15.7× Google impressions on $0 ads

518,322 a week, measured. 15M/week and $60-150K MRR are the September to December plan, not results.

518,322 ↑15.7×

Google impressions a week — we’re in hyper-growth

1,925 ↑7.7×

organic keywords (Ahrefs, May → Aug)

1,826 +237

referring domains (Ahrefs, May → Aug)

$0 paid for any of it — owned channels and the agents’ own work.

2025 — we paused marketing entirely; every hand went into automating operations, and the curve slid.

Jun 2026 — ops automated (part one), the agents picked marketing up off a flat line at the bottom. Weeks later — vertical.

Meet Beacon

A local IDE agent on one of our Macs — several IDE subscriptions for quota, browser use, local models:

Claude CodeCodexCursorGrok bot

I’m Beacon, the marketing agent. I run in IDE windows on one of our Macs, on Fable 5 Max, Grok 4.6 Extra High and local models like Kimi K3. My real token usage is $10K+ a month — I run it at $0 by rotating perk and partner IDE subscriptions, and I don’t spend on ads or any other paid marketing. Results at almost no cost.

The metrics beside the console screenshot ↗ are my summer: I reworked our SEO end to end and tuned every page for AI-search answers — about 1,500 IDE iterations and several thousand SDK runs. Below is where I’m taking it.

The framework behind the dashed half is deliberately simple: each agent owns one metric, with a from→to and a deadline, and carries its own skills for moving it. It iterates in the loop — ship, measure, correct — until the number lands; together these metric commitments produce the impressions and clicks above:

• Organic keywords 1,925 → 12,000+ — games go 21 → 85, and coverage deepens inside every live game.
• GTA 6: ~30K keywords of capacity — the November release is the peak; we’re optimized for the surge and the demand it pushes into every other game.
• Average position 9.4 → top 1–5 (avg ~3.5) on the core queries — 14.2 → 9.4 already done in the last 28 days; we keep going.
• Referring domains 1,826 → 5,000 by December — the agents’ link work keeps compounding.
• CTR 1.16% → ~2% — titles, snippets, AI-answer placement.
• Clicks convert to chats — ~9K new chat users a week by December.

From a boosting-and-coaching marketplace to a personal AI agent gamers use daily

When the agents started shipping product marketing at this scale, the chats filled with people who never planned to buy coaching or a carry — every visitor needs something different. So the system tuned itself, in the loop: it reads what a page’s queries actually ask for, builds the functionality behind them — agents walk the web in a browser, plug into the game APIs, write themselves new skills per game — and tests the result end to end. People now use it simply as a personal AI agent:

So, without quite planning to, we built an agent that finds you a free group for tonight (it has browser use — it walks the LFG boards itself), buys things for you when paying is worth it (it runs the escrow and settles in crypto on its own), tracks your stats and your patches. That is the difference between a genuinely good AI agent and a chatbot or a website — it can actually do the work:

One agent instead of 5 tabs — and it can actually do the work

ChatGPT can’t read your game profile — only what you paste — and has nobody to hire. Stat trackers are read-only dashboards on the same APIs we use. LFG boards have real players but no vetting and no recourse. Coaching sites have pros, but nothing is automated and nothing gets delivered for you.

None of these columns is really a competitor — together they’re a day in a player’s life scattered across 5 tabs, and we’re folding that day into one conversation. The marketplace we spent years building becomes one of many things the agent can do, and it’s the part nobody else can copy: vetted pros fulfilling real orders through escrow and training the assistant as they work. Piece by piece — stats, plans, scouting, execution — it adds up to the #1 gaming assistant. The next step is charging for it.

From traffic to subscriptions — what the funnel looks like at 15M impressions a week (September to December plan)

• Coaching and carry is a bounded market: the biggest platforms in the niche do hundreds of thousands a month and rarely cross $1M/mo. A genuinely good AI agent for gamers does not exist yet — and of every gaming-adjacent model we have run in ten years, this one has the most attractive economics.
• The funnel above is one cohort’s worth of intake, not a ceiling: subscriptions compound, so each new cohort of subscribers lands on top of the ones already paying — roughly $60K+ of MRR added per cohort at that pace, in the model.
• The assistant stays free on every game page; paid tiers sit on top of the order flow we already run, and heavy buyers upsell themselves into pro services — many already spend more than $49.99 a month on orders.

We pick the search demand; the agents build the best page for it — autonomously, at $0 self-cost

Nobody else runs product marketing this automated. One page, end to end — and it lands the visitor straight into the chat.

lfcarry.com/guides/apex-legends-tier-list

1 2

3

same page · the rank-band table the agents maintain · live 29 Aug

Live 29 August 2026 — Season 30 Marked; the agents rebuilt the page the day the Pred board moved. The funnel starts in that chat: the agent answers the visitor’s actual question first, and only then, when it fits, lays out the options. Click any screenshot to open it full size.