Alexon.eth说：Kaito往左，Ferryboat往右

Kaito veers left, while the Ferryboat turns right.
Kaito往左，Ferryboat往右

说起来有渊源，@_kaitoai 创业的时候正好我们已经做了一个很类似的产品，只不过最后走向了不同的方向。2年前，我们的结论是：当时只有我们和Kaito当时数据是准的。

There is a history between us when @_kaitoai started his venture, coinciding with a similar product we had developed at the time, albeit taking divergent paths eventually. Two years ago, our consensus was: only we and Kaito had accurate data back then.

很多人DM我很多关于Kaito的问题，我算了一下定价，NFT的赔率已经不足够让我们下单（前段时间再次Covid，错过。）所以这是一个与利益无关的post，那既然回复别人也是回复，不如我统一写了发在Twitter上，省的重复碎片的聊好几次，顺便试试Kaito的Yap算法到底权重如何。

由于Kaito最近的付费版我没有用过，也确实利益无关，我主要沿着Kaito和Ferryboat自己做的产品思路上聊起，其中能解决一些：
-【Kaito为什么这么贵，到底跟API有没有关系】
-【Kaito为什么转型】
-【Kaito到底的想象力是什么】
yap噪音里，没有实际传播的问题。

Many have DM’d me with various questions about Kaito. After calculating the pricing, the odds in NFTs were insufficient for us to place an order (missed out due to recent Covid ). Hence, this is a post unrelated to interests.
Since responding directly to each individual would be repetitive, I decided to consolidate my thoughts and share them on Twitter, avoiding fragmented discussions multiple times. This also serves as an opportunity to test the weight of Kaito’s Yap algorithm.

Having not utilized Kaito’s recent paid version, and with no vested interests, my focus lies on discussing the product trajectories of Kaito and Ferryboat. This dialogue can address certain queries within the noise of yap, such as
-“Why is Kaito so expensive, and does it relate to APIs?” -“What spurred Kaito’s transformation?”

• “What is the essence of Kaito’s imagination?”
  issues that haven’t been effectively communicated amid the yap.

????Origin: Why Twitter?
原点：为什么是Twitter？

2022年，正值低谷期，不善社交的情况下，在思考怎么把Twitter的数据量化，找到世界上关于Crypto真实的信息波动，获取”上帝视角“。因为当时基于对流量的认知，我们认为Crypto世界下，流量结构非常单一：

-Twitter&Youtube对应【公域流量】
-Discord&Telegram对应【私域流量】。

In 2022, amidst a downturn and a lack of social skills, the focus shifted to quantifying Twitter data, aiming to capture the true fluctuations of crypto information worldwide and attain a “God’s eye view.” At that time, based on our understanding of traffic, we perceived the traffic structure in the crypto world to be remarkably simple:

Twitter and YouTube corresponded to “public domain traffic”
Discord and Telegram corresponded to “private domain traffic.”

在过往创业的世界观下，公域通常解决曝光和线索，大白话来说通常解决种草问题，对应的是CPM（千次曝光）、CTR（点击率）、CPA（线索）这类数据作为广告投放的衡量指标；私域通常解决信任问题，完成最后的转化率。

当然，这个概念是混合的，比如你Follow了Alexon，在你把我加入了list后，timeline你定期刷这个list，也处于私域的信任状态。这里就不展开了，这是增长和营销的小伙伴们的工作。

在这个结构下，私域流量是无法加权重来量化的，标准化过程也很难，毕竟在第一步要获取所有Discord&Telegram群组进入，就已经宣布死亡。所以，正常智商下都不会优先选择去采集这个部分。

Within the perspective of previous entrepreneurial experiences, the public domain typically addresses exposure and leads – in simple terms, it usually tackles seeding issues.
This corresponds to metrics such as

-CPM (Cost Per Mille)
-CTR (Click-Through Rate)
-CPA (Cost Per Action)

which serve as key indicators for advertising placements. On the other hand, the private domain typically addresses trust issues and final conversion rates.

Of course, this concept is nuanced. For instance, if you follow Alexon and subsequently add me to a list, regularly checking that list on your timeline, you are operating in a state of trust within the private domain. I won’t delve further into this as it falls within the realm of growth and marketing professionals.

In this framework, weighting private domain traffic for quantification is challenging, and standardizing the process is equally difficult. After all, the process of accessing all Discord and Telegram groups is already a daunting task, signaling a challenging journey even before the first step. Hence, under normal circumstances, it is unlikely that one would prioritize harvesting this segment.

公域流量下 Youtube：长视频形态制作时间长，更新频次慢，长尾效应强，长期来看是更优质的具备信任基础的流量，但时效性慢。简单来说，Meme这种高周转的内容天生就不可能在Youtube的场域下形成，它更适合：视频版Podcast、Defi、教程类比如Airdrop、Depin矿机这类需要长展示的内容。 Kaito的逻辑同样适用于Youtube，我们也做了，它在适合观察一些类似Kaspa、Helium类型的标的上更加显著。

Twitter：无疑是时效性最快和集中度最高的平台，这在任何一个行业里都是少见的集中度高。所以它非常适合监测品牌的心智，情绪的波动（没错，我们一直也沿用了”心智“这个消费品的词汇，非常震惊Kaito用的也是mindshare，毕竟我们完全不认识。） Twitter监测Meme和事件型驱动的Altcoin上更加显著。

所以，我看到 @Punk9277的交易背景时候，也一点都不奇怪，应当大家做这个事情的原点都是一样的。

Within public domain traffic, YouTube stands out for its lengthy video content production, slow update frequency, strong long-tail effects, and the establishment of a trust-based, high-quality flow over time, albeit with slow timeliness.
Simply put, content with high turnover rates like memes inherently struggle to thrive within the realm of YouTube. It is better suited for video-based podcasts, DeFi, tutorial content such as airdrops, or demonstrations of products like mining rigs.
Kaito’s logic similarly applies to YouTube; we have also ventured into this territory, where it notably excels in observing targets similar to @KaspaCurrency and @helium_mobile.

Twitter, on the other hand, undoubtedly boasts the fastest timeliness and highest concentration of information, a rarity across industries. Therefore, it is exceptionally well-suited for monitoring brand sentiment and emotional fluctuations. We have consistently adopted the term “mindshare” for consumer goods, which aligns surprisingly well with Kaito’s use of the term, despite our lack of prior acquaintance. Twitter is particularly effective in monitoring memes and event-driven altcoins.

Thus, when I encountered @Punk9277’s trading background, it came as no surprise to me. It appears that the origins for all of us engaged in this activity share common roots.

????Why expensive？
为什么Kaito这么贵？

这其中有两个原因：API和Twitter法规问题。

1）API：由于要商用，Kaito应该也是demo时期用了一些和我们一样省钱的奇技淫巧。后续Musk收购Twitter后，调整了API的使用规则，堵了很多擦边调用API的方法。后续融资后，应当就是直接调用的Twitter API，确实很贵。

2）Twitter的法规问题：即使是商用，我们记得也是每个月有一个调用次数的上限，我就没去确认数字了，有心的yapper可以去看看API文档。 结论：如果大规模便宜的给to c用户，理论上是扛不住的。

所以，在这个限制条件下贩卖一个调用次数的容量，只能to b是经济效益最大的选择。且按照Kaito履历，to b冷启动要比to c简单。应当也是最先卖给了交易机构和基金（因为当时的产品形态下，主要还是以cashtag和hashtag为主，捕捉事件和趋势。）

There are two primary reasons for this:
API constraints and Twitter’s regulatory issues.

-API Constraints:
During the demo phase, Kaito likely employed some cost-saving tricks similar to ours due to commercial usage requirements. Following Musk’s acquisition of Twitter, there were adjustments made to the API usage regulations, closing off many workarounds for borderline API calls. Subsequently, post-funding, they probably switched to direct use of the Twitter API, which indeed comes at a high cost.

-Twitter’s Regulatory Issues:
Even for commercial purposes, there used to be a monthly limit on the number of API calls – I did not verify the exact figure, but interested yappers can refer to the API documentation. Conclusion: It would theoretically be unsustainable to provide large-scale, inexpensive services to “to C” users.

Therefore, under these constraints, selling a volume of API call capacity is most economically advantageous for “to B” users. Considering Kaito’s background, launching to B clients is likely simpler than to C. It is probable that they initially targeted trading firms and funds (as at that time, the product primarily revolved around cashtags and hashtags, capturing events and trends).

????Kaito的产品方向可能
Kaito’s product direction might entail

教人创业的事情我干不来，我只能分享一下我们做的尝试。

-关键词为主：alpha工具、交易策略
我们目前自用的版本就是这个，难点在于洗数据，每家洗数据的思路也不尽相同。由于整个Twitter的垃圾数据非常多，如果单纯采集Crypto的数据，会发现无意义的新闻号和政治号特别多，如果单纯用Engagement来评判，一定是最后食用垃圾食品。故而，只能不停迭代出一个【准确】的趋势，多数情况下大家都给与不同规则的账号赋予不同权重，或者用某个自己设立的list作为原点扩散。

这应该也是前几天中文yap没分的原因之一，的确我们也调低了中文的权重，因为当时我们实测的结果是英文CT的传播需要24-48h才能传导到中文区。如果以信息链来看，大部分时间下确实是英文内容主导叙事，所以相应的，铭文这个东西我们就不会有什么参与感，因为监测不到。

我们也尝试过用量化的指标做策略去测试，时测下来如果以Binance已经list过的top100来看，不足够显著，而小市值和meme则需要强人工的干预，有效但不足够标准。
这个我认为也是
@Punk9277
在
@Mable_Jiang
的Podcast上说的想做一个通用产品的潜台词。我相信Kaito早期应该也试过这个方向，得出了一个类似的结论后，发现继续往这个方向走的话，越走越不可能通用。毕竟如果能直接赚钱，谁还去研究产品化呢。

Kaito’s product direction might entail
Teaching entrepreneurship is not my forte; I can only share some of our attempts.

-Primarily Keyword-Oriented: alpha tools, trading strategies

Our current in-house version revolves around this. The challenge lies in data cleansing, with each entity having unique approaches. Given the abundance of spam data on Twitter, focusing solely on crypto data reveals a plethora of irrelevant news and political accounts. Relying solely on engagement metrics often leads to consuming junk information. Therefore, we continuously iterate to discern an accurate trend. In most cases, different rules are applied to assign varied weights to accounts, or a self-curated list serves as the starting point for dissemination.

This might be one of the reasons why the recent Chinese discussions were not segmented correctly. Indeed, we reduced the weighting for Chinese content since our tests indicated that it takes 24-48 hours for English content to disseminate into the Chinese sphere. From an information flow perspective, English content predominantly shapes narratives, rendering us less involved in Chinese discussions due to our inability to monitor them.

We’ve also experimented with using quantitative metrics for testing strategies. Results show that focusing on the top 100 assets listed on Binance lacks significant impact. Small-cap assets and memes necessitate strong manual intervention, effective but lacking standardization.

I believe this aligns with
@Punk9277
‘s implied desire to create a universal product on
@Mable_Jiang
‘s podcast. Early on, Kaito likely explored this avenue and arrived at a similar conclusion – further progression in this direction leads to decreasing universality. After all, if profitability is readily achievable, who would delve into product development?

????垂直GPT类、新闻类：

由于商用原因，Kaito的付费版理论上是采集了推特原文但是不能公开的。

我们由于不商用所以是自己查看原文上会方便一些。这些原文的语料显然是可以给ChatGPT再喂一遍的，毕竟实时数据OpenAI是不开放的，所以我们会看到在这条路的探索上，
@Galxe
做了
@AlvaApp
， 但是alva的数据基本都是点对点合作了类似rootdata这类之后，写了prompt后展示出来，他们的小伙伴
@LHuang19342
送了我一年的VIP，最近我用它查询一些Crypto百科问题还是比较方便的。

但alva也有自己的问题，应该是洗数据的问题，但依旧提高了效率。这个东西我们2年前也做过，我身边的朋友还体验过，但是到了付费环节测试下来意愿不强，毛利算下来一般，我们就停止了。

查询一些社区新闻，比如“ eliza和Eliza之争到底是什么？”这类问题上，拥有Twitter数据的人总能更快让AI来告诉你。Kaito应该也做了类似的产品。

这个类型的产品的问题是底层都调了GPT的接口去读一遍语料，所以收费的部分基本都贡献给了OpenAI，如果能自己建个模型的话才有盈利的空间，但是投入比较大。

其实这个方向上，我们一直认为
@realMaskNetwork
是更适合的。多说一句， Mask我们没有交流过，但是他们的Github显示非常努力，每次有新玩意都加班加点开发。比如半年前的Solana的Blink刚推出，我们全网在看谁是最快能拿出来能用的版本，Mask是公开资料里我们能找到里最快的，当时我们都准备建仓了，但是不知道为什么好像没发？看起来并没有给我一个shill mask的机会。
@suji_yan

????Vertical GPT and News Categories:

Due to commercial constraints, Kaito’s premium version theoretically gathers original Twitter content that cannot be publicly disclosed.

As we do not operate commercially, it is more convenient for us to view the original content. Clearly, this corpus of original text can be fed back into ChatGPT, given that real-time OpenAI data is not accessible. Thus, in exploring this path,
@Galxe
developed
@AlvaApp
. Alva’s data primarily involves peer-to-peer collaborations akin to rootdata, where prompts are written and displayed.
Their associate
@LHuang19342
graciously granted me a year of VIP access, which I recently used to conveniently explore some crypto-related encyclopedia queries.
However, Alva also faces its own challenges, likely related to data cleansing, yet it has undoubtedly enhanced efficiency.
We ventured into a similar endeavor two years ago; friends in my circle even experimented with it. However, during testing at the premium stage, enthusiasm waned, and profit margins seemed average, prompting us to halt operations.

For sourcing community news, such as inquiries like
-“What is the dispute between
@elizawakesup
and
@elizaCommunity
all about?” individuals with access to Twitter data can promptly employ AI to provide answers. Kaito likely delved into developing a similar product.

The challenge with this type of product lies in utilizing GPT interfaces to read through the corpus. Consequently, the revenue generated essentially contributes to OpenAI. Profit potential exists if one can establish their own model, albeit requiring significant investment.

In this regard, we have always believed that
@realMaskNetwork
is better suited. It is worth mentioning that we have not directly communicated with Mask, but their GitHub activity demonstrates remarkable diligence, with overtime work put into each new development.
For instance, when Solana’s Blink was introduced six months ago, the entire community was eager to see who could swiftly present a usable version.
Mask, as per publicly available information, was among the fastest. We were on the brink of establishing positions at that time, yet for some reason, it seems like it didn’t materialize? It appears they didn’t provide me with an opportunity to promote Mask.
@suji_yan

????KOL的广告产品
KOL advertising products

起因都是因为本身就给过了权重，所以做这个方向是相对比较顺畅的。无论是分时数据还是读关注list都可以给KOL本身的【正确性】和是否属于【核心圈层】（Kaito我看是取了个Smart Follower的名字）

-正确性：
kaito目前应该还没投入人力做，否则fud的kaito的应该不会给分。由于拥有分时数据其实可以用时间戳反复加权提升事件正确的KOL的权重，也可以相应阅读正向和负向的情绪。所以这个部分可以反复筛选KOL。

-核心圈层：
这个我想替发中文内容的KOL说句话，这个其实不是中文KOL的错。底层原因是因为中文区看得懂英文区，英文区看不懂中文区。叠加这两年中心确实英文主导叙事，所以确实容易被降权。但是上帝视角来看，英文区的垃圾话KOL同样很多，无论什么语言，垃圾信息一定是大于有效信息的。

所以建议Kaito考虑一下，找几个你们认为有价值的KOL，比如
@hotpot_dao
，重新读一遍中文list，给中文重新打个权重，在TGE这个时候团结一下更多的人。

????KOL advertising products

The root cause lies in the inherent weighting given initially, making this direction relatively smooth to pursue. Whether it’s real-time data or reading through follower lists, both can assess the KOL’s own “accuracy” and whether they belong to the “core circle” (Kaito appears to have adopted the name “Smart Follower”).

Accuracy:
Kaito likely hasn’t allocated resources to this yet; otherwise, they wouldn’t receive negative ratings. Leveraging real-time data allows for repeated timestamp weighting to enhance the accuracy of events associated with KOLs, enabling the interpretation of positive and negative sentiments. This iterative process aids in filtering out KOLs effectively.

Core Circle:
Speaking on behalf of Chinese content creators, this isn’t solely the fault of Chinese KOLs.

Fundamentally, the issue arises because the Chinese community understands English content, while the reverse isn’t always true. Coupled with the prevailing English narrative dominance in recent years, it’s easier for Chinese voices to be marginalized.

However, from a macro perspective, there are also numerous subpar English-speaking KOLs; regardless of language, junk information tends to outweigh valuable insights.

Hence, I recommend Kaito to consider selecting a few KOLs deemed valuable, such as
@hotpot_dao
, revisiting Chinese lists, reassessing their importance, and fostering unity among more individuals during this TGE phase.

为什么Kaito要转KOL数据？以及为什么我认为是对的？

Kaito已经产品化做的很好了，目前这个yap策略也是很聪明的。主要聪明的点有以下两个：

-战术上：
免费利用KOL的杠杆抢注意力宣发 从我们的视角下，Crypto的流量不存在【信息流投放】逻辑，只存在【分销】和【裂变】逻辑。

过往两年，所有在Twitter上拿到流量的基本都是这两个形式，今天的yap其实就是升级版的Friendtech和tips，只不过当年这些项目不具备筛选有效KOL的手段，所以误打误撞造成了色情博主也能获取空投。

而Kaito识别出了好内容+好KOL，利用Airdrop节省了大笔营销费用，也给了大部分Smart账号一个理由去shill。等于是直接分销到了所有KOL而不需要Agency去投放，加上个landingpage去做社交炫耀。 （这个裂变的部分我认为还可以再改进，我认为Kaito的裂变率不足够高。）

Why did Kaito pivot towards KOL data? And why do I believe it’s the right move?

Kaito has already excelled in productization, and the current yap strategy is quite clever. The main points of cleverness are as follows:

-Tactically:
Leveraging KOL influence for attention-grabbing promotions.

From our perspective, the flow of traffic in the crypto space doesn’t adhere to the logic of “information flow advertising” but rather operates on the principles of “distribution” and “viral propagation.”

Over the past two years, those gaining traction on Twitter have largely followed these two patterns. Today’s yap is essentially an upgraded version of Friendtech and tips, except that these projects in the past lacked the means to filter out effective KOLs, inadvertently leading to instances where even adult content creators could receive airdrops.

Kaito has identified quality content and influential KOLs, utilizing airdrops to save significant marketing costs and providing most Smart accounts with a reason to promote.

This approach essentially distributes to all KOLs without the need for agencies to advertise, complemented by a landing page for social showcasing. (I believe there is room for further improvement in the viral propagation aspect; Kaito’s viral rate, in my opinion, is not yet high enough.)

2）战略上：
通过KOL的绑定账号，脱离Twitter API的限制
实际上，这个才是重点。

我相信Kaito往下走要面临的终极问题是：如果Twitter修改规则，那么如果保障产品的运行。 朝令夕改在如今Musk主政的Twitter面前并不是危言耸听，那么接下来到二级市场要接受所有用户的的”市梦率“考核，这个问题是绕不过的。

利用TGE的最大热度绑定最多的KOL形成SBT，因为是主动授权，调用起来就以后不再有障碍。这个举动目前看起来是挺成功的，形成一个网络效应，汇聚越来越多的KOL和to B的客户。这样可以保证核心资产是Supply和Demand两端的数据实际上主动权从Twitter转移到了Kaito。

-Strategically:
By linking accounts with KOLs, breaking free from Twitter API limitations

In reality, this is the crux of the matter.

I believe the ultimate challenge that Kaito will face moving forward is: How to ensure the product’s operation if Twitter changes its rules？

Given the unpredictable nature of Twitter under Musk’s leadership, sudden rule changes are not unfounded. Consequently, transitioning to the secondary market will entail assessing the “market dream rate” of all users, a challenge that cannot be circumvented.

Leveraging the peak hype of TGE to bind the maximum number of KOLs to form a Strategic Binding Team (SBT) is crucial. Because this is done through proactive authorization, future utilization should encounter no barriers.

This move appears to be quite successful at present, creating a network effect that attracts more and more KOLs and B2B clients. This approach ensures that the core assets, which are the data at both the Supply and Demand ends, effectively shift the decision-making power from Twitter to Kaito.

结束之前，说个有意思的八卦

作为一个普通的创业者也是苦笑不得。两年多前，我们第一次尝试用这个东西去做产品，完成度还不错，朋友甚至还介绍了四家VC看这个产品，具体是哪几家我就不@了。因为不是典型精英背景，除了一家愿意跟投，剩下三家跟我打concall的人都没当回事。我至今不知道他们boss知不知道有过这个concall，后来认识之后也没问过。这就是为啥我们最后也没产品化，走向了不用和人打交道的纯自用alpha工具的道路，也是为什么每次我们关注一个东西都比较早的原因。

更有趣的是，其中有个打过concall的员工，我昨天在timeline上刷到目前在吹Kaito牛逼赚Yap。希望如果看到的话，以后你多看产品少问Connection。

Before we conclude, here’s an interesting anecdote:

As an ordinary entrepreneur, it’s often a bitter amusement.

Over two years ago, we made our first attempt to develop a product using this technology, and it was fairly well-received.

Friends even introduced the product to four venture capitalists for consideration—though I won’t mention which ones. Due to not having a typical elite background, aside from one firm willing to co-invest, the other three VCs I had conference calls with seemed indifferent.

To this day, I’m unsure if their bosses were even aware of these calls, and after we eventually met in person, I never inquired. This is why we ultimately didn’t commercialize the product, instead opting for a path that doesn’t involve engaging with others—a route focused on developing alpha tools solely for personal use. It’s also the reason why we tend to be early adopters when it comes to new interests.

What’s even more intriguing is that one of the employees who participated in those conference calls was seen boasting about Kaito’s success with yap on their timeline yesterday. If you happen to come across this, remember to focus more on the product and less on connections in the future.

这可能就是被迫i人只能出来扯淡的原因，Yap一下获取点行业影响力，给下一次创业做点准备。这一年多来，只要产品做的好，无论有没有知名VC支持，找过我们的，我们都在有限范围内给予了最大的善意。做个好人，行业就善意一点。

正好昨晚睡前 @Atlantropaz来找我聊Kaito，说他们和kaito也有合作，我看了一下 @noise_xyz 好像是利用了kaito的数据预测什么？我老婆说学弟的项目要好好看，但是我什么都看不到。最近在思考和一个生态深度合作 ，目前Taste最接近的也是Kaito的客户 @megaeth_labs。

感谢大家这一年多来的支持。

祝福 @_kaitoai 成功，但无论你是不是Smart Follower，我都乐意互动（有效信息的情况下）。

This may be why introverted individuals are sometimes compelled to engage in casual conversation—just to gain a bit of industry influence with a yap, preparing for the next entrepreneurial venture.

Over the past year, as long as the product is well-executed, regardless of whether it has the backing of prominent VCs, those who have approached us have been met with the utmost goodwill within our limited scope. Be a good person, and the industry tends to reciprocate with kindness.

Coincidentally, last night before bed, @Atlantropazreached out to discuss Kaito, mentioning their collaboration with Kaito. Upon looking into @noise_xyz, it seems they are utilizing Kaito’s data for predictions.

My wife mentioned that we should closely examine the project of a younger colleague, yet I am unable to focus on anything. Lately, I have been contemplating a deep collaboration within an ecosystem, and currently, Taste’s closest client is also a customer of Kaito,
@megaeth_labs
.

Thank you all for your support over this past year.

Wishing success to @_kaitoai, I am open to interacting with you, whether you are a Smart Follower or not, as long as there is valuable information to share.