password on each sensor. Now imagine you have 200 sensors across a building and IT changes the WiFi password for security reasons. Every single sensor stops working. You need to reconfigure them all, one by one.</p><p>And there’s the security angle. WiFi devices are IP-addressable, which means they’re exposed on your network like any other connected device. That’s a significant attack surface. In an era where building systems are increasingly targeted by ransomware, adding hundreds of IP devices that you can’t easily patch is a real concern.</p><p>For high-bandwidth applications like cameras or displays, WiFi makes sense. For a temperature sensor sending 20 bytes every 15 minutes, it’s overkill with serious downsides.</p><h4>Zigbee: good for lighting, less relevant elsewhere</h4><p>Zigbee has its place. If you look at commercial lighting control systems from Signify, Osram, or Legrand, many run on Zigbee. The mesh topology works well when you have powered devices (like light fixtures) that can act as repeaters.</p><p>But here’s the thing: Zigbee was designed for short-range, high-density networks. It operates at 2.4 GHz, the same crowded band as WiFi and Bluetooth. In a modern office building with dozens of access points and hundreds of Bluetooth devices, interference becomes a real issue.</p><p>More importantly, Zigbee requires a mesh network. That means careful planning, coordinator placement, and ongoing management. Each hop adds latency. And while the spec supports 65,000 devices theoretically, real-world deployments with mixed sensor types get complicated quickly.</p><p>For lighting control in new construction? Zigbee is a solid choice. For retrofitting a building with environmental sensors, leak detection, and occupancy monitoring? There are better options.</p><h3>What makes LoRaWAN different</h3><p>LoRaWAN solves the specific problems that matter in facility management.</p><p>First, range. A single gateway covers 4 to 6 concrete floors. I’ve seen deployments where one indoor gateway handles an entire small office building. Compare that to Zigbee, where you need a coordinator plus multiple routers to cover the same area.</p><p>Second, battery life. We’re talking 10+ years for many sensor types. That’s not marketing, it’s physics. LoRa modulation is extremely efficient, and sensors only transmit for milliseconds at a time.</p><p>Third, and this is often overlooked: simplicity. LoRaWAN devices don’t need network configuration. You provision them once with the network server, and they just work. No passwords to manage. No IP addresses to track. When you scale to hundreds of sensors, this operational simplicity matters more than technical specs.</p><p>The tradeoff is bandwidth. LoRaWAN is slow, typically 0.3 to 50 kbps. But for sensor data, you don’t need more. Temperature readings, door status, air quality measurements, they’re all tiny packets.</p><h3>Concrete use cases</h3><p>Let me walk through what I’m seeing in actual deployments.</p><h4>Temperature and air quality</h4><p>This is where most facility managers start. Room-level temperature monitoring sounds basic, but it enables demand-driven HVAC. Instead of conditioning spaces on a fixed schedule, you respond to actual occupancy and comfort levels.</p><p>One Dubai commercial building reported 25% energy reduction after deploying LoRaWAN sensors with smart HVAC controls. The building now adjusts ventilation when CO₂ rises in meeting rooms, before anyone notices the air getting stale.</p><p>In Hong Kong, hundreds of LoRaWAN sensors went into major hospitals through the government <a href="https://www.emsd.gov.hk/filemanager/en/content_1406/GWIN_Introduction.pdf">GWIN network</a>. They track air quality, cold storage compliance, and room occupancy across hundreds of thousands of square meters.</p><h4>Door and window monitoring</h4><p>Open windows in climate-controlled buildings waste energy. Simple magnetic contact sensors on LoRaWAN detect when a window stays open and can trigger alerts or automatic HVAC adjustments.</p><p>The same sensors work for security: monitoring access to roof hatches, electrical rooms, or restricted storage areas. You get a clear picture of activity when the building should be empty.</p><h4>Occupancy tracking</h4><p>Understanding how spaces are actually used is critical for post-pandemic workplace planning. LoRaWAN PIR sensors detect presence without identifying individuals, avoiding the privacy concerns of cameras.</p><p>Microshare reports <a href="https://microshare.io/eversmart-clean/">customers achieving 20% reduction in cleaning costs</a> by switching from scheduled cleaning to usage-based cleaning. If a meeting room wasn’t used, don’t clean it. If a restroom sees heavy traffic, prioritize it.</p><p>Swisscom deployed <a href="https://www.swisscom.ch/en/business/enterprise/success-stories/dienstleistung/post-immobilien.html?srsltid=AfmBOoqXXxZi4Wrqy5C_xNqr6FG-DJA-lZEzzW3pv0OT1sZD7LwkvSMH">LoRaWAN for inspection rounds at Post</a> Real Estate Management. Work times dropped 30%.</p><h4>Customer satisfaction</h4><p>Here’s one people don’t always think of. Wireless feedback terminals with smiley buttons can run for years on LoRaWAN. Taqt (which acquired Skiply in 2024) now equips over thousands sites worldwide, including hundreds of shopping centers and airports.</p><p>At Zurich Airport, Microshare combines occupancy sensors with feedback terminals. If multiple users report a problem in a restroom, it triggers immediate intervention. Otherwise, cleaning follows usage patterns.</p><h4>Leak detection</h4><p>This is where ROI gets obvious. Water damage claims total $300 billion annually worldwide. Insurance costs have jumped 300% since 2017 in some markets.</p><p>A New York condo that was deemed uninsurable after two water events saw premiums go from $150,000 to $850,000. After installing <a href="https://lora-alliance.org/lora-alliance-press-release/lora-alliance-reports-accelerating-adoption-of-lorawan-technology-in-smart-buildings/">320 LoRaWAN sensors via ProSentry</a>, they detected 31 leaks and caught a freezing door incident before damage escalated.</p><p>In Switzerland, MClimate deployed flood detection at the Millennium Business Center in Lausanne. Immediate alerts prevent the kind of cascade damage that results from overnight or weekend leaks.</p><h3>Integration with existing BMS</h3><p>This is where things have changed dramatically in the past two years.</p><p>Modern LoRaWAN gateways now speak native BMS protocols. You don’t need middleware or custom integration. The gateway translates LoRaWAN sensor data directly into BACnet objects or Modbus registers.</p><p>Kerlink and Microshare demonstrated this at the Royal Hospital for Neuro-disability in London. In just a few days, they deployed sensors for bed occupancy, energy use, fridge temperatures, and equipment tracking. The system integrated without impacting core IT infrastructure.</p><p>Milesight released a BACnet driver for Tridium’s Niagara Framework, which is one of the most common building automation platforms. LoRaWAN sensors can now appear as native devices in Niagara, no coding required.</p><p>Wattsense offers gateways with built-in support for Modbus, BACnet, M-Bus, KNX, and others. You can map sensor values to register addresses in a web interface.</p><p>The practical implication: facility managers don’t need to replace their BMS. They can extend it wirelessly, adding sensors where wiring would be prohibitively expensive.</p><h3>Why 2025 was the tipping point</h3><p>A few factors converged this year.</p><p>Regulation is pushing buildings toward granular monitoring. France’s BACS decree mandates energy consumption reductions. New York’s Local Law 157 requires gas detection devices. These aren’t optional, and LoRaWAN offers the fastest path to compliance.</p><p>The ecosystem is mature. The LoRa Alliance reports smart buildings as the fastest-growing vertical, with double-digit deployment growth in 2025. You can source certified sensors from dozens of manufacturers. There’s no vendor lock-in.</p><p>And the economics work. A LoRaWAN network costs roughly 1/10th of cellular for the same coverage. Sensors are cheaper than wired alternatives when you factor in installation labor. Kerlink notes that LoRaWAN infrastructure is about ten times less expensive than cellular, with no spectrum licensing fees.</p><h3>The bottom line</h3><p>LoRaWAN won’t replace everything. WiFi still makes sense for high-bandwidth devices. Zigbee works for dedicated lighting systems. BLE handles asset tracking in dense environments.</p><p>But for the core facility management use cases: environmental monitoring, leak detection, occupancy tracking, equipment status, energy metering, 2025 confirmed that LoRaWAN is now the default choice. It solves the right problems at the right price point, with an ecosystem that’s mature enough for enterprise deployment.</p><p>What to expect in 2026? More consolidation among solution providers. Deeper BMS integration, probably with AI-driven analytics layered on top. And continued pressure from energy regulations pushing buildings toward granular monitoring.</p><p>If you’re still running buildings on manual inspections and zone-level HVAC, the gap is widening. The buildings that figured this out in 2024–2025 already have a structural advantage in operating costs and tenant satisfaction. The rest will catch up eventually. But in facility management, eventually is expensive.</p><p>Water leak detection remains the easiest entry point. The sensors are cheap, the risk mitigation is clear, and insurance adjusters understand the value. From there, you can expand to occupancy, air quality, and beyond. If you haven’t started yet, make it a 2026 resolution.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=28ccae7cfdc7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Critical IoT Applications You Didn’t Know Were Running on LoRaWAN]]></title>
            <link>https://medium.com/lorawan-experience/the-critical-iot-applications-you-didnt-know-were-running-on-lorawan-eea21c351b14?source=rss-dd84dd98a597------2</link>
            <guid isPermaLink="false">https://medium.com/p/eea21c351b14</guid>
            <category><![CDATA[iot]]></category>
            <category><![CDATA[lorawan]]></category>
            <dc:creator><![CDATA[Jérôme Chambard]]></dc:creator>
            <pubDate>Mon, 25 Aug 2025 12:39:27 GMT</pubDate>
            <atom:updated>2025-08-25T12:39:27.031Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Why hospitals, mines, and chemical plants trust “simple” sensor networks with safety</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zfdBa3JQyE_oP_nkYCee0g.png" /></figure><h3>LoRaWAN has a perception problem</h3><p>Most people know LoRaWAN as that “low-power sensor network thing.” Great for reading water meters. Perfect for tracking packages. Brilliant for smart agriculture.</p><p>All true. But that’s like saying a Swiss Army knife is great for opening bottles.</p><p>The same technology monitoring your tomato plants is quietly handling safety systems in chemical plants, managing alerts in hospitals, and monitoring air quality in underground mines. The protocol hasn’t changed. Just how we use it.</p><p>Yet every week, someone tells me LoRaWAN “can’t handle critical applications” because it’s “basically one-way communication.”</p><p>Time to set the record straight.</p><h3>The bidirectional truth nobody wants to hear</h3><p>LoRaWAN is bidirectional. Full stop.</p><p>Your device sends data up. The network sends data down. Revolutionary, I know.</p><p>But here’s where it gets interesting and everyone gets confused:</p><p><strong>Class A devices</strong> open two tiny receive windows after each transmission. It’s like checking your mailbox after sending a letter. You don’t stand there all day waiting — you check at specific times. This isn’t a limitation. It’s genius. Your battery lasts 10 years instead of 10 hours.</p><p><strong>Class C devices</strong> keep their ears open all the time. Like that friend who never puts their phone down. Instant response, but they need to be plugged in.</p><p>The “experts” who dismiss this as inferior to cellular have never tried to power an LTE modem in the middle of nowhere for a decade. Good luck with that.</p><h3>The paranoid approach that actually works</h3><p>I’ve seen too many IoT deployments die silent deaths. Device stops talking. Nobody notices for weeks. By then, your frozen vaccines are spoiled, your worker is injured, or your pump has been leaking toxic waste into the groundwater.</p><p>The fix is stupidly simple: Make your devices neurotic.</p><p>Split your communication into two channels:</p><ul><li><strong>Your actual data</strong> (temperature, location, whatever)</li><li><strong>A heartbeat that never stops</strong> (tiny “I’m alive” messages)</li></ul><p>Think of it like proof of life in a hostage situation. Regular check-ins. Miss one? Sound the alarm.</p><p>Viamed Salud deployed LoRaWAN across 13 hospitals in Spain handling 850,000 patient consultations a year. Their panic buttons ping every 60 seconds. Excessive? Their lawyers don’t think so. Neither do the nurses who rely on those buttons.</p><p>The battery trade-off is real. More paranoia = less battery life. But when your alternative is someone dying alone in a storage room because their panic button was in “power save mode,” you make the trade.</p><h3>Why “good enough” isn’t good enough</h3><p>Most IoT traffic is “fire and forget.” Send the temperature reading. Hope it arrives. Move on.</p><p>That’s fine for your smart garden. It’s not fine for your emergency systems.</p><p>LoRaWAN’s confirmed messages are like registered mail with tracking. The device sends data and waits for confirmation. No confirmation? It tries again. And again. Changes frequency. Boosts power. Whatever it takes.</p><p>A hospital in Hong Kong covers hundreds of thousands of square meters with this approach. Every emergency button press gets through. Not 99%. Not 99.9%. Every. Single. One.</p><p>But here’s what the vendors won’t tell you: Most developers are too lazy to implement confirmed messages properly. They use the default settings, blame the protocol when it fails, and switch to expensive cellular modems that drain batteries in months.</p><h3>Let’s talk about range (and why your WiFi is embarrassing)</h3><p>Your home WiFi barely reaches your backyard.</p><p>A single LoRaWAN gateway covers an entire city district, if it is equipped with the right antenna, the right filters, and is positioned on a high point.</p><p>Not theoretically. Actually. Right now. In production.</p><ul><li><strong>Urban range:</strong> 2–5 km of concrete jungle</li><li><strong>Rural range:</strong> 15+ km of actual jungle</li><li><strong>Record:</strong> 766 km (ok, you need water between the 2 points…)</li></ul><p>The secret? It’s not actually secret. Sub-GHz frequencies penetrate obstacles like they’re not there. Chirp spread spectrum laughs at interference. The signal literally works below the noise floor. Your microwave oven can’t kill it. Your neighbor’s router can’t touch it.</p><h3>The reliability numbers that shut up the skeptics</h3><p>Real deployments. Real data. No marketing BS:</p><ul><li><strong>Packet loss under normal conditions:</strong> 0%</li><li><strong>Packet loss with heavy interference:</strong> 18% (WiFi would be dead)</li><li><strong>Network availability SLA:</strong> 99.9% (that’s 8.76 hours downtime per YEAR)</li><li><strong>Battery life:</strong> 10 years (not a typo)</li><li><strong>Devices per gateway:</strong> 1,000+ simultaneously</li><li><strong>Message delivery time:</strong> 99% within 10 seconds</li></ul><p>A UK medical trial achieved 100% accurate transmission of diagnostic data over 6 km. No retries. First shot. Every time. These are urine tests that determine if someone needs immediate antibiotics or can wait. That’s life-or-death reliability.</p><h3>Why redundancy isn’t optional (it’s mandatory)</h3><p>Every critical system I’ve designed has one principle: Assume everything will fail. Plan accordingly.</p><p>Your LoRaWAN device transmits. Every gateway in range receives it. Not one. All of them. The network sorts out the duplicates. Built-in redundancy with zero configuration.</p><p>Gateways need multiple backhaul options. Fiber primary, cellular backup. Both fail? Satellite kicks in. I’ve seen installations with three redundant paths. Paranoid? The chemical plant that avoided a catastrophe during a flood doesn’t think so.</p><p>Network servers? Distribute them geographically. One data center burns down? Traffic routes to another. Automatically. The devices never know the difference.</p><p>Production sites report running 16,000 devices across 300 gateways, processing 125,000 messages daily. Their downtime last year? Zero. Not “five nines.” Zero.</p><h3>The unglamorous truth about Class C</h3><p>Everyone obsesses over Class A battery life. But Class C is where shit gets real.</p><p>Always listening. Always ready. Sub-second response times.</p><p>Fire alarm triggers? Every device in the building knows immediately. Worker enters restricted zone? Supervisor gets alerted instantly. Gas leak detected? Ventilation systems activate NOW.</p><p>No polling delays. No schedule windows. No “sorry, I was sleeping” excuses.</p><p>Yes, it needs power. So what? Your smoke detector is plugged into the wall too... If not, the batteries may run out without you even noticing. Some things are worth the wire.</p><h3>Heartbeats: The rhythm of paranoia</h3><p>Your gateways should ping every 30 seconds. Miss two beats? Failover engages.</p><p>End devices? Depends how paranoid you need to be:</p><ul><li>Panic buttons: Every 10 minutes</li><li>Door sensors: Every 6 hours</li><li>Temperature monitors: Daily</li><li>Tank level sensors: Weekly</li></ul><p>But here’s the critical part everyone screws up: Randomize the timing. Slightly.</p><p>If 1,000 devices all ping at exactly noon, your network chokes. Add random delays. Spread the load. This isn’t rocket science, but you’d be amazed how many “professionals” miss it.</p><h3>The certifications that separate toys from tools</h3><p>Real safety systems meet real standards:</p><p><strong>IEC 61508</strong> defines failure rates. SIL 3 means less than one dangerous failure per 10 million hours. LoRaWAN systems achieve this with proper design.</p><p><strong>ATEX/IECEx</strong> certification for explosive environments. Oil rigs don’t fuck around with consumer-grade protocols.</p><p><strong>FDA compliance</strong> for medical devices. When the government trusts it for patient monitoring, maybe your smart building can handle it too.</p><p>These aren’t participation trophies. They’re proof that serious people trust LoRaWAN with serious consequences.</p><h3>Stories from the trenches</h3><p><strong>3 million water meters</strong> across France detect leaks before they become floods. Not a pilot. Not a trial. Three. Million. Production. Devices.</p><p><strong>Underground mining operations</strong> monitor oxygen and gas levels in real-time. Workers get alerts when conditions become dangerous. They evacuate based on these warnings.</p><p><strong>Barcelona’s smart city</strong> runs 20,000 parking sensors. Sounds trivial until you realize the reduced traffic emissions are measurably improving air quality. Sometimes “good enough” changes the world.</p><p><strong>Remote diagnostic devices</strong> in the UK achieve perfect transmission over 6 km for medical tests. Results get to doctors faster, enabling quicker treatment decisions.</p><h3>The uncomfortable truth about alternatives</h3><p>Let’s be honest about why people default to cellular for critical IoT:</p><ol><li>It’s familiar (everyone has a phone)</li><li>It’s “everywhere” (except when it’s not)</li><li>Telcos promise 99.99% uptime (in cities)</li></ol><p>But for critical IoT:</p><ul><li><strong>Power consumption kills you</strong> (recharge weekly or die)</li><li><strong>Coverage gaps are real</strong> (basements, rural areas, inside metal buildings) and you can’t do anything about that</li><li><strong>Complexity adds failure points</strong> (SIM cards, data plans, carrier lock-in)</li><li><strong>Cost scales badly</strong> (monthly fees per device add up fast)</li><li><strong>Network sunsets hurt</strong> (remember 2G? 3G is next)</li></ul><p>Your cellular modem needs 100x more power than LoRaWAN. Now imagine maintaining thousands of devices in hard-to-reach places. Still comfortable?</p><h3>How to build systems that don’t suck</h3><p>After seeing enough failed deployments, here’s what actually works:</p><p><strong>Use confirmed messages for anything critical.</strong> Every alarm. Every alert. Every emergency. Get confirmation or retry until you do.</p><p><strong>Deploy more gateways than you think you need.</strong> Overlapping coverage isn’t waste. It’s insurance.</p><p><strong>Implement heartbeats religiously.</strong> Silent failure is the worst failure.</p><p><strong>Plan for the apocalypse.</strong> Multiple backhauls. Geographic redundancy. Battery backups. If it can fail, assume it will.</p><p><strong>Test your worst-case scenarios.</strong> Cut the power. Kill the primary gateway. Flood the basement. See what survives.</p><p><strong>Monitor everything.</strong> Not just devices. Gateways, backhauls, servers, battery levels, signal strength, temperature, humidity. Paranoia is a feature.</p><h3>The wake-up call</h3><p>The real shift in perspective comes when you see the failure modes.</p><p>Cellular-based emergency systems work great — until you’re in a basement, or the local tower gets overloaded during a disaster, or your battery dies after three days.</p><p>LoRaWAN keeps working. Through power cuts, network congestion, even when cellular towers are down. Messages route through multiple gateways to redundant servers. The infrastructure assumes failure and routes around it.</p><p>This isn’t about technology religion. It’s about building systems that work when everything else fails.</p><h3>Stop believing. Start testing.</h3><p>I don’t want you to believe me. I want you to test it yourself.</p><p>Buy a development kit. Set up a gateway. Put a device in your basement, your car, that dead zone where nothing works. Send some messages. Confirm they arrive.</p><p>Then try the same with WiFi. Or cellular. Watch them fail where LoRaWAN succeeds.</p><p>The myth that LoRaWAN isn’t reliable enough for critical systems needs to die.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=eea21c351b14" width="1" height="1" alt=""><hr><p><a href="https://medium.com/lorawan-experience/the-critical-iot-applications-you-didnt-know-were-running-on-lorawan-eea21c351b14">The Critical IoT Applications You Didn’t Know Were Running on LoRaWAN</a> was originally published in <a href="https://medium.com/lorawan-experience">LoRaWAN Experience</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Vibe coding is the future of Home Automation]]></title>
            <link>https://medium.com/@jerome.chambard/vibe-coding-is-the-future-of-home-automation-0e2eb982e3f0?source=rss-dd84dd98a597------2</link>
            <guid isPermaLink="false">https://medium.com/p/0e2eb982e3f0</guid>
            <category><![CDATA[lorawan]]></category>
            <category><![CDATA[home-automation]]></category>
            <category><![CDATA[vibe-coding]]></category>
            <dc:creator><![CDATA[Jérôme Chambard]]></dc:creator>
            <pubDate>Tue, 12 Aug 2025 09:31:58 GMT</pubDate>
            <atom:updated>2025-08-12T09:31:58.165Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Why vibe coding is the future of Home Automation?" src="https://cdn-images-1.medium.com/max/1024/1*b0-Mj9SqMElgJ1AeetI5mg.png" /></figure><h3>How I Used Claude Desktop to Build My Smart Home Dashboard (And Why You Should Try)</h3><p><em>When building personal projects, pleasure comes first. Here’s how I applied vibe coding to create the exact home automation system my family needed, without the hassle.</em></p><p>Last Saturday morning, coffee in hand, I stared at my ChirpStack dashboard. Functional? Yes. Family-friendly? Not even close. My wife couldn’t tell which doors were open without decoding cryptic device names and timestamps. I needed something better.</p><p>In the past, this realization would have meant kissing my weekend goodbye. But I’d been using vibe coding for other projects and knew there was a better way. I opened Claude and started describing my vision for the perfect home automation dashboard. What happened next reminded me why this approach is transforming how we will build personal projects in the future.</p><h3>The Magic of Vibe Coding</h3><p>Vibe coding is what happens when you stop thinking like a programmer and start thinking like a dreamer. Instead of writing code, you have a conversation. Instead of debugging syntax, you refine ideas. It’s the difference between assembling IKEA furniture and describing your dream living room to an interior designer who happens to work at the speed of light.</p><blockquote>“I want a dashboard that shows which doors are open in my house. Make it look modern, work on phones, and be so simple my kids can use it.”</blockquote><p>That’s it. That’s the prompt that started everything.</p><h3>The Conversation That Built My Dashboard</h3><p>What struck me wasn’t just that Claude understood what I wanted: it was how the conversation evolved:</p><p><strong>Me</strong>: “I have a ChirpStack server at 192.168.50.163 with door sensors. I want a simple web dashboard that shows which doors are open.”</p><p><strong>Claude</strong>: “I’ll help you create a Flask-based dashboard. Let me structure this with a modular design so you can easily add different sensor types later…”</p><p>Wait, what? I hadn’t mentioned making it modular. But Claude was already thinking ahead, anticipating that someone building a door sensor dashboard would probably want to add temperature sensors later. This wasn’t just code generation, it was like having a conversation with someone who’d built dozens of these systems before.</p><h3>Why This Felt Revolutionary</h3><p>Let me paint you a picture of the traditional Saturday home automation project:</p><ul><li>9 AM: Start researching ChirpStack’s API documentation</li><li>10 AM: Still reading about gRPC and protobuf</li><li>11 AM: Finally write first line of code</li><li>Noon: Basic API connection working (maybe)</li><li>2 PM: Realize you need to handle edge cases</li><li>4 PM: Start working on the UI</li><li>6 PM: Family asks why you’ve been at the computer all day</li><li>8 PM: “It’s almost working, I swear!”</li></ul><p>Here’s how my Saturday actually went:</p><ul><li>9 AM: Describe what I want to Claude</li><li>9:15 AM: Deploy the generated code</li><li>9:30 AM: First version running (with issues)</li><li>10 AM: Debug why some sensors show “unknown”</li><li>10:30 AM: Realize the data format varies between sensors</li><li>11 AM: Work with Claude to handle edge cases</li><li>Noon: Show actually working dashboard to family</li><li>1 PM: Add features based on family feedback</li><li>2 PM: Done. Happy. Still have half the weekend left!</li></ul><p>The efficiency wasn’t even the best part. It was the joy of seeing my idea come to life without the usual frustration of implementation details.</p><h3>The Perfect Storm: LoRaWAN + ChirpStack + AI</h3><p>Before we dive deeper into the magic, let me explain why this combination is so perfect for home automation hobbyists:</p><p><strong>LoRaWAN</strong> is like the quiet superhero of IoT. While everyone’s fighting with WiFi passwords and dead batteries, LoRaWAN sensors just… work. I installed sensors years ago, and they’re still running on the same batteries. The signal reaches everywhere — basement, garage, even the shed at the back of the garden.</p><p><strong>ChirpStack</strong> is your private LoRaWAN network server. It runs on a Raspberry Pi, costs nothing monthly, and gives you complete control. No cloud, no subscriptions, no company watching when you open your doors. It’s professional-grade software that happens to be free and open source, thanks to the brilliant <a href="https://www.brocaar.com/">Orne Brocaar</a>.</p><p><strong>AI assistance</strong> is the final piece that makes this accessible. ChirpStack is powerful but complex for beginners. Its API uses gRPC with protobuf — not exactly beginner-friendly. But when you can describe what you want in plain English, those complexities melt away.</p><h3>The Dashboard That Emerged</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/390/1*vbeq738tldZ3pH2maMJjbA.png" /></figure><p>What materialized from our conversation wasn’t a rough prototype — it was exactly what I’d imagined, though getting there required some real-world adjustments. The first version worked for most sensors, but a few showed “unknown” status. Turns out, different sensor models report data in slightly different formats. Twenty minutes of debugging with Claude’s help, and we had code that handled all the variations.</p><p>The dashboard now shows each sensor as a card, with open doors highlighted in red at the top of the list. The status updates in real-time, the interface scales beautifully on phones, and it even shows how long ago each sensor last reported.</p><p>But here’s what really captured the magic: I didn’t ask for any of those specific features. Claude inferred them from understanding what a door sensor dashboard should do. When I mentioned it was for family use, it automatically made the status indicators large and color-coded. When I said it should work on phones, it created a responsive design that actually looks better on mobile than desktop.</p><h3>The Hobbyist’s Paradise</h3><p>This is where vibe coding truly shines — and where it’s important to be honest about what “shines” means. As hobbyists, we’re not building banking software or medical devices. We’re building cool stuff for our homes, where an afternoon of tinkering can produce something genuinely useful. Yes, you’ll debug. Yes, you’ll hit edge cases. But instead of spending that time reading documentation, you’re having a conversation about your specific problem with your specific hardware.</p><p>Behind your home firewall, you can focus on what matters: Does it work? Is it useful? Does it bring joy? Most importantly — can you build exactly what YOU want without being constrained by someone else’s idea of what a smart home should be?</p><p>My dashboard checks all three boxes. Could a professional developer critique the code? Probably. Do I care? Not even slightly. My family can now see at a glance which doors are open, and that’s what matters.</p><h3>The Iterative Magic</h3><p>The real magic of vibe coding reveals itself during iteration. After using the dashboard for a day, I had new ideas:</p><p>“Can you make it so doors that have been open for more than an hour flash?”</p><p>Five minutes later, done.</p><p>“I’d like a summary at the top showing total open doors.”</p><p>Three minutes. Done.</p><p>“Can we add a dark mode?”</p><p>Ten minutes, including testing. Done.</p><p>Each iteration felt like a collaborative spell. Describe the desired outcome, watch it materialize, debug the inevitable quirks, refine as needed. Sometimes Claude’s first attempt would miss something — like assuming all sensors report boolean values when some report strings. But that’s where the conversation shines: “The door shows as unknown, here’s what the data looks like…” leads to “Ah, I see the issue. Let me update the parser to handle string values too.”</p><p>No stackoverflow searches, no documentation diving, just collaborative problem-solving.</p><h3>Beyond Code: The Mindset Shift</h3><p>Vibe coding isn’t just about using AI to write code — it’s about changing how we think about personal projects. Instead of “Can I build this?” the question becomes “What would be cool to have?”</p><p>This shift is profound. When implementation complexity no longer determines what you build, your imagination becomes the only limit. Want your lights to flash when the washing machine finishes? Describe it. Want a dashboard that shows which plants need watering based on soil sensors? Just ask.</p><h3>The Technical Foundation (For the Curious)</h3><p>For those wondering about the nuts and bolts, here’s what makes this stack so magical:</p><p>The LoRaWAN sensors I use cost about $25 each and include:</p><ul><li>Door/window sensors that detect open/closed state</li><li>Temperature and humidity sensors</li><li>Motion detectors</li><li>Water leak sensors</li><li>Even soil moisture sensors for the garden</li></ul><p>They all communicate with a single Dragino LoRaWAN gateway (about $200) that covers my entire property. The gateway forwards data to ChirpStack running on an old computer, which stores everything locally.</p><p>The beauty is that once this foundation is in place, building new applications becomes a matter of imagination rather than implementation.</p><p>On AI side, Claude Desktop was my choice because it integrates seamlessly with GitHub. By connecting my repositories to dedicated projects, I can maintain my codebase more efficiently and keep my development workflow organized.</p><h3>Your Own Magic Awaits</h3><p>If you’re sitting on a home automation idea because it seems too complex, too time-consuming, or too technical, I have news for you: those barriers no longer exist.</p><p>Here’s your starter spell:</p><ol><li>Set up ChirpStack</li><li>Get a LoRaWAN gateway and a few sensors</li><li>Open Claude or another AI assistant</li><li>Describe what you want to build</li><li>Watch the magic happen</li></ol><p>Don’t re-use the code of someone else: the joy is in the journey, describing your vision and watching it come to life. Your conversation with Claude will be different from mine, because your home and needs are different.</p><h3>The Future is Already Here</h3><p>We’re living in an unprecedented time for hobbyists. The tools available to us — from $25 sensors with multi-year battery life to AI that can transform ideas into working code — would have seemed like science fiction just a few years ago.</p><p>Vibe coding isn’t about becoming a better programmer or building production-ready systems. It’s about removing the friction between imagination and implementation. It’s about spending your Saturday morning building something cool and your Saturday afternoon actually using it.</p><p>The smart home industry wants to sell you expensive hubs and lock you into their ecosystems. But with LoRaWAN sensors, ChirpStack, and an AI assistant, you can build something better: a system that’s exactly what you want, respects your privacy, and brings a little magic to your daily life.</p><h3>Join the Revolution</h3><p>The next time you think “I wish I had a dashboard that…” or “Wouldn’t it be cool if my house could…”, don’t dismiss it as too complex. Open an AI assistant, start describing your vision, and prepare to be amazed.</p><p>Welcome to vibe coding. Welcome to the future of home automation. Most importantly, welcome to actually building those cool ideas you’ve been dreaming about.</p><p>The magic is real, and it’s waiting for you.</p><p><em>What would you build if implementation wasn’t a barrier? Share your home automation dreams in the comments.</em></p><p><em>Follow for more adventures in hobbyist home automation and the joy of building things that make life a little more magical.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0e2eb982e3f0" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>