Everlyn AI Depth Investigation: Using AI as a Banner to Create Web3's Tiger Skin

During the National Day holiday, I saw several KOLs in the Crypto Assets field praising Everlyn AI on Twitter— a “video generation” project similar to Sora, which piqued my curiosity and prompted me to do some research.

The first impression is very stunning: an AI project led by a former Meta research scientist, claiming to create a groundbreaking open-source AI video generation model and build a decentralized Web3 protocol. Its Crypto Assets narrative combines AI, intellectual property (similar to Story Protocol), DePIN (io.net is also one of its investors), and creator economy among several popular tracks. When a project gathers so many hot spots and has a number of well-known endorsements, it is indeed impressive.

However, after a thorough investigation, I developed deep doubts about this project. Perhaps my sources of information are limited, but I cannot find any substantial trace of the project in the Web3 world. In simple terms, the project lacks publicly verifiable protocol addresses and block explorers to observe its on-chain activities. Aside from the fact that its token issuance is real, the rest of the Web3 elements are as elusive as “air”.

This article will provide a comprehensive analysis of Everlyn AI based on publicly available information, helping readers understand the true nature of this project.

Overpackaged AI and the Illusory Promises of Blockchain

The core proposition of the project

The core selling points of Everlyn AI's external promotion include:

AI field

  1. Everlyn-1 Video Generation Model: Claimed to be the “first open-source autoregressive video model”, it generates videos frame by frame in a manner similar to GPT's “next token prediction”. This contrasts with the diffusion models commonly used by competitors, which have slower generation speeds. Academic research has confirmed that autoregressive (AR) models have potential in decoding speed and generating long sequences, but they are prone to error accumulation; on the other hand, diffusion models excel in generation quality but come with high computational costs. Everlyn chooses the autoregressive route, aiming for breakthroughs in speed and length, and claims it can generate a 1080p resolution, 8-second long video in 4 seconds while reducing costs by 10 times.
  2. Open Source Advocacy: A core differentiating feature of Everlyn AI is its “open source” nature, committing to provide “completely open source model weights,” which directly challenges the industry status quo of closed-source models like OpenAI's Sora.

Web3 Field

Regarding Everlyn's innovative highlights in AI, the existing descriptions are relatively uniform. However, the descriptions of its Web3 features are varied, with some content not even found in official documentation. The author summarizes its Web3 narrative mainly includes the following points:

  1. With the rampant spread of deepfake technology, the demand for verifying the authenticity of digital content has become more urgent than ever. Everlyn claims to be building a dedicated Layer 1 public chain that provides tamper-proof provenance proof and copyright verification for AI-generated content by recording timestamps and creator information on-chain. As a result, videos created through Everlyn will leave a public and transparent record on the blockchain, making it easy to identify and trace if they are used to create deepfake content. This narrative is similar in logic to IP platforms like Story Protocol and NFTs.
  2. Everlyn has built a DePIN platform where users can contribute their GPUs to provide computing power for “video generation” and earn rewards. This narrative may be related to io.net's involvement in investment, thereby triggering the market's association with its DePIN narrative.
  3. In addition, it also includes the more common narrative of the creator economy, where creators can earn token incentives by publishing videos on the Everlyn platform.

These claims precisely hit the pain points of the current market: the high cost of AI video generation, the monopoly of closed-source models, and the difficulty in verifying content authenticity. However, there is a huge gap between promotion and reality.

AI part: real but over-packaged

Despite the strong background of the team, there are obvious contradictions in their technical promotion.

  1. Terminology Confusion: The project claims to use a “self-regressive model” architecture, but mentions “xDiT (Distributed Diffusion Transformer)” in its promotional materials. The diffusion transformer is a core component of diffusion models and represents a different technical approach than self-regressive models. This confusion of fundamental concepts is concerning.
  2. Current Status of the Codebase: Although there is the Everlyn-Labs organization on GitHub, its codebase mainly contains the relevant code for the academic research projects published by team members (such as ANTRP, Wasserstein-VQ), and is not a unified, production-level video generation system. Furthermore, there is no blockchain-related code in the team's GitHub codebase.
  3. Performance validation missing: The project claims to be “the fastest video generator on Earth,” but lacks independent third-party benchmark verification. In the open-source community, models such as CogVideoX and CausVid have gained wide recognition, while the actual competitiveness of Everlyn-1 has yet to be confirmed.

Web3 Section: Nonexistent “Ghost”

This is the most serious problem of the entire project. The project claims to build a Decentralized Video AI Layer, however, after a thorough search, the author was unable to find any relevant Web3 technology white papers, developer documentation, public testnets, block explorers available to observe on-chain activities, or any code repositories related to Web3 technology?.

In summary, the Web3 part of Everlyn, except that the issuance of the coin is real (its token is on BSC, with the address 0x302DFaF2CDbE51a18d97186A7384e87CF599877D), everything else is only at the conceptual narrative level, completely lacking technical details.

Its Web3 part, in the absence of code implementation, only has hollow narratives (it is worth mentioning that Everlyn's official Web3 narrative is relatively restrained, and many eye-catching statements come from the excessive interpretation of certain KOLs), but is eager to issue tokens. This behavioral pattern gives the author a familiar scent of “cutting leeks”.

The Truth Behind Celebrity Endorsements

The credibility of a project largely depends on its founding team. The author has 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 have indeed worked at Meta and have made significant contributions in the field of image AI.

In addition to the two founders, all the promotional materials for the project highlight that Turing Award winner and Meta's Chief AI Scientist, Yann LeCun, is an advisor for Everlyn AI, as a key selling point to demonstrate the academic achievements of this project. However, according to my investigation, LeCun's identity as an “advisor” is quite dubious.

The only official source currently found to support this claim is a tweet from the Everlyn team, stating, “We’re honoured 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 relevant posts on his X account, and found no information mentioning Everlyn AI or his advisory role. Moreover, LeCun's posts and interviews mainly discuss general AI topics, with little reference to Web3, and no mention of the Everlyn project.

Therefore, the author believes that this time involving LeCun is a typical example of “celebrity effect” marketing - creating an illusion of a formal endorsement relationship by amplifying a trivial interaction.

The Players Behind: Investment Logic and Marketing Machine

Strategic Investment of Mysten Labs

Everlyn AI has secured $15 million in funding led by the Sui development team Mysten Labs, with a valuation of $250 million. However, it is noteworthy that neither the official blog of Mysten Labs nor the channels of the Sui Foundation have released an official statement regarding this investment.

This “silence” suggests to the author that the investment in Mysten Labs is at best a strategic talent acquisition—acquiring a top-notch AI team and its technology to enrich the Sui ecosystem, rather than endorsing the project's tokenomics or decentralized roadmap.

Interpreting the Real Signals of “Binance Listing”

Being able to go live on Binance is also a highlight of the project's promotion, but we need to accurately understand the real signal behind it:

  • The LYN token is launched on the “Binance Alpha” platform, not the main spot trading area of Binance.
  • Binance Alpha is a high-risk sandbox specifically for trading “emerging digital assets not listed on the Binance exchange.”
  • Binance clearly warns: Being labeled as an Alpha asset does not mean it will be listed on the main site in the future. The asset faces higher volatility and risk, potentially losing the entire investment, and withdrawals may not be possible.

Therefore, the launch of the Alpha platform on Binance is not a quality endorsement, but rather a trading experiment conducted under the premise of isolating risks.

Kaito-driven Marketing Machine

The amazing popularity of Everlyn AI is no coincidence; behind it is a systematic marketing promotion machine.

  • Role of Kaito: Everlyn AI utilized Kaito's Capital Launchpad platform in its $2 million public fundraising. Kaito describes itself as an “AI-driven vertical search engine,” and its launchpad allocates quotas based on indicators such as users' social reputation. It is noteworthy that Everlyn AI's airdrop task explicitly requires users to post in Kaito's “yaps” section.
  • Paid Promotion Mechanism: Kaito's business model includes the “Yap-to-Earn” competition, where project parties establish a reward pool to “incentivize creators to produce content about the company.” This directly confirms the author's suspicion: the extensive support from numerous KOLs for Everlyn is a result of paid promotion.

The conclusion is obvious: the massive market volume of Everlyn AI does not stem from the organic appeal of its technology or the spontaneous enthusiasm of the community, but rather from a costly marketing campaign meticulously orchestrated through professional platforms like Kaito. By directly linking posting (“yapping”) with rewards (points, airdrop weights), it has created clear economic incentives for KOLs to promote the project without the need for in-depth verification of its quality. The high “heat index” on platforms like RootData is a result of this artificially manufactured activity, rather than a true reflection of the project's fundamental progress.

Conclusion

In summary, the author personally believes that Everlyn AI, a recently popular star project that has been hyped by many KOLs, is highly suspicious and deserves caution. The founder attracts capital by leveraging their AI reputation and collaborates with crypto marketing teams to issue tokens. The AI part may be genuine, but it is still difficult to verify whether its performance is as excellent as advertised. However, based on the author's observations, its Web3 component is entirely an “air project” or a “ghost project”, with only narratives and no code, yet eager to issue tokens, presenting a rather ugly image of being in a hurry to “cut leeks.”

The Everlyn AI case reveals a concerning phenomenon in the Crypto Assets field: using legitimate technical teams and endorsements from top investors to provide a veneer of credibility for token projects that lack substance. This “decentralized drama” is more deceptive than pure scams because some parts of it (AI research) are real, but the investment vehicles (tokens and so-called protocols) are built on fictional foundations.

For investors, the core lesson is: do not relax scrutiny on project infrastructure just because the team is excellent and KOLs are promoting it. In the Web3 space, the absence of fundamental elements such as technical whitepapers, testnets, and open-source code is a more significant danger signal than any marketing hype.

This article is based on publicly available information and does not constitute investment advice. Investing in Crypto Assets carries significant risks; please make decisions cautiously.

View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
0/400
No comments
Trade Crypto Anywhere Anytime
qrCode
Scan to download Gate App
Community
English
  • 简体中文
  • English
  • Tiếng Việt
  • 繁體中文
  • Español
  • Русский
  • Français (Afrique)
  • Português (Portugal)
  • Bahasa Indonesia
  • 日本語
  • بالعربية
  • Українська
  • Português (Brasil)