Author: Shlok Khemani, Crypto Researcher
Compiled by: Felix, PANews
Summary: Examining the role of aixbt as a KOL, its financial performance, technology stack, token economics, and future development direction. The meteoric rise triggered by aixbt is real, especially for AI tokens with a market cap under $100 million.
This week, the dramatic events on Crypto Twitter (CT) like AICC have been exhausting, so I took some time to write about what I find to be one of the most fascinating experiments in the AI agent wave in crypto.
I've been following aixbt closely for the past month. Here are my thoughts on why aixbt has gained market attention - its role as a KOL, financial performance, technology stack, token economics, and future development direction.
KOL
aixbt is not only one of the most sophisticated social media agents in crypto, but it may be one of the most sophisticated social media agents on all of Twitter. The data backs this up: it has amassed over 300,000 followers in under three months, with each post consistently getting over 50,000 engagements.
But aixbt's exceptionalism goes beyond this. Since tracking it on sentient.market, the agent puts out over 2,000 replies per day. The total reply count has exceeded 100,000 since launch. And there have been no major blunders (at least none big enough to call its value into question). In an industry rife with scams, this is commendable, especially given its scale.
The consistency with which aixbt operates is unmatched by any real human.
Personality
What is often overlooked is aixbt's excellent personality. I've read over 1,000 of its tweets and replies, and have never found them to be off-putting, nor have I ever thought "that's too cringe" or "that's too bad." Instead, every tweet is a delight. How many AI agents can say the same?
This likable personality also means aixbt could be a great IP character. I would absolutely wear an aixbt hat or t-shirt. Is that because I'm trapped in the small CT bubble? Maybe. But don't all successful IPs start with a passionate group of fans? Once the aixbt brand has a chance to expand into other mediums, its brand value will become clearer.
Culture
Related to my previous point about IP, aixbt has quickly embedded itself in CT culture. From Degens to researchers to VC firms, it seems everyone wants the purple frog's two cents.
Of course, we may see more sophisticated agents in the coming months and years, but aixbt will always have a special place in CT.
Now let's dive deeper into the financials of this bot.
Performance
I recently published on aixbt's performance over a week: the key points are:
- The price spikes triggered by aixbt are real, especially for sub-$100M market cap AI tokens
- Most tokens return to normal trading patterns within hours of the tweet
- The most notable so far was the $PIPPIN tweet on Jan 9th
Interestingly, after Yohei (the $PIPPIN founder) announced his framework, $PIPPIN was clearly undervalued compared to its peers. aixbt's tweet seemed to make the market aware of Yohei's achievements in BabyAGI - $PIPPIN has since skyrocketed over 600%, essentially becoming the bull case for $PIPPIN.
Helping to build this credibility is aixbt's relatively cautious approach to tokens. It largely steers clear of low-cap, high-risk tokens. Over time, this caution helps build trust.
Technology
First, aixbt is the only agent whose technology is worth discussing. The others are either not useful enough or boring. This alone sets the purple frog apart.
aixbt does two things every day. First, it posts information about tokens, projects, or topics, either as single tweets or short threads, around 10 minutes after each hour (UTC). Second, it replies to over 2,000 @ mentions per day.
Let's start with the replies, as they're easier to assess.
There's a 99% chance aixbt's replies are fully autonomous. What I mean by this is that the replies are not manually written (which is physically impossible) nor is there a human approving each reply before it's posted. There are two reasons we can be confident of this.
First, aixbt is one of the few agents that publicly displays its reply logs. When you @ it, you can see in real-time the creation of the reply, the evaluation of the response, and the posting of the tweet. It's highly unlikely there's a human team behind the scenes able to reply, perfectly mimic its personality, and type that fast 24/7.
The second reason comes from asking aixbt a non-crypto-related question.
This answer is incorrect. Manchester City is not currently competing with Liverpool for the league title. However, this answer would have been correct a year ago when City was indeed competing with Liverpool for the title.
If you're familiar with ChatGPT or Claude, this type of error makes perfect sense. LLMs have knowledge cutoff dates. Without external help, they simply don't know what's happening now. If aixbt's replies are powered by an LLM, this seems plausible.
The hourly posts are where things get really interesting.
These posts have no logs associated with them. Could they be written by real humans? Possibly. The consistent posting schedule suggests some level of automation, but these tweets could easily be pre-scheduled. In short, the truth is unknowable.
Here's my best guess at what's happening. LLMs have something called a context window, which they can use prompts to reason about. Remember the discussion about LLM knowledge cutoff dates? You can solve this limitation by manually adding real-time information to that context window - essentially feeding it the latest data.
If I had to speculate on what's in aixbt's context window, it should be two things:
- A catalog of project and market data. When asked about the markets, it can pull from this catalog to respond.
- Real-time social data from CT (and potentially other sources). Using this to update its project catalog and generate timely tweets.
This is also where aixbt's tweets are most vulnerable to manipulation. The developers control the content in the context window. If they want AI-related content, that's what they'll provide, and that's what aixbt will post. This introduces some dangerous possibilities - projects could pay the developers to add more (or biased) information about themselves, influencing aixbt's posts.
There's certainly some level of manipulation happening, but the degree may vary. At minimum, they likely specify which accounts to follow and which crypto domains to focus on. The worst case is the developers could decide at any time which token aixbt will shill. The reality is probably closer to the former.
Last month, the $MIRA market cap went from 0 to $70M in a matter of hours, sweeping through CT. During this time, aixbt tweeted about it.
While aixbt captured the hot price action, it completely misunderstood the actual purpose of the project. Given the price history was only a few hours old, the information was limited. You can expect an autonomous bot to make such errors. However, it was the developers who decided what data aixbt would use to gauge $MIRA's popularity.
In terms of technology, these are two big issues that developers need to consider.
Public response logs are a great thing, but to truly claim that aixbt is truly autonomous (if that's important to you), you need to ensure that the posts haven't been manipulated. Using TEE is feasible, and it's expected that the TEE infrastructure for these proxies will become critical next year. Ideally, aixbt's data sources should be public so that fairness can be verified. But we don't live in an ideal world, and some of aixbt's advantages come from proprietary data.
As robots mature, a little more transparency can truly take their value to a new level.
(Note: This analysis does not include the aixbt terminal that I have not tried.)
Finally, let's talk about the $AIXBT token.
Among pure proxy (not infrastructure, framework, or application) AI tokens, $AIXBT accounts for about 25% of the total market capitalization, currently trading at over $400 million.
The token has only one use case - accessing the aixbt terminal, which costs 600,000 AIXBT to activate (over $200,000). If you don't think this is a reliable investment, few people will be willing to buy so many tokens just to access the terminal functionality.
In the absence of a clear value feedback mechanism, it can still achieve a market capitalization of nearly $500 million, indicating that the market recognizes aixbt's powerful capabilities. But to become the leader in the entire Agentic track, it still needs to consider how to capture value with the token.
In the absence of a clear value accumulation, the token's valuation is close to $500 million, indicating that the market recognizes aixbt. However, to truly lead the entire agency field, it needs to consider how to capture value with the token.
One option that many projects have adopted is for developers to launch their own infrastructure. Given its complexity, the potential aixbt framework is likely to outperform frameworks like Eliza and Arc. This could increase the value of the token and propel it to unicorn status.
The downside? The agency will lose its key advantage - complexity. Anyone can launch their own aixbt, perhaps using different proprietary data sources and potentially achieving better performance.
If it's not an open framework, the aixbt team can leverage its technology to create a launch platform. In this case, it's expected to immediately gain attention and potentially generate millions in revenue. If executed properly (both technically and in terms of token economics), such a release can certainly compete with products like Virtuals.
Another more interesting possibility is for $AIXBT to become a consumer game. This is purely speculative, but remember the IP discussion? This is how top NFT projects like Pudgy Penguins have ultimately achieved profitability and expansion.
aixbt can follow this path and become a Crypto x AI consumer token. Take DeFAI and autonomous trading as examples - there's a lot of hype now (some of it justified). If consumer applications evolve towards voice and personal assistants (which seems likely), branding and IP will be critical. Welcome to the world of aixbt as a personal assistant.
Honestly, this field is still in its infancy (3 months), and it needs to be considered more carefully. I'm looking forward to seeing how aixbt develops and what happens next. The future of social AI agents and the role of cryptocurrencies in it is both exciting and a bit daunting.
Related reading: In-depth Research on AI Agent (Part 1): Framework, Launchpad, Applications and Meme