Finance YouTube has no accountability. alphaYT exists to add it — every pick time-stamped, benchmarked against SPY and QQQ, and ranked by consensus so you can see what holds up before you risk a dollar.
It started like most side projects do: watching a YouTube video when I should have been doing something else.
I came across a video by adriconomics titled "I Watched 200 Finance Videos. Buy These 14 Stocks!". In it, he manually watched finance channels, logged the stocks creators mentioned, and compiled them. It was a neat project, but as a developer, whenever I see someone doing repetitive manual work, my brain immediately asks the same question: Can I automate this?
For a while, the idea just sat in the back of my mind. I didn't act on it immediately.
Then August 2026 arrived, and the LLM ecosystem went a bit wild. xAI released Grok 4.6, Meta dropped Muse Spark 1.2, and DeepSeek released DeepSeek Flash 0731. That model was absurdly capable for its price point. At the same time, I had opencode with the oh-my-openagent plugin installed on my laptop, but I hadn't really pushed it to its limits yet. I was itching to test what an agent setup could actually do if given a real, long-running, multi-step task — something I could fire off and let run autonomously for hours to see where it would break.
I realized I already had the second key piece: the Gemini API. Since Gemini can parse full YouTube video URLs directly without needing yt-dlp, Whisper transcriptions, or local audio processing, the extraction pipeline was suddenly much simpler.
So I gave the agent a single high-level prompt: replicate the idea behind the video, but scale it up and build a complete system to automatically audit the picks and benchmark their performance against the market.
I watched the initial setup unfold. Different models spent the first hour debating system architecture, schema design, and execution flows before settling on a plan of 43 discrete tasks. Then I let it run.
When I checked back, I was honestly blown away. What started as a chaotic stress-test for an agent ended up as alphaYT: a full, production-ready terminal auditing 20 finance YouTubers, computing split-adjusted entry prices, tracking open/closed positions, and serving precomputed consensus views through Next.js and Supabase. You can see the exact methodology behind every calculation.
I'm keeping the site completely free to access. Down the road, I might place a few non-intrusive ads just to cover the millions of API tokens consumed during backfills and daily ingestions, but the core data remains public.
opencode with oh-my-openagent
DeepSeek Flash 0731, Grok 4.6 (which I regret because of the price), Muse Spark 1.2 (during the community version stage), GLM 5.2
Gemini API (google-genai direct video parsing)
Python 3.12 (uv), yfinance (curl_cffi), Supabase Postgres
Next.js (App Router), TypeScript, Vercel
If you have feedback, feature requests, or want to collaborate on the project, feel free to drop me an email at youmassi@gmail.com.
The board is live — five tiers from Watchlist (1–3 creators) to Maximum Consensus (8+), each position benchmarked and each ticker one click from its full audit trail.
Explore the consensus →View the leaderboard →Read the methodology →