$0.58 to break down a company's two years of growth

Someone turned "researching how a company grew" into a reusable skill.

It's called REA-growth, open source, runs on treg. Give it a company name, and it pulls out all of that company's moves over two years: traffic curves, keywords, ad placements, social media posts, backlinks, job postings, then aligns them month by month. Finally it gives you a growth roadmap you can copy.

A complete breakdown of Lovable costs $0.58 in data.

First, what is treg

treg positions itself as the "OpenRouter for tools." Thousands of API endpoints, from SEO backlinks, social trends, person and company info enrichment, to ad management, web scraping, image and video generation, all in one catalog. Pay per call, starting at one cent per call.

It solves a very specific pain point. For an agent to do real work, the tools it needs usually sit behind various subscription walls: Semrush $139/month, Moz $99/month, Crunchbase $99/month, Apollo $59 per seat. You wouldn't buy these just to run one analysis. treg keeps these accounts on hand and charges you fractions of a cent per use.

You can also connect your own keys. Any agent on the team can call them, but credentials never leave the server.

Lovable's two years, laid out

The subject of this breakdown is Lovable. In October 2024, it raised a $7.5M pre-seed, and that month its estimated US organic search traffic was 367 visits.

Over the next eight weeks, they did three things: gave every app a real backend (Supabase), hooked up Google Ads tracking, and launched an affiliate program paying up to $100 per new subscription.

By January 2025, traffic had grown 12x.

The real inflection point was February 9, 2025. They started adding a badge to apps published by free users. If you published an app on the free plan, a small pop-up appeared in the bottom right: "Edit with Lovable."

This is an old playbook: hire free users as your sales force. The difference is the copy. "Edit with" gives you an urge to click. The hardest thing about any new tool is a blank page; one click skips that step. To date, 20,132 websites carry this link, more than 250k backlinks, of which over 130k come from apps users built themselves on lovable.app.

Eight months, $100M ARR.

More interesting are the moves that didn't drive growth

In spring/summer 2025, X was the loudest channel. The biggest post got 2.6M views. The breakdown found no difference in the growth curve before and after big posts. By the time the post arrived, it was already surging.

In March 2026, a YouTube brand film got hundreds of thousands of views. Growth actually slowed.

Instagram creator reels got plenty of likes, but traffic went from rising to flat. None of the 14 founder story carousels made it into the top quartile of their Instagram performance.

The July Google campaign pushed 17 new ads live in four days; traffic peaked that month and then went down. It was like stepping on the gas in a car that was already slowing.

Looking at the timeline, the only paid moves that really line up with growth acceleration were the first wave: November 2025 to February 2026, Google only, and the first video ad ran for nearly a year. Advertisers don't let a losing ad run that long. That period also overlapped with a $330M Series B, so it's impossible to separate how much came from ads.

What can be separated gives one judgment: things that bring views don't bring users; putting usable apps in front of people brings users.

Where the money went

The $0.58 breaks down like this:

Failed calls were not billed. In the end, 343 dated company actions were aligned with monthly traffic.

How to use it

This skill runs in agents like Claude Code, Codex, and Hermes. First install treg, then have the agent read this guide:

treg.to/llms.txt

Then tell it: follow treg.to/skills/rea-growth/SKILL.md to reverse-engineer my competitor's growth.

The skill itself is open source, in the superdesigndev/treg repo on GitHub. Besides rea-growth, the same repo includes several ready-made skills: lead-signals for finding prospects, recruit for turning LinkedIn job posts into a candidate shortlist, and make-ugc for making AI UGC videos.

This kind of research used to take an analyst weeks. Now an agent runs it in ten minutes, and the bill is under a dollar. The change here isn't the analysis itself; it's that the barrier is low enough to casually run it on ten companies each.

One line is worth remembering. Putting "Edit with" on something someone else made works better than any brand film. The badge is a distribution move, not a marketing move.

Original links:

• REA-growth intro page
• GitHub repo
• Meng To's tweet
• Jason Zhou's original post