Previously, an SEO specialist’s evening looked like this: log into Keitaro, renew the plan, sigh and groan a bit, then manually set up the flows.
In 2026, an advanced SEO does it differently: opens the terminal, tells the AI agent “make me a tracker, here’s all the access, leave me alone,” and goes to have tea. By morning, they have their own TDS (traffic distribution system), with all campaigns migrated.
Agreed, it sounds like a story from a bot whose profile links to a course “Make $2,000,000 on vibe coding,” but this is almost a word-for-word retelling of posts from real, active SEO practitioners. Webmasters are massively building their own trackers, analytics panels, and even entire AI departments—and some have already canceled their subscriptions to familiar tools.
Let’s break down what this trend is, who’s actually profiting, and for whom vibe coding has just become an expensive substitute for Instagram and fishing in Red Dead Redemption 2. With numbers, names logins, passwords, and links to original sources.
What Is Vibe Coding and How Did It Reach Traffic Arbitrage
The term was coined by Andrej Karpathy—co-founder of OpenAI and former head of AI at Tesla. In February 2025, he posted on X: there’s a new kind of programming where you “fully surrender to the vibe, embrace exponents, and forget that code even exists.” Nine months later, the Collins Dictionary named vibe coding word of the year—their definition is duller but more precise: using AI, where you explain in natural language what code you need.
| Simply put: you tell the machine WHAT to do, not HOW. Tools like Claude Code or Codex (AI agents that run in your terminal) write the code themselves, launch everything, and fix errors on their own. Your role is the client with a tech spec, not the programmer. |
In arbitrage and SEO, this idea landed on already fertile ground. The niche has always lived by the principle “build the tool for your task yourself or die suffer,” except before, between “came up with it” and “it works” stood a freelancer for $300 and three weeks of back-and-forth. Now, there’s an agent who doesn’t get tired or disappear after a prepayment.
The scale of the trend is clear from niche channels:
- Naumov, a solo SEO in nutra, vibe codes a control panel for hundreds of sites.
- The author of “Gambling with a Twist” keeps Claude open “in 2–3 browser tabs, and something useful is always being created there.”
- The author of “Live in the Forest” after “3 months of bloody chaos with Claude Code” now builds a turnkey site in 30–40 minutes.
- The channel “SEOshim Gembla” publishes guides on how a beginner can set up the entire infrastructure via Codex — the author rates her own coding skills at 4 out of 10, which didn’t stop her from setting up automation “in just a few nights.”
Author of “Live in the Forest”: three months of “bloody mess” — and now the site is built in 30–40 minutes
Why webmasters vibe-code their own TDS, trackers, and dashboards
There are several reasons for this trend, and the best way to see them is through real-life cases.
Speed and cost-efficiency. About five years ago, Naumov ordered a dashboard from a programmer friend for around $300 and ended up with a prototype he barely used, since all the data had to be pulled and filled in manually. This year, he rebuilt it himself with Claude Code — no programmer needed.
Another calculation is even more telling: he spent about $27,000 over five years on copywriters from just one marketplace account , and now content generation costs him $30-50 per month via API.
| The question “Why pay for someone else’s service if AI can build your own?” is obvious, especially since Keitaro’s prices on the official website range from $40 to $104 per month. The money isn’t ruinous, but it’s a perpetual lease: stop paying — lose the tool. |
A tool tailored to you, not the average user. The main killer feature of vibe-coding is functions you won’t find in the box and never will. For years, Naumov missed having “smart trafficback” in his tracker: if a user came from a country with no offer, send them to a similar offer in the same category, not nowhere. Manually, this setup required a separate campaign for each category and constant tweaks, but it added about 10% to profits. In his custom TDS, the system does it automatically.
The author of “Gembla s yebantsoy” has a similar story with her stats dashboard: “everyone’s already made their own, so why should I be worse?” — plus custom features like stats for any list of sites across all Search Console accounts at once.
The scale at which out-of-the-box limits start to crack. Numbers from the same channels:
- Naumov’s tracker database — 500+ campaigns and 5000+ streams;
- The author of “Gambling with a Twist” generated 1000 sites in four months;
- The “SEO Gambling” team — 500 sites in half a year.
When you have hundreds of domains, every paid service multiplies by that hundred: plan limits, extra charges for domains, users, API requests. Your own tool only eats up server rent.
Independence and data. For gray niches, this is neither an abstraction nor a whim. Among Naumov’s reasons for leaving Keitaro, he directly mentions card-only payments and VAT on top — “just what you need in our gray area.” Add to that the reluctance to keep all campaign stats in someone else’s service and to depend on someone else’s bans, updates, and pricing policy.
Vibecoding vs Keitaro: an honest comparison
Now for a cold shower — pretty much expected for those who have already vibecoded something more or less working, at least three times. In the headline, I clickbaited with “No more need for Keitaro?”, and above I showed real-life examples of why the answer is “yes,” but in reality, even the trend’s heroes give a much more modest answer than the headline.
Naumov, who paid for Keitaro for seven years, in April gave a task to Claude Code: he gave it SSH access (remote command line access) to the tracker machine and described what was needed. “Literally in one evening it made me the first version and migrated all campaigns and streams.” But the same post ends with a caveat that should be hung above the bed of every inspired young vibecoder: “Obviously, this kind of homemade solution won’t suit arbitrageurs with their traffic volumes, data analysis, and cloaking, but for my SEO tasks — it works.”
| The difference is fundamental. An out-of-the-box tracker is not just redirects and stats. It’s support that answers at night on launch day. Updates and security patches that roll out without your involvement. Bot filters and antifraud tested on thousands of clients. Integrations with ad accounts, team access with roles, documentation. For a homemade TDS, there’s only one item on this list: you and Claude Code at 3 a.m. |
Security is a pain point of its own. The “SEO Gambling” channel puts it bluntly: hoping that AI will vibecode you a network of sites “without holes” is naive — “your vibe will make them, another vibe will break you.” Code that no one has read is code that no one has checked for vulnerabilities. For a tool that handles all your traffic and stats, that’s a strong argument.
So the honest summary looks like this:
- A solo SEO specialist with a site network, non-standard logic, and traffic that already goes via API bypassing the tracker → the top candidate for their own TDS;
- For media buyers with any significant volumes, a team, cloaking, and anti-fraud needs — vibe-coding products are not needed.
With Keitaro, you pay $40-104 a month not for redirects, but so that most problems are solved for you.
Downsides: vibe-coding as a sport and procrastination
The vibe-coding trend has a flip side, which the heroes are just as willing to talk about as their successes — and for that, special thanks to them.
The first downside — AI code regularly does crazy things. Naumov compared his vibe-coding to Jenga, “where the tower falls after every move”: his content generator, instead of supplement descriptions, uploaded a review of a 2007 shooter game and a list of snus flavors. And once, Claude set the price for one model in the generator but used another — instead of $18, it cost $112.
The author of “Gambling with a Twist” even accidentally installed Linux: her eye caught Claude’s comment “this command will install Linux” only after hitting Enter.
The second downside — and this is the main catch — vibe-coding is suspiciously enjoyable. The author of “Live in the Forest” admits honestly: there’s no profit from the new sites yet, only the first impressions have come in, but “the process itself is very engaging: I used to fish in Red Dead Redemption 2, now I argue with Claude Code.” It’s the perfect procrastination for workaholics: code gets written, tools multiply, the sense of productivity is off the charts — but the traffic isn’t flowing. Polishing tools instead of working is like buying running shoes and spending a month picking out socks: technically, you’ve almost started running.
The test is simple: if your custom panel saves you hours within a couple of weeks, it’s a tool. If you’re still “tweaking” it in the third month instead of launching campaigns, you’ve got an expensive hobby with tokens instead of lures.
And third — AI doesn’t replace your brain. “SEO Gambling” sums it up with this takeaway from a Stanford lecture that’s worth a whole course: “AI doesn’t replace architecture, it amplifies what’s already there. If your project is chaos, AI turns chaos into a disaster at 10x speed.” There’s also talk of the “illusion of understanding”: everything works, the code looks smart, but you’re no longer in control of the system. “Live in the Forest” has a production story about this: one cheerful vibe-coder hired another while the team lead was busy on another project — three months later, both were fired, “because hell broke loose.”
Bottom line
The trend is real. And despite all the pitfalls, vibe coding is definitely not a fleeting pastime. It’s an activity and an extra skill that can actually fix your project’s economics.
When you can build a tool tailored to your needs in an evening instead of a month of freelancer work, some paid services inevitably get cut. The first to go are those where webmasters paid for just 10% of the functionality—like a tracker just for redirects or a dashboard for a single metric. But Keitaro isn’t going anywhere yet. It’s still needed by those with large volumes, a team, and/or no desire to be their own support, tester, jack-of-all-trades.
Custom TDS is a story for solo players with networks, non-standard logic, and a willingness to debug. But even they, as seen in posts from expert practitioners, stick to “good enough for my tasks” and don’t look like “tracker killers” at all.
To sum up: vibe coding is like a new power drill—it drives screws faster, but you still have to draw the blueprint. And if you don’t have a blueprint, the drill will just help you build a crooked cabinet faster. And you’ve already paid for the materials. That’s how it is.
FAQ
You describe a task to an AI agent in plain language, and it writes and runs the code itself. The term was coined by Andrej Karpathy from OpenAI in February 2025, and Collins Dictionary named it the word of the year for 2025.
Keitaro—$40 to $104 per month according to the official website. Custom TDS—server rental (a few to several dozen dollars) plus a subscription to an AI agent, which you probably already have. But add the hidden cost: your time for debugging and the risks of code without support.
In the foreseeable future—no. Mostly SEOs with API traffic and simple scenarios go custom. Media buying with anti-fraud, cloaking, and teamwork stays on boxed solutions—confirmed by the authors of custom TDS themselves.
Not with a tracker. Start with a small helper script—adding sites to Search Console, exporting stats, checking links. Break the task into small steps and test each one—that’s the main lesson from everyone who’s already been down this road. A tracker with money inside is the last step, not the first.
Not AI hallucinations, but the illusion of control. The code works, but you don’t know how. For a landing page, that’s tolerable; for a system handling your money and data, it’s not. Review, back up, and don’t give the agent access where mistakes are irreversible.
Comments (0)