Everlyn-1 video generation model : Claimed to be the "first open-source autoregressive video model," its frame-by-frame video generation method is similar to GPT's "next word prediction." This contrasts with the slower diffusion models commonly used by competitors. Academic research has confirmed that autoregressive (AR) models have potential in decoding speed and generating long sequences, but are prone to error accumulation. Diffusion models, on the other hand, excel in generation quality but are computationally expensive. Everlyn chose the autoregressive approach, aiming to achieve breakthroughs in both speed and length, claiming to be able to generate 1080p resolution, 8-second videos in just 4 seconds while reducing costs by 10x. Open Source: A core differentiating selling point of Everlyn AI is its open source nature, promising "fully open source model weights," a direct challenge to the industry's status quo of closed-source models like OpenAI's Sora. Existing documentation provides relatively consistent descriptions of Everlyn's innovative AI features. However, descriptions of its Web3 features vary widely, with some not even found in official documentation. I summarize its Web3 narrative as follows: With the proliferation of deepfake technology, the need to verify the authenticity of digital content has become more urgent than ever. Everlyn claims to be building a dedicated Layer 1 public blockchain that will provide immutable proof of origin and copyright verification for AI-generated content by recording timestamps and creator information on-chain. In this way, videos created through Everlyn will leave a public and transparent record on the blockchain, making it easy to identify and trace if used to create deepfakes. This narrative shares similarities with the logic behind IP platforms like Story Protocol and NFTs. Everlyn has built the DePIN platform, where users can contribute their GPU computing power to "video generation" and receive compensation. This narrative may be related to io.net's investment, sparking market associations with its DePIN narrative. Furthermore, it also encompasses the more common creator economy narrative, where creators receive token incentives for publishing videos on the Everlyn platform. These claims precisely address the current market pain points: the high cost of AI video production, the monopoly of closed-source models, and the difficulty of verifying content authenticity. However, there's a significant gap between hype and reality. The AI component: Real but overly packaged. Despite the team's impressive background, there are glaring contradictions in their technical promotion: Confusion in terminology: The project claims to utilize an "autoregressive model" architecture, yet its promotional materials refer to "xDiT (Distributed Diffusion Transformer)." The Diffusion Transformer is a core component of the Diffusion Model and represents a distinct technical approach from the Autoregressive Model. This confusion of fundamental concepts is concerning. Codebase Status: While the Everlyn-Labs organization exists on GitHub, its codebase primarily contains code related to team members' published academic research projects (such as ANTRP and Wasserstein-VQ), and does not represent a unified, production-grade video generation system. Furthermore, the team's GitHub repository does not contain any blockchain-related code. Lack of Performance Verification: The project claims to be "the fastest video generator on the planet," but lacks independent third-party benchmark verification. While models like CogVideoX and CausVid have already gained widespread recognition in the open-source community, Everlyn-1's actual competitiveness remains unproven. Web3: The "Ghost" of Non-Existence This presents the most serious problem for the entire project. The project claims to build a decentralized video AI layer, but after a thorough search, I was unable to find any relevant Web3 technical white papers, developer documentation, public testnets, block explorers for observing on-chain activity, or any code libraries involving Web3 technology.
In short, as for the Web3 part of Everlyn, except for the real existence of coin issuance (its token is located on BSC, the address is 0x302DFaF2CDbE51a18d97186A7384e87CF599877D), everything else remains at the conceptual narrative level and is completely lacking in technical details. Despite lacking any code implementation for its Web3 platform and offering only hollow narratives (it's worth noting that Everlyn's official Web3 narrative is relatively restrained, with many of the eye-catching claims stemming from overinterpretations by key opinion leaders), the company is eager to issue a token. This pattern of behavior reeks of a familiar "leek-cutting" scheme. The Truth About Celebrity Endorsements: A project's credibility depends largely on its founding team. I've verified the backgrounds of the two founders, Dr. Harry Yang (Co-founder & CTO) and Dr. Ser-Nam Lim (Co-founder & Head of Research/CEO), and found no issues. Both individuals have indeed worked at Meta and have made significant contributions to image AI. Besides the two founders, all materials touting the project cite Turing Award winner and Meta's Chief AI Scientist, Yann LeCun, as an advisor to Everlyn AI as a key selling point, bolstering the project's academic achievements. However, my research has proven that LeCun's "advisor" status is highly inflated. The only official source supporting this claim is a tweet from the Everlyn team stating, "We're honored to welcome Yann LeCun as an academic advisor to Everlyn." However, I checked LeCun's personal website (yann.lecun.com), Meta AI page (ai.meta.com), LinkedIn profile, and all related posts on his X account, and found no mention of Everlyn AI or his advisory role. Furthermore, LeCun's posts and interviews primarily discuss general AI topics, with little mention of Web3, and certainly no mention of Everlyn. Therefore, I believe that this use of LeCun's influence is a classic example of "celebrity marketing"—magnifying a trivial interaction to create the illusion of a formal endorsement. The Players Behind the Scenes: Investment Logic and Marketing Machines Mysten Labs' Strategic Investment Everlyn AI secured $15 million in funding led by Mysten Labs, the Sui development team, at a valuation of $250 million. However, it's worth noting that neither Mysten Labs' official blog nor the Sui Foundation's channels have released an official statement regarding this investment. This silence leads me to believe that Mysten Labs' investment is, at best, a strategic acqui-hire—acquiring a top AI team and their technology to enrich the Sui ecosystem, rather than an endorsement of the project's token economics or decentralized roadmap.
Interpreting the Real Signal of “Binance Listing”
Being listed on Binance is also a highlight of the project's publicity, but we need to accurately understand the real signal behind it:
LYN tokens are listed on the "Binance Alpha" platform, not the Binance main site spot trading area.
Binance Alpha is a high-risk sandbox dedicated to trading "emerging digital assets not listed on the Binance exchange."
Binance explicitly warns that being labeled an Alpha asset does not mean it will be listed on the main platform in the future. These assets are subject to increased volatility and risk, and may result in total investment loss and inability to withdraw cash. Therefore, listing an Alpha asset on Binance is not an endorsement of quality, but rather a trading experiment conducted under the premise of isolating risks. A Kaito-driven marketing machine Everlyn AI's astonishing popularity is no accident; it's backed by a systematic marketing and promotional machine. Kaito's Role: Everlyn AI used Kaito's Capital Launchpad platform for its $2 million public offering. Kaito describes itself as an "AI-powered vertical search engine," and its launchpad allocates quotas based on metrics such as users' social reputation. Notably, Everlyn AI's airdrop explicitly requires users to post in Kaito's "yaps" section. Paid Promotion Mechanism: Kaito's business model includes "Yap-to-Earn" competitions, in which the project establishes a reward pool to "incentivize creators to produce content about the company." This directly confirms my suspicion: the widespread support for Everlyn by influencers (KOLs) is the result of paid promotion. The conclusion is clear: Everlyn AI's enormous market buzz stems not from the organic appeal of its technology or the spontaneous enthusiasm of its community, but rather from a costly, carefully orchestrated marketing campaign through specialized platforms like Kaito. Directly linking posting ("yapping") with rewards (points, airdrop rewards) creates a clear financial incentive for influencers (KOLs) to promote the project without conducting in-depth research on its quality. High "hot indexes" on platforms like RootData are the result of this artificial campaign, not a true reflection of the project's fundamental progress. Conclusion: In summary, I personally believe that Everlyn AI, a recently popular project touted by many influencers (KOLs), is highly suspicious and warrants caution. The founders leveraged their AI reputation to attract capital and partnered with crypto marketing teams to issue tokens. The AI component may be genuine, but whether its performance lives up to the hype remains difficult to verify. However, based on my observations, the Web3 component is a complete phantom project, boasting only narratives but no code, yet rushing to issue tokens, betraying the unsightly desire to profit from the situation. The Everlyn AI case highlights a cautionary tale in the cryptocurrency space: leveraging legitimate technical teams and top-tier investor endorsements to provide credibility for token projects lacking substance. This "decentralization drama" is more deceptive than a pure scam because some elements of it (the AI research) are genuine, but the investment vehicles (the tokens and the so-called protocols) are built on a fabricated foundation. The core lesson for investors is: Don't neglect scrutiny of a project's infrastructure simply because the team is impressive or KOLs are touting it. In the Web3 space, the absence of fundamental elements like technical white papers, testnets, and open-source code is a more significant red flag than any marketing hype. This article is based on analysis of publicly available information and does not constitute investment advice. There are significant risks in cryptocurrency investment, please make your decision with caution.
JinseFinance
decrypt
Beincrypto
Cointelegraph
Bitcoinist