AI model training is shifting into a scarcity problem: human-origin data is limited, AI-generated data is compounding, and the last fresh permissioned datasets should become more valuable. That makes RDDT central, because it sits on timely human discussion, has only 7% user growth, and is already seeing 5% of all content ever written on the platform arrive in the last quarter. The setup extends to licensing and M&A: top model companies will have to pay for RDDT data, renewal prices should reset much higher, and a 20% acquisition probability is enough to justify related trades. Google, XAI, X, YouTube, Yelp, and SHUT are all framed through how much human signal they can offer or how they may react to RDDT news.
AI training is heading toward a self-referential loop: human source data will become too small versus AI output, so the training signal will "build on itself" and risk getting worse as AI data is replicated again and again.
Owning the last fresh, high-signal permissioned human data is framed as a no-brainer for big AI labs, because they need real human judgment rather than older books or stale archives.
GOOG/Google is said to have a data advantage from video data and email data, while some other AI labs have to license data or scrape it.
XAI is described as likely advantaged by x.com data, while X itself is treated as strong because it is topical and timely.
Reddit data is singled out as highly human because it captures back-and-forth arguments, differences of opinion, and the "journey," not just answers.
X is described as more limited than Reddit even though it has similar debate-style behavior, because its strongest use cases are narrower topics like technology, politics, and finance.
YouTube is framed as breadth plus original content, but more YouTube content is now being created by AI, making newer Gemini releases something to watch.
A person’s X profile allegedly began posting after AI agents were placed on it, with the agents watching episodes and writing posts.
RDDT plunged after META announced a competitor to Reddit, but the reaction is viewed as overdone because META lacks the connection points and the right structure to compete.
The structural gap is that Facebook/Meta is organized around personal relationships, while Reddit is organized around shared interests; Threads is cited as proof that people do not want to dig into a Meta-owned platform to share deep opinions.
Reddit is described as not broken, with a strong user experience and value proposition, but it is also under attack from AI content infiltrating the platform.
Over the next 5 years the world is expected to be consumed by AI slop, which should increase demand for platforms where humans still want to read other humans’ opinions and discussion.
RDDT is called the number one destination in the world for human interactions and content, and a precious platform if it keeps executing.
RDDT user growth was only 7% in the last reporting cycle, but 5% of all the content ever written on Reddit was written in the last quarter, implying engagement is surging unless AI generated much of it.
It is not possible to tell how much of that content lift is AI-driven, but the entire Reddit corpus is said to be available on the dark web.
RDDT lets researchers use its data for free while charging commercial users, and outside the U.S. it is unclear whether every AI training operation is paying.
Top model companies are expected to pay and not be able to sneak in and steal Reddit data; eventually dozens and dozens of huge tech entities are expected to pay for it.
Reddit data pricing is expected to rise exponentially as deals expire and renew, and each higher renewal should become a major story if the cost reset is real.
Real-time topical data is said to require platforms like X or Reddit, because old books and other older data do not provide up-to-date information.
RDDT is expected to be acquired someday or sign one or more major licensing deals at rates significantly above today, creating a short 5-to-10-minute trading window and a tiny arbitrage opportunity in related names.
YELP and SHUT are named as smaller correlated reactions if RDDT gets an acquisition bid or a major licensing headline, with the claim that the rest could pop on that news.
RDDT was bought more just now, with the position described as moderate and just outside top-four/top-five sizing.