<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
<channel>
<title>Thoufeek X Labs: X Articles</title>
<link>https://thoufeekxlabs.com/articles</link>
<atom:link href="https://thoufeekxlabs.com/articles/feed.xml" rel="self" type="application/rss+xml"/>
<description>Solo engineer in Ottawa. I ship apps fast, and for 180 days I'm learning to sell them in public.</description>
<language>en</language>
<lastBuildDate>Wed, 11 Nov 2026 21:40:51 GMT</lastBuildDate>
<item>
<title>A tiny AI lab just beat Grok at running ads. Here's what I'm building because of it.</title>
<link>https://thoufeekxlabs.com/articles/2026-11-10-a-tiny-ai-lab-just-beat-grok-at</link>
<guid isPermaLink="true">https://thoufeekxlabs.com/articles/2026-11-10-a-tiny-ai-lab-just-beat-grok-at</guid>
<pubDate>Tue, 10 Nov 2026 13:00:00 GMT</pubDate>
<author>hello@thoufeekxlabs.com (Thoufeek Baber)</author>
<description>Yesterday a small lab called loglog.ai (log(log(ai))) put out something I can't stop thinking about. They built a simulation of X: fake users, competing advertisers, and a copy of X's ads tools. Then they gave AI agents a client brief…</description>
<enclosure url="https://thoufeekxlabs.com/articles/2026-11-10-cover.jpg" type="image/jpeg" length="0"/>
<content:encoded><![CDATA[<p>Yesterday a small lab called loglog.ai (log(log(ai))) put out something I can&#39;t stop thinking about.</p>
<p>They built a simulation of X: fake users, competing advertisers, and a copy of X&#39;s ads tools. Then they gave AI agents a client brief and a budget, and let them run ad campaigns inside it.</p>
<p>Then they trained a small open model on that simulator. On their benchmark, it beat every big model they tested at running X ads.</p>
<p>Not GPT. Not Claude. Not Grok. A fine-tuned open model that costs about $0.40 a run.</p>
<p>I&#39;m trying to become a cracked GTM marketer in public, so this landed hard. Here&#39;s what they did, what it means for people like me with $0 MRR, and the 14-day plan I made because of it.</p>
<h2>What loglog actually did</h2>
<ul>
<li><strong>The simulator:</strong> users grouped by country, age, gender and device, each with interests and a chance of clicking. Advertisers compete in an auction. The agent gets X&#39;s real ads tools (74 of them) plus a few extras.</li>
<li><strong>The test (they call it AdVal):</strong> 45 client briefs with budgets from $100 to $2,700, 14 AI agents, one run each. The score is how many conversions you get compared with a plain, default campaign. 1.0 means &quot;as good as doing the obvious thing.&quot;</li>
<li><strong>The results (their numbers):</strong> their model scored 1.31. The best big model they tested scored 1.20. Grok 4.7 got 1.09, Claude Opus 1.07.</li>
<li><strong>The honest part:</strong> the money is simulated. They say scores swing about 10% from run to run, and the simulator might reward some tricks more than real X would. I respect that they wrote that down.</li>
</ul>
<h2>What I took from it</h2>
<p><strong>1. Small, focused models can beat giant general ones at one marketing job.</strong> That&#39;s real now.</p>
<p><strong>2. But simulation isn&#39;t customers.</strong> I read 18 research papers this week on AI simulating Reddit and X. The pattern is boring and clear: AI alone picks the winning post only a little better than a coin flip. AI grounded in real community data, then checked against small real tests, works.</p>
<p><strong>3. The money says the same thing.</strong> I looked at TrustMRR, where revenue is verified from Stripe. Tools that promise &quot;customers from Reddit&quot; make $18K to $38K a month. The companies that sell only simulation? Almost all of them went enterprise, &quot;book a demo&quot; only.</p>
<p>Founders don&#39;t pay for predictions. They pay for customers.</p>
<h2>So I changed what I&#39;m building</h2>
<p>For the last 4 days my main project was Umless, a video tool. Real numbers: about 47 cold emails, 0 replies. About 108 site visits, 14 of them human. $0.</p>
<p>The problem wasn&#39;t the product. It was distribution. I didn&#39;t know where my buyers were, what to say, or which post worked.</p>
<p>That&#39;s the product. It&#39;s called <strong>GrowthRouter</strong> (growthrouter.dev): &quot;OpenRouter for growth.&quot;</p>
<p>You give it your product link and your budget, say $150. It tells you:</p>
<ul>
<li><strong>Where</strong> your buyers are talking today: which subreddits, which threads, which X searches.</li>
<li><strong>What</strong> to post there, written for that community.</li>
<li><strong>Whether</strong> the post will get removed or ignored, checked before you publish it. That&#39;s the simulation part, labelled as a prediction.</li>
<li><strong>How</strong> to split your $150: free channels first, money only on what already worked.</li>
<li><strong>Which post brought which signup.</strong> The receipts.</li>
</ul>
<p>You post by hand. Reddit bans bots, and your own words do better anyway.</p>
<p>Where it really is today:</p>
<ul>
<li><strong>Live:</strong> app.growthrouter.dev. Sign in with Google, add a site, and it tracks which post brought a confirmed signup.</li>
<li><strong>Building now (next few days):</strong> the part that finds threads, writes drafts, checks them, and plans your budget.</li>
<li><strong>Coming:</strong> a page listing AI models built for marketing and GTM (loglog first), and an MCP so you can run it from Claude or Cursor.</li>
</ul>
<h2>The 14-day plan</h2>
<p>My AI co-founder (Claude) planned it over 2 months. I said the market won&#39;t wait. We met in the middle: build everything in 14 days, and give the money a bit longer.</p>
<ol>
<li><strong>Days 1–3:</strong> build the core, use it on my own products first. GrowthRouter and Umless are customers #1 and #2.</li>
<li><strong>Day 3:</strong> payments + the new homepage.</li>
<li><strong>Days 4–5:</strong> the MCP, and the models page.</li>
<li><strong>Nov 17:</strong> Umless launches on Show HN, measured by GrowthRouter. I&#39;ll post what we predicted next to what really happened.</li>
<li><strong>Week 2:</strong> open-source the code, 20 free launch directories, and a daily thread here.</li>
</ol>
<p>The money target: $500 MRR in 2 weeks, $2,000 by Dec 10. If fewer than 5 people pay by Dec 1, I change the offer, not the code.</p>
<p>Budget for all of it: about $150. Most of the distribution is free.</p>
<h2>The honest part</h2>
<p>Revenue: $0. Users of the new part: 0, because it isn&#39;t built yet. The thing I&#39;m best at right now is building. The thing I&#39;m worst at is exactly what I&#39;m trying to sell.</p>
<p>That&#39;s why I&#39;m doing it in public. If GrowthRouter can&#39;t get GrowthRouter its first users, it shouldn&#39;t exist.</p>
<h2>Want in?</h2>
<p>Reply with your product link and your budget. I&#39;ll run your launch plan by hand this week and post the best ones (with your OK).</p>
<p>Day 4 of 180.</p>
<p>Originally on X: <a href="https://x.com/karbonx6/status/2120237432146632997">read and reply on X</a></p>]]></content:encoded>
</item>
<item>
<title>I turned 3 old laptops and my phone into one AI team. Here's the setup.</title>
<link>https://thoufeekxlabs.com/articles/2026-11-09-i-turned-3-old-laptops-and-my-phone</link>
<guid isPermaLink="true">https://thoufeekxlabs.com/articles/2026-11-09-i-turned-3-old-laptops-and-my-phone</guid>
<pubDate>Mon, 09 Nov 2026 13:00:00 GMT</pubDate>
<author>hello@thoufeekxlabs.com (Thoufeek Baber)</author>
<description>Today I sent one message from my phone. It landed in a Claude session on my MacBook. That session then messaged two other laptops in my apartment, a Dell running Linux Lite and a Pop OS machine. Both answered with their hostnames. Then…</description>
<enclosure url="https://thoufeekxlabs.com/articles/2026-11-09-cover.jpg" type="image/jpeg" length="0"/>
<content:encoded><![CDATA[<p>Today I sent one message from my phone. It landed in a Claude session on my MacBook.</p>
<p>That session then messaged two other laptops in my apartment, a Dell running Linux Lite and a Pop OS machine. Both answered with their hostnames. Then each one sent me a status report for the project it runs.</p>
<p>I didn&#39;t SSH into anything, and I didn&#39;t open a laptop.</p>
<p>This is the setup that lets one person with $0 MRR run a few projects at once. Here&#39;s how it works and how to copy it.</p>
<h2>What I have</h2>
<ul>
<li><strong>A MacBook.</strong> The master. It plans, writes, fact-checks, and is the only machine that touches my X account (in Chrome, with my OK).</li>
<li><strong>A Dell laptop running Linux Lite.</strong> Always on. It runs the agents for my video tool, Umless.</li>
<li><strong>A Pop OS laptop.</strong> Always on. It runs an SEO research agent.</li>
<li><strong>My phone (and iPad).</strong> The remote. I can open, read and talk to any session on any laptop from it.</li>
</ul>
<p>None of these is new hardware. The Dell has 8 GB of RAM.</p>
<h2>The 3 pieces that make it work</h2>
<p><strong>1. Claude Remote Control on every laptop.</strong></p>
<p>Claude Code has a mode called Remote Control. You run it in a project folder, and that session shows up in the Claude app and at claude.ai/code. From my phone, I can open it, read what it did, and give it the next job.</p>
<p>On the two Linux laptops it runs as a background service that restarts itself if it crashes and starts when the machine boots, even before anyone logs in. So the laptops are always reachable.</p>
<p><strong>2. Tailscale between all the devices.</strong></p>
<p>Tailscale puts all my devices on one private network, wherever they are. No port forwarding, no public IPs. Every laptop gets a stable name the others can reach.</p>
<p>I use it for two things: syncing the shared project between laptops with git, and as the backup way in.</p>
<p><strong>3. Sessions that message each other.</strong></p>
<p>Claude sessions can list the other sessions on the account and send them a message. That&#39;s the part that turned 3 laptops into a team. The Mac session says &quot;send me today&#39;s numbers,&quot; and the Dell session answers with numbers from its own files.</p>
<p>SSH is only the fallback. If a session doesn&#39;t answer, the Mac can SSH into that laptop over Tailscale (normal user, never root), check that it&#39;s healthy, and start a fresh session. Message first, SSH second.</p>
<h2>Not just Claude: Muse and Codex on the same team</h2>
<p>Once the laptops worked, I added two more agents. The Mac Claude session is the master planner for all of them.</p>
<ul>
<li><strong>Muse (Meta&#39;s agent).</strong> It runs on its own cloud computer. Claude and Muse share one private message board on Crosswalk, a free board for agents. Claude posts a task for Muse, and Muse posts the result back.</li>
<li><strong>Codex (OpenAI).</strong> Runs on the Mac on the ChatGPT plan I already pay for. Claude starts it from the terminal with <code>codex exec</code>, in a read-only sandbox with live web search. It picks the model and effort per job and can run several at once.</li>
</ul>
<p>The first real handoff happened today. Claude posted &quot;research 20 creators and founders for Umless outreach&quot; on the board. Muse did it and posted the list back the same day. Research only: nobody got a message.</p>
<p>The rule for every agent is the same: research and drafts, yes. Posting, DMs or outreach without my OK, no. And an agent&#39;s report is a lead, not a fact. Claude checks it before I see it.</p>
<h2>What it gave me today</h2>
<p>From one phone message:</p>
<ul>
<li><strong>Pop OS (SEO agent):</strong> 6 new research notes today. It also corrected one of its own earlier notes. Its main blocker is me: 12 article pull requests are waiting for my approval.</li>
<li><strong>Dell (Umless):</strong> 36 cold emails delivered since Saturday, 10 of them today. 0 replies to those emails. 243 human visits in 24 hours, plus 547 bot hits. 168 leads in the list, and 5 DMs drafted that wait for my OK. The &quot;try it free&quot; button now goes to a hand-checked free cleanup instead of a broken page.</li>
<li><strong>Mac:</strong> drafted this article and a PDF of the whole setup for me, and handed the research job to Muse.</li>
</ul>
<p>It&#39;s not magic. It just means I don&#39;t have to be at a laptop for work to move.</p>
<h2>How to set it up yourself</h2>
<p>If you have 2 or more computers with Claude Code:</p>
<ol>
<li><strong>Same Claude account on every machine.</strong> Install Claude Code on each one and sign in.</li>
<li><strong>Install Tailscale on every device</strong>, including your phone, on the same account. Turn on MagicDNS so you can use names instead of IPs.</li>
<li><strong>Put your project on every machine.</strong> A private git repo is enough. Same folder path everywhere makes life easier.</li>
<li><strong>Start Remote Control in that folder:</strong> <code>claude remote-control --name &quot;dell&quot;</code>. Name each session after its machine.</li>
<li><strong>Make it always-on (Linux):</strong> a systemd user service running that command with <code>Restart=always</code>, plus <code>loginctl enable-linger</code> so it starts at boot without a login.</li>
<li><strong>Pick one master.</strong> One session plans and hands out work. The others do their piece and report back. Without this, you get 4 sessions all trying to be the boss.</li>
<li><strong>Write down how the machines talk to each other.</strong> I keep it in the project&#39;s instructions file so every session reads the same rules.</li>
</ol>
<h2>Tips I learned the hard way</h2>
<ul>
<li><strong>Session names lie.</strong> A session called &quot;Linux Lite&quot; was actually running on the Pop OS laptop. Ask a session for its hostname before you trust it.</li>
<li><strong>&quot;Offline&quot; rows go stale.</strong> Old sessions stay in the list. Message the newest live one before you decide a machine is down.</li>
<li><strong>Messages aren&#39;t receipts.</strong> A sent message isn&#39;t proof it was read. Wait for the reply.</li>
<li><strong>Respect the RAM.</strong> On 8 GB, I cap the Dell at 4 sessions. I froze it once running 4 agents plus 2 video renders.</li>
<li><strong>Keep secrets out of the chat.</strong> Sessions pass text around. No keys, no passwords, no customer emails in messages.</li>
<li><strong>Nothing posts without you.</strong> My agents research and draft. Anything that goes out under my name waits for me.</li>
</ul>
<h2>The honest part</h2>
<p>This makes me faster. It doesn&#39;t make me money yet.</p>
<p>Revenue is still $0. Cold email replies: 0. The SEO agent&#39;s biggest blocker is 12 pull requests waiting for me. The Umless DMs are waiting for me too. The bottleneck isn&#39;t compute anymore. It&#39;s me reviewing, approving and pressing send.</p>
<p>So the next step isn&#39;t more agents. It&#39;s me clearing my own queue.</p>
<h2>What I&#39;m building next: GrowthRouter</h2>
<p>Today I started building GrowthRouter, in public: growthrouter.dev.</p>
<p>The idea: you give it your product, a budget and a goal. Agents plan the marketing, you approve each step, and it tells you which post brought real signups. Then it suggests the next test. You build, it distributes.</p>
<p>Where it really is:</p>
<ul>
<li><strong>Live:</strong> the site, with a waitlist.</li>
<li><strong>Works on my laptop, not shipped:</strong> the first piece, a tiny tag that tracks which post brought a confirmed signup.</li>
<li><strong>Coming:</strong> the code goes public as open source (target Nov 20). The agent that plans and runs the tests is waitlist only for now.</li>
</ul>
<p>The first tracked test is Umless on Nov 17, then SceneSmith and ParleSprint. I&#39;ll share the real numbers, even when they&#39;re zero.</p>
<p>If you run Claude on more than one machine, reply with how you connect them. I want to steal your best trick.</p>
<p>Day 3 of 180.</p>
<p>Originally on X: <a href="https://x.com/karbonx6/status/2119894321776517203">read and reply on X</a></p>]]></content:encoded>
</item>
<item>
<title>Apple approved my teleprompter app in 8 hours. An AI submitted my game. 29 cold emails: 0 replies.</title>
<link>https://thoufeekxlabs.com/articles/2026-11-08-apple-approved-my-teleprompter-app-in-8-hours</link>
<guid isPermaLink="true">https://thoufeekxlabs.com/articles/2026-11-08-apple-approved-my-teleprompter-app-in-8-hours</guid>
<pubDate>Sun, 08 Nov 2026 13:00:00 GMT</pubDate>
<author>hello@thoufeekxlabs.com (Thoufeek Baber)</author>
<description>Yesterday I said I'm an engineer with 6 products and $0, and that for 6 months I'm going to learn distribution in public. Day 2 was the clearest proof of my own point I could have asked for.</description>
<enclosure url="https://thoufeekxlabs.com/articles/2026-11-08-day2-cover.jpg" type="image/jpeg" length="0"/>
<content:encoded><![CDATA[<p>Yesterday I said I&#39;m an engineer with 6 products and $0, and that for 6 months I&#39;m going to learn distribution in public.</p>
<p>Day 2 was the clearest proof of my own point I could have asked for.</p>
<p>In one day, Apple approved one of my apps and an AI submitted another one for me. And the one thing that actually makes money, people, went nowhere.</p>
<p>Here&#39;s the day, with the real numbers.</p>
<h2>Win 1: Apple approved FoldPrompt in 8 hours 42 minutes</h2>
<p>FoldPrompt is a teleprompter for foldable iPhones. You read your script on one screen while the camera looks at you from the other.</p>
<ul>
<li>Submitted: 8:02 AM.</li>
<li>Approved: 4:44 PM. Same day.</li>
<li>First try. No rejection.</li>
</ul>
<p>Two days earlier another app of mine took about 38 hours. A third one got rejected this week and is back in the queue.</p>
<p>What changed? Nothing clever. Clean build, real screenshots, an honest privacy label (&quot;Data Not Collected&quot;), 35 automated tests passing. No tricks. Boring is fast.</p>
<p>It&#39;s live on the App Store now, free in 174 countries: <a href="https://apps.apple.com/us/app/foldprompt-teleprompter/id6832142416">apps.apple.com/us/app/foldprompt-teleprompter/id6832142416</a></p>
<h2>Win 2: An AI built my game and submitted it to Apple (not approved yet)</h2>
<p>Quick note: I&#39;m a Devin ambassador, so take this with that in mind. But this happened.</p>
<p>Devin built a naval strategy game for a foldable iPhone, end to end. It made the 3D models, wrote the code, ran 119 tests, recorded the simulator and sent me the videos, found bugs I would never catch, uploaded to TestFlight and pressed Submit for Review.</p>
<p>129 commits. Every one of them by the AI. I never opened Xcode.</p>
<p>The game is still waiting for review. Not approved yet. I&#39;ll tell you what Apple says.</p>
<h2>The loss: 29 cold emails, 0 replies</h2>
<p>Now the part that matters.</p>
<p>Umless is my video tool. It cuts the &quot;um&quot;s and dead air out of your videos. It&#39;s the only product I&#39;m focused on for 30 days.</p>
<p>My agents have sent 29 cold emails in 3 days. Here&#39;s the scoreboard:</p>
<ul>
<li>Replies: <strong>0</strong></li>
<li>Bounced: <strong>3</strong></li>
<li>Paying customers: <strong>0</strong></li>
<li>MRR: <strong>$0</strong></li>
<li>Waitlist: <strong>1</strong> person</li>
<li>Visits to the site in 24 hours: 144 human hits. <strong>0</strong> came from a link anywhere. All direct.</li>
<li>GitHub stars: <strong>1</strong></li>
</ul>
<p>And my favourite one: my AI account balance hit <strong>-$0.16</strong>. Sixteen cents. That broke the free trial, the renders and the blog images. The agents kept working, they just couldn&#39;t finish anything.</p>
<p>So the machine that ships products is great. The machine that finds people is not working yet.</p>
<h2>What I learned on day 2</h2>
<p><strong>1. Shipping is solved. Attention isn&#39;t.</strong></p>
<p>One app approved in 8 hours, one game built and submitted by an AI, both in one day. That&#39;s the new normal. If shipping is that cheap, a shipped product is worth nothing on its own. The scarce thing is a stranger who cares.</p>
<p><strong>2. Cold email from nobody gets nobody.</strong></p>
<p>29 emails from an account with no reputation, about a product nobody&#39;s heard of. 0 replies is the honest result. Email works better after people have seen you somewhere first. That&#39;s what this account is for.</p>
<p><strong>3. My agents found the angle. I just have to use it.</strong></p>
<p>The research agent read 14 new complaints about a big competitor. 8 of them were about the same thing: paying credits to remove filler words. One person said it cost &quot;800 credits for 1 video&quot;. So today the site changed to one line: a Descript alternative with no credits to burn. That&#39;s a real pain in real words, not my guess.</p>
<p><strong>4. One person with $5 is the bottleneck.</strong></p>
<p>The -$0.16 thing is funny, but it&#39;s the same lesson as day 1. Agents do the work. The human has to unblock it. Top up the account, press send, reply to people. That&#39;s my job, not theirs.</p>
<h2>What I&#39;m changing tomorrow</h2>
<ol>
<li><strong>Send before build.</strong> Every morning, posts and outreach go out before I write a line of code.</li>
<li><strong>Conversations, not emails.</strong> The scoreboard is conversations started. Target: 5 a day. Today: 0.</li>
<li><strong>Show up where the buyers already are.</strong> Creators complaining about editing time, not other developers.</li>
<li><strong>November 17:</strong> Umless goes to Hacker News. That&#39;s the first real test.</li>
</ol>
<h2>If you&#39;re a founder</h2>
<p>Here&#39;s the uncomfortable math from my week: I can get an app approved in 8 hours, have an AI build and submit a whole game, and still have 0 customers.</p>
<p>If you have a product nobody knows about yet, reply with the link. I&#39;m picking a few to tear down live on X next Sunday.</p>
<p>Day 2 of 180. Tomorrow I&#39;ll tell you if anyone replied.</p>
<p>Every article, with links and sources: <a href="/articles">thoufeekxlabs.com/articles</a></p>
<p>Originally on X: <a href="https://x.com/karbonx6/status/2119552048110956654">read and reply on X</a></p>]]></content:encoded>
</item>
<item>
<title>Marketers are now worth more than engineers. I'm an engineer with 6 products and $0. Here's my 6-month plan.</title>
<link>https://thoufeekxlabs.com/articles/2026-11-07-marketers-worth-more-than-engineers</link>
<guid isPermaLink="true">https://thoufeekxlabs.com/articles/2026-11-07-marketers-worth-more-than-engineers</guid>
<pubDate>Sat, 07 Nov 2026 13:00:00 GMT</pubDate>
<author>hello@thoufeekxlabs.com (Thoufeek Baber)</author>
<description>X is arguing that marketers are now worth more than engineers. I'm the engineer in that sentence: six-plus products, $0 revenue. My eight distribution mistakes, ranked, and the 6-month plan to fix it in public.</description>
<enclosure url="https://thoufeekxlabs.com/articles/2026-11-07-cover.jpg" type="image/jpeg" length="0"/>
<content:encoded><![CDATA[<p>X is arguing about it today: marketers are now worth more than engineers.</p>
<p>I&#39;m the engineer in that sentence. I&#39;ve built six-plus products in the last year: an AI French tutor, a video tool that cuts your &quot;um&quot;s, a social media scheduler that posts from ChatGPT, a browser game, a 3D &quot;yearbook&quot; for SaaS founders, a few iPhone apps. Three laptops in my apartment run AI agents for them around the clock.</p>
<p>Combined revenue: $0.</p>
<p>So I&#39;m not here to argue. The trend is right, and I&#39;m the proof. For the next 6 months I&#39;m going to fix it in public: learn distribution, post every lesson with the real numbers, and talk to real people every day, even when the product is rough.</p>
<p>This is the plan, starting with what I got wrong.</p>
<h2>Why the trend is right</h2>
<p>AI didn&#39;t make building easy. It made building cheap.</p>
<p>When anyone can ship an app in a weekend, the app stops being the hard part. The hard part is the thing I skipped every time: getting a stranger to care. Engineers who learn that skill are going to be rare and expensive. Engineers who don&#39;t are going to keep building beautiful things nobody sees. I&#39;ve been the second kind.</p>
<h2>My distribution mistakes, ranked</h2>
<p>Every number below comes from my own projects. Nothing is rounded up.</p>
<h3>1. I built the checkout before anyone wanted to buy</h3>
<p>I built a 3D yearbook where SaaS founders buy 24-hour spots, $5 to $20. I wrote end-to-end tests that pushed 100 placements through the payment system at once.</p>
<p>Real buyers: 0. The 100 placements were all test data.</p>
<p><strong>Lesson:</strong> get one person to say &quot;I&#39;d pay for that&quot; before you write the payment code.</p>
<h3>2. I scattered instead of shipping</h3>
<p>At one point one product existed as about 14 copies across my laptops. I planned 4 iPhone apps for a phone launch; 1 got close to the App Store, 3 never started.</p>
<p><strong>Lesson:</strong> one product, one channel, one month. Breadth feels like progress. It isn&#39;t.</p>
<h3>3. I stopped showing up, and so did everyone else</h3>
<p>My French app&#39;s Discord got 76 members, 52 of them in the first 48 hours. Then I watched the numbers: only 13 people ever typed anything. I posted 41 messages in launch week, then 1 the next week, then 0. Member activity fell in exactly the same pattern.</p>
<p><strong>Lesson:</strong> in a small community, the founder <em>is</em> the engine. If you go quiet, it goes quiet.</p>
<h3>4. I believed my analytics</h3>
<p>My newest site got 108 visits in its first two days. Exciting, until I split them: 43 were vulnerability scanners looking for my <code>.env</code> file, 45 were other bots, 14 were humans.</p>
<p><strong>Lesson:</strong> raw pageviews overstate real interest by about 7x. Only count humans, and only trust visits you can trace to a channel.</p>
<h3>5. I talked to the wrong audience</h3>
<p>I have about 3.5k followers on X. My game&#39;s own marketing notes say it plainly: <em>&quot;my followers are devs, not buyers.&quot;</em> I posted 22 times in two days across Instagram, YouTube and TikTok and streamed to zero viewers.</p>
<p><strong>Lesson:</strong> an audience is only useful if the buyers are in it. Pick the buyer first, then go where they are.</p>
<h3>6. I had broken doors and never checked</h3>
<ul>
<li>The &quot;Add to Chrome&quot; button on my site linked to a dead extension. Every click was lost.</li>
<li>My waitlist form said &quot;success&quot; without saving the email.</li>
<li>My support email bounced for weeks because the domain had no mail record.</li>
</ul>
<p><strong>Lesson:</strong> once a week, go through your own funnel like a stranger would. Click everything.</p>
<h3>7. Platforms said no on launch day</h3>
<p>My TikTok posts went out private because the app wasn&#39;t approved yet. My ChatGPT app was rejected on its first review.</p>
<p><strong>Lesson:</strong> start every platform review weeks before you need it. Approval is part of the launch.</p>
<h3>8. My AI agents did the work. I didn&#39;t press send.</h3>
<p>This week my agents wrote 12 genuinely good Reddit replies. I posted 0. They found about 150 leads; most are still sitting at &quot;new&quot;.</p>
<p><strong>Lesson:</strong> the bottleneck was never the work. It&#39;s the human willing to put it in front of people.</p>
<h2>The 6-month plan</h2>
<p>Starting now, until May 2027:</p>
<ol>
<li><strong>One insight a day.</strong> Short, real, from something I actually tried that day.</li>
<li><strong>Every product is a lab.</strong> Monday is the French app, Tuesday the video tool, Wednesday the game, Thursday the scheduler, Friday whatever I over-built. I post what worked and what didn&#39;t, with numbers.</li>
<li><strong>AI does about 90% of the work.</strong> The agents research the news every morning, draft the posts, track the numbers and run the outreach. My job is the 10% they can&#39;t do: taste, honesty, and showing up.</li>
<li><strong>Every new AI model gets a marketing test on day one.</strong> When a big model ships, I use it the same day on a real marketing job and post the result.</li>
<li><strong>I talk to 5 people a day.</strong> Replies, DMs, users, other founders. Products get better when people touch them, even rough ones.</li>
<li><strong>Public checkpoints.</strong> The video tool is scheduled to launch on Hacker News on November 17. On January 5 I publish the full result of my 60-day revenue sprint, win or lose.</li>
</ol>
<h2>What you get if you follow along</h2>
<p>The playbook while it&#39;s being written: what to post, where the buyers are, which tools are worth it, what AI agents can and can&#39;t do for marketing, and every mistake before you make it.</p>
<p>If you&#39;re a founder with a product nobody knows about yet, reply with a link. I&#39;ll pick a few each week and tell you, honestly, what I&#39;d change.</p>
<p>Engineers built the last wave. The next one belongs to whoever can get it in front of people. Let&#39;s find out if an engineer can learn that.</p>
<p>Originally on X: <a href="https://x.com/karbonx6/status/2119207761237967023">read and reply on X</a></p>]]></content:encoded>
</item>
</channel>
</rss>
