In the group, the discussion about open-source bots mostly boils down to one consensus:

No one will open source a bot that makes money.
It sounds like common sense, but this principle has become sharper in the AI era—
Previously, the moat was "ability to write code." Now, GitHub hands it over to AI, which can read the entire architecture in seconds. The rewriting cost drops from two months to two hours.
The moat shifts from "ability to write code" to "ability to strategize."
I'm also doing this layering myself—
Polymarket-toolkit I open source (the repository for calling such tools). It has high reuse value, low threshold, and open-sourcing is like building a brand.
For market-making/taker strategies in the H series, I do not open source. These include sigmaD1 calibration, market reprice thresholds, and empirical parameters for adverse selection. Making these public is like giving researchers a direct path—once the edge is commoditized, it’s gone, regardless of whether there was an original edge.
There’s also a middle ground: methodology can be shared, but specific parameters cannot. The paid source code package for pm-quant (a strategy + encrypted zip + 1-on-1 deployment support). The paid barrier = filtering out peer competition.
Someone pointed out a master’s address, gabagool, in the group: runs better at 5 min / 15 min / hourly levels, "regardless of market conditions, it just flies." His code is not on GitHub. The market has already voted with its feet.
So, the idea that "open-source bots are unprofitable" is a survivor bias—those who make money have no motivation to open source.
You see repositories with thousands of stars, full of praise, and ongoing updates—most are content products disguised as quant strategies: open source is an entry point to attract subscriptions/payments, not the core revenue.
Tool-layer open source, strategy-layer closed source, methodology shareable, parameters not shareable—this boundary is quite clear in the PM quant community.
The moat hasn't disappeared; it’s just shifted from "ability to write code" to "ability to strategize," from "tools" to "judgment."
Data scraping / research addresses on PM: