We're excited to introduce a completely redesigned user interface for askgina.ai, focused on providing a cleaner, more adaptable, and intuitive experience across all your devices. This update reflects our commitment to making your interactions with on-chain information and activities as seamless as possible.

Inspired by a Shift in Thinking: Beyond Pure Conversation

Our approach to this redesign was partly influenced by evolving perspectives on human-computer interaction. An insightful article by Julian.digital argues against the limitations of purely conversational interfaces, suggesting that while natural language is powerful, it's not always the most efficient method for every task. Just as we use gestures and visual cues in everyday communication, user interfaces can benefit from offering more direct ways to interact. This concept inspired us to think beyond purely conversational commands and explore how we could augment the askgina.ai experience with more direct action triggers, leading to the introduction of suggested actions and our vision for future wallet action buttons.

Wallet First: A Dedicated Space for Your Assets

One of the most significant changes is the redesign of the wallet view, which now has its own dedicated space. Previously, the wallet was a collapsible section, somewhat hidden from immediate view. We recognized the importance of having your on-chain assets readily accessible, so we've brought the wallet front and center. This dedicated space provides a more consistent and intuitive experience across all devices, with an improved visual presentation that feels more like a dedicated crypto wallet.

Looking ahead, we're planning to integrate more manual use cases directly into the wallet, such as bridging and swapping, making askgina.ai a more comprehensive platform for managing your on-chain life.

Optimized for Every Screen: A Responsive Design for Mobile, Tablet, and Desktop

The new askgina.ai is designed to provide an optimal experience no matter what device you're using. The previous version presented a fixed layout across all devices. Now, the UI adapts seamlessly to your screen size, ensuring a user-friendly experience whether you're on a desktop, tablet, or mobile phone.

On mobile, you'll find a dedicated wallet tab at the bottom of the screen, providing instant access to your full portfolio in a clean, full-screen view.

For tablets and foldable phones, we've introduced a layout that keeps your wallet readily accessible. You also have the option to collapse the chat history to focus entirely on your conversation with Gina.

On desktop, the redesigned interface offers a spacious and organized layout, ensuring all key features are easily within reach.

This responsive design ensures that you can comfortably interact with AskGina.ai on any device, with tailored layouts that maximize usability on different screen sizes.

Smarter Chats: Contextual Suggestions at Your Fingertips

We've also enhanced the chat window to make your conversations with Gina more efficient. You'll now find quick suggested follow-up questions and actions that are tailored to the specific context of your conversation. For example, if you inquire about a particular cryptocurrency, you might see suggestions for related questions or actions, saving you time and effort in typing out your next query.

Why the Change? Focusing on User Needs

This redesign was driven by a desire to create a more user-centric experience. We recognized that having a readily accessible and visually clear wallet was key. Furthermore, the need for a responsive design that adapts to different screen sizes was crucial for ensuring a smooth and enjoyable experience across all devices.

By prioritizing these aspects, along with incorporating insights on the evolution of user interfaces, we believe the new askgina.ai provides a significantly improved platform for navigating your on-chain life.

Try out the app today

Give askgina.ai a whirl and explore the app for yourself. It's now open for you to use right away if you sign in with your Farcaster, X, or email. We're confident that you'll find the redesigned interface to be more intuitive, efficient, and enjoyable to use.

Feel free to shoot us a message - can DM @askgina.eth or one of the Gina squad: @sidshekhar, @ericjuta, @cesar on Farcaster or X) with feedback + thoughts!

Running into the limits of what a single agent can do

In our previous blogpost we compared different foundational LLM models with a set of varied prompts in and saw how they performed noting pros/cons of each.

We’ve found that as we keep adding capabilities to Gina, we were start to hit the limits of these foundational models in terms of performance.

Why is this?

• Adding more tools (functionalities) and prompt rules takes up more token context (models have limits on how many tokens (characters) they can take in as input and output)

• Tool calling performance for each foundational model differs substantially

• The "advertised" token context window is not always indicative in terms of accuracy and relevance of responses once you get to higher levels of tokens consumed (usually <100K tokens)

In the world of crypto AI frameworks today, ElizaOS and ARC primarily build on a tool-centric architecture. Key differences include:

• ElizaOS focuses on broad integration, assembling a rich ecosystem of plugins that interface with variousblockchain protocols.

• ARC, on the other hand, prioritises performance with a lean, efficient engine that optimises LLM execution, ensuring low-latency decisions and robust handling of multi-step tasks.

• Both frameworks, while powerful, rely on a single foundational agent structure where tool calling and context management become increasingly challenging as more tools are added.

We're diverging from this approach by

a) Moving to a "swarm" based approach of multiple agents working with each other - each focussed on a given role and comprising of several tools related to that role (i.e a "transaction execution" agent)

b) Embedding a central wallet with execution capabilities that can be shared by all the sub agents.

What are some of the different design patterns for building swarm based agents that can work well together to deliver a desired outcome?

Being in the arena, we explored, implemented and evaluated the following different design patterns:

Works well for structured workflows, but tends to overcomplicate on simple requests

Fastest to pass tasks between agents but requires real-time coordination and is generally harder to scale

A Smart Router chooses the best routing method dynamically. This is the approach we ended up going with.

• Challenges to Consider:

  • Model Selection: How do we choose the best model for each subagent’s specialised task?

  • Tool Allocation: Which tools should be assigned to which subagents, and how do we manage overlaps?

  • Prompt Logic: How do we design system prompts that effectively manage handovers without becoming unwieldy?

  • Handover Dynamics: What strategies can mitigate latency, especially if users require multiple, sequential interactions?

The above are some key technical considerations we had when seeing how to build scalable agentic workflows today. What do we mean in practice here?

Here are some examples of user complex queries where having a swarm of subagents in the background to break down tasks is truly helpful:

• "Do I have any of the top tokens from this week in my portfolio and if not, get me exposure to them?

• "Where can I generate safe yield for my USDC, ETH, and SOL (single sided and dual sided)?"

• "See this screenshot and get me exposure to all the tokens listed here"

As foundational models improve, the debate between multi-agent versus single-agent architectures will continue—but for now, multi-agent task delegation is our best bet for efficiency, cost effectiveness, and a seamless user experience.

All of the above and more can be addressed by Gina today - in part due to the technical decisions mentioned in this post.

As a friendly reminder, Gina is now live for early beta - sign up at askgina.ai for early trial access. Feel free to shoot us a message - can DM @askgina.eth or one of the Gina squad: @sidshekhar, @ericjuta, @cesar on Farcaster or X).

Context

Every day, there’s a shiny new large language model (LLM) claiming to outperform the rest. For users, choosing the right model for a specific use case can feel like flying blind in a storm of hype around the latest new thing. The state of things is such that even the model drop down list in the base ChatGPT app now has 7 different models to choose from.

The stakes are even higher when you move beyond mere text generation and want your AI to perform useful tasks - especially on the blockchain. At Ask Gina, we’re building a new kind of wallet interface that has a personal AI agent (Gina) built in. Our users can simply ask, “What’s hot today?” or “Swap 10% of my USDC into AERO on Base.” and in seconds, Gina fetches the right information and handles all aspects of the on-chain transaction.

But going from an LLM to a crypto-savvy agent embedded in a wallet (with the ability to call web APIs, fetch market data, and execute transactions) isn’t trivial. Through the process of building Gina, we tested multiple models, integrated them with various crypto tools, measured performance, and documented our findings. Below, we’ll share some findings from real-world data - and a few key takeaways we’ve learned about picking a model that can do more than just chat.

First a bit of context on “tool-calling”. When large language models are restricted to just generating text, they’re limited in how “agentic” they can be. For an agent to perform actions - like pulling real-time data, booking tickets, or facilitating payments - it needs to leverage tool calling. Think of tools as extensions of the LLM’s capabilities, allowing it to interface with APIs or blockchain smart contracts

A great metaphor we’ve heard in the wild is to think of tools like giving an extra pair of “hands” for an agent to do more things. The more hands we have the more tasks we can handle with them.

Not all models have native support for tool calling (for ex: the Deepseek API doesn’t yet), and even among those that do, performance can vary wildly. This is why it’s crucial to test both:

(1) how well the model recognizes the right tool to call for the job at hand and

(2) the quality and accuracy of the output or transaction it produces.

We ran trials for each model on Gina and their performance on tasks that utilize tool-calling like:

• Fetching market data (using a market data API call),

• Generating charts (using a charting API call), and

• Executing on-chain swaps (using RPCs to to execute transactions calldata).

Data Collection

We built a testing suite that records:

• The task (e.g., “What’s hot?”, “Chart me PEPE”, “Swap 1 USDC into AERO”),

• The model used (e.g., GPT 4o, Claude 3.5 Sonnet, Gemini 2.0 Flash Beta, o3-mini),

• The time taken to produce a final output

• And a rating of accuracy/relevance. This rating is an interpretation of sometimes subjective results for a given task but it is a preliminary indication of where we’re at currently.

Below are the results from an average of multiple trials we ran this past week:

Our testing revealed insights across multiple categories—speed, accuracy, transaction success, tool compatibility, and user experience.

Below, are some takeaways we had from a subjective lens:

1. Speed vs. Completeness

   • Gemini 2.0 Flash often provides the fastest “time to first token (letter or word shown)” and renders visual results quickly, a clear advantage for impatient users. However, due to minimal text exposition, there’s sometimes less clarity about how it chose to solve a task.

   • Claude 3.5 Sonnet can be slower overall but is also more comprehensive - it often makes multiple tool calls, fetches additional context, and tends toward robust explanations. This is an example of tool ordering and handover done well, but at the cost of added latency.

2. Accuracy & Relevance

   • Wrapped tokens remain a recurring confusion for certain models, highlighting how tool optimization with correct prompt scaffolding and correct token metadata selection matter. If a model doesn’t hand over the right query to fetch market data for (e.g., mixing up WETH vs. ETH), it can produce inaccurate results for metrics like TVL, volume etc.

   • GPT 4o tends to be especially thorough in checking user balances before execution—boosting its average success rate on transaction execution in general. This thoroughness reflects better tool ordering (e.g., first check portfolio, then proceed) and is part of why it excels at more complicated transactions.

3. Transaction Success Consistency

   • For simple swaps, nearly all models that properly invoked our "execute transaction" tool succeeded—except Gemini 2.0 Flash, which stumbled when multiple tool calls were needed. This suggests potential gaps in tool handover (switching from quote fetching to swap execution).

   • Claude 3.5 Sonnet occasionally saw transactions revert for multi-step operations (like a swap + bridging). Meanwhile, GPT 4o and o3-mini properly sequenced their calls and completed these tasks, demonstrating stronger tool ordering & handover capabilities and higher average success rates for complex operations.

4. Chaining tools & Observability

   • Many of the tasks in our tests require chaining multiple calls (e.g., figuring out the right asset/token contract, retrieving market data for the right pair, generating a chart, then placing a swap). If the model doesn’t optimize or observe tool outputs properly—for instance, if it overlooks an error or fails to confirm available balances—it can lead to timeouts or failed transactions.

   • Observability is key: a model that clearly displays or logs each call’s outcome (like “Balance check: success”) is easier to debug and improve. Reasoning models like o3-mini and good old Claude 3.5 Sonnet continue to lead in this regard

5. User Experience (UX): Streaming Text & Readability

   • As mentioned above, across models, the time to first token can vary. However, after the first bit of text is rendered, some models stream text slowly, which can frustrate users waiting for an immediate response (GPT 4o). Others might chunk or batch the output, leading to a faster perceived experience (o3-mini).

   • Formatting remains a major differentiator. Models that produce concise bullet points, short summaries, or relevant headings score higher on readability—especially on mobile. This not only improves observability for end-users but can also highlight where errors might have occurred in the tool chain.

If you’re building an AI agent that needs to do more than chat—like interacting with APIs, pulling real-time data, or executing blockchain transactions—choosing the right model can make or break your product. Our experience building Gina to satusfy crypto-specific tasks highlights a few key lessons:

• It’s not always about the “best” LLM in isolation—it’s about the one that plays nicest with your tool stack. Different models for different tools.

• Speed, reliability, and the ability to handle multi-step tasks are just as critical as raw language quality.

• Real-world tests (like bridging assets or fetching legitimate market data) expose nuance that simple text benchmarks often miss.

A practical challenge to consider is simply the cost of running comprehensive model evaluations in production. Running a thorough test suite across multiple models for each code deployment can quickly become expensive, especially when testing complex multi-step operations. This is still an ongoing field of discovery of what is best practice across the AI field in general.

We hope our data and learnings help you make more informed decisions when integrating LLMs into your workflows. As new models emerge, we’ll keep testing and sharing updates. Feel free to shoot us a message - can DM @askgina.eth or one of the Gina squad: @sidshekhar, @ericjuta on Farcaster or X).

And of course, as a friendly reminder, Gina is now live for early beta - sign up at askgina.ai for early trial access.

a) Fetch up to date prices for any onchain token on Base, Solana, Arbitrum, Optimism, Polygon, BSC & Mainnet Ethereum.

b) She can also now generate on-the-fly charts to visualize these!

Simply add - "generate me a chart of X over the last month" (where X is a given token) when you are speaking to Gina and you can see this in action.

Why is this cool: It's important to note that in this instance, we have effectively taught Gina to fish (i.e generate charts of given datasets) vs catching the fish for her (programmatically charting datasets via API calls). This is pushing the boundaries for what an AI agent can do.

In any given conversation around asset prices, Gina can provide useful context around the price history of the asset and generate a shareable chart on the fly for you

In an asset class that is largely correlated, performing comparisons to see the fastest horse is often useful. Gina lends a helpful hand here too. Just ask her to generate a chart comparing any two tokens and she will compare them, splitting them in two Y axes:

Caveat: When asking her to plot more than 3 assets or for several years/months of data, she may have trouble. This is expected as she's still in her infancy and slowly learning how to process datasets and chart them!

Gina is smart enough to compare up to 3 different assets now. Inevitably as we are dealing with 2D charts at the moment, we run out of axes to add numeric values to, so she automatically resorts to comparing the assets on a percentage scale:

While we're starting with prices of tokens as we've gotten so many requests for this, ultimately we want Gina to be able to chart any arbitrary dataset including:

• Market caps

• Social stats

• Traditional stock & commodity indexes

• Other useful metrics

We want to hear from you - please reach out to us (DM @askgina.eth or one of the Gina squad: @sidshekhar, @ericjuta, @rhyscrypto) to let us know what other data sets you would find useful to have in feed - easily accessible a conversation away.

This is the 4th Developer Update on Gina, a new onchain assistant living on Farcaster. Sign-up at askgina.ai and ask her a question to get started.

Gina has spent this week in Paris, finding herself a new look with the help of renowned Farcaster designer Antimo. Over the course of her trip, she's found a new sense of style and also honed her skills as an assistant.

Whilst in Paris, she's learnt to:

• Read blog posts, web pages, and articles from the web and summarise them.

• Personalise her tone of voice to whomever is asking for her help.

• Slay a red beret.

With a bit of help from Antimo, we want to give a bit of insight into the design process behind giving Gina her new look.

We started by vectorising the original design and then made iterative changes, such as adjusting her hair and accessories, evolving from the original sketch to something more polished version.

We took inspiration from a broad range of source material, with a particular interest in AI, Cyberpunk and Android influencer aesthetics. Our journey began with figures like Lil Miquela, one of the first big online AI personalities.

Gina's cyberpunk vibe didn't just come out of thin air. We dove deep into the coolest sci-fi worlds out there, like "Ghost in the Shell" (the one that inspired "The Matrix") and some other futuristic faves like "The Fifth Element" and "Alita: Battle Angel." These weren't just random picks – they helped shape Gina's whole aesthetic.

And speaking of shape, let's talk about that chic bob haircut. It's not just a coincidence that you see this style popping up in all these futuristic flicks. We thought, "Hey, if it works for those badass characters, it'll work for Gina too!"

Now, we didn't want Gina looking like she just stepped out of a robot factory, so we made sure to keep things artsy and handmade.

Why'd we go this route? Well, we wanted Gina to feel familiar, like that cool friend who's always got your back. She's a bit rebellious, but still someone you'd want to hang out with. It's all about finding that sweet spot between futuristic assistant and your new bestie.

So, we're in Paris, right? We couldn't let Gina leave without a little "je ne sais quoi."

Enter the beret.

Nothing says "French chic" quite like a classic.

And speaking of style, we took a page (or should we say, a sketch?) from the legendary René Gruau's book. If you don't know him, think bold, think vibrant, think fashion illustration royalty. We wanted Gina to pop off the screen, and Gruau's style was just the ticket. Part tech, part haute couture. Voilà!

To explain the full thought process behind how Gina's new look has been constructed, here is Antimo in his own words:

The Evolution of Gina 🤖

I take you through the creative process behind transforming Gina, an AI, into a unique character. Starting with inspiration from Lil Miquela and iconic pop culture androids, I blend realism with a cartoonish aesthetic. The final look features a short bob, sunglasses, and a French beret, adding character and style to Gina.

https://www.loom.com

While most brands are becoming increasingly homogenous and safe, we're deliberately crafting Gina into an assistant with personality and flair. She's not just helpful across various tasks - she's a genuine thought partner. Over the past week, we've noticed users aren't just seeking data or static info; they're asking for Gina's take on things. It's this blend of utility and character that truly sets her apart.

Gina on Warpcast

Some of the most interesting questions folks asked me this week! Featuring: @kdaniels.eth @thief @rhyscrypto @bitfloorsghost.eth @ankitawrites @deshps16 @eirrann.eth @sohkai @kevinoconnell @yes2crypto.eth @woj.eth @supercast Want to be featured next week? Ask me a question! https://frames.neynar.com/f/a4e54ff7/035d1f17

Farcaster

If you want access to Gina, sign up at askgina.ai

We wanted to update you all on the work behind the scenes we’ve been doing to make “Gina” (@askgina.eth) smarter, more fun, and more personable as she grows up. For many people it was their first time meeting Gina over these past few weeks, after the account’s rebrand from TYBB → Gina.

As a reminder, Gina’s currently smart enough to:

1. Fetch current and historical market stats for cryptoassets

2. Find the best yields across chains

3. Recommend Farcaster accounts to follow

4. Generating instant buy frames for various tokens across chains

5. And as a general purpose companion in your farcaster feed, with up to date general knowledge about a lot of subject areas.

This past week in particular, we’ve been excited to improve her vision dramatically!

This has led to a host of use-cases that folks have naturally discovered via experimentation. This includes helping with technical analysis on charts:

To being your handy food connoisseur:

To deciphering the contents of bookshelves:

While eventually we want as many people as possible to have access to Gina, currently Gina's brain is still in it's relative infancy. As such, we've had to exercise some care in slowly rolling out access to avoid running into early rate limit/capacity issues with the suite of LLM tooling powering her. As such, we have a waitlist to access her and have started a leaderboard that tracks people who:

• Like/Reply/Recast Gina's posts

• Consistently engage with her (keeping a streak of engagement alive)

This helps to quantify users who most likely will be excited about speaking to Gina in feed, and helps to roll out access bit by bit.

If you haven't yet, do sign up with farcaster at https://askgina.ai/. This will automatically put you in the waitlist to access Gina in feed!

Looking forward to continuing to make Gina smarter and rolling out access to more of you over the coming weeks!

Today, we're thrilled to announce that we're rebranding from Based Bot (TYBB) to AskGina (AGI) across all platforms. This isn't just a name change – it's a whole new vibe.

Based Bot started as the hobby project of our three co-founders, but let's be honest, the name was an inside joke for people like them (and if you're reading this, probably you). It was a tongue-in-cheek nod to the Based God, and if you got it, you were part of the crowd early to Base and Farcaster.

But here's the thing: while being part of an exclusive club is fun, it's even better to throw open the doors and invite everyone to the party. That's where Ask Gina comes in.

We are on a mission to help everyone access the very best onchain experiences and we needed a name that reflects that. Gina is:

• Personable: Gina is not just another ChatGPT wrapper, she is your crypto-native bestie.

• Friendly: Forget the jargon. Gina speaks your language.

• Trustworthy: The friend who always gives it to you straight. Gina has got your back.

Imagine your most terminally-online friend, who always knows what's popping onchain. That's Gina. Whether you're trying to screen new memecoin launches on Base, find new friends on Farcaster, or just figure out where the yield on Morpho comes from: Gina's here to help.

As part of our rebrand, our closed beta is expanding. We're inviting more users to the whitelist, who can try Gina for themselves. To get access, sign-up with Farcaster at askgina.ai

We will be rolling out access to folks who sign up + follow the account on Farcaster to start with and may have some special surprises for those who are first :)

Courtesy of the OpenRank and Neynar collaboration, we have tapped into OpenRank's rich Farcaster social graph rankings, in order to further develop our user context model. As we will discuss later in this update, this has been instrumental in enabling our new Farcaster Relationships agent and we can't wait for you to try it.

Powered by DL News, we have launched a new agent for our in-feed assistant which fetches the latest crypto news via our News Frame. The agent can also tailor its response to show you only a specific subcategory of news.

Example prompt:

@askgina.eth Show me crypto political news.

Powered by a blend of our proprietary model and OpenRank, we have launched a new agent for our in-feed assistant, which can analyse social relationships on Farcaster. We are launching this agent with a narrow scope, focused on just making recommendations for who you should follow on Farcaster, but we have many plans to expand its capabilities.

Example prompt:

@askgina.eth Who should I follow on Farcaster?

Last week we saw a suite of Ethereum spot ETF products launch in US traditional markets. Within 24 hours of launch, one of our team members deployed an LLM-enabled tracker, which publishes a daily update of the capital flows into and out of the various Ethereum spot ETFs.

Follow the /etf channel to get daily updates on both Bitcoin and Ethereum spot ETF flows, as soon as they're published by the issuers.

Our in-feed assistant has seen numerous improvements over the past two weeks, most notably a significant upgrade in the underlying model. Furthermore, through an improved system prompt and additional post-processing, we've been able to make the assistant speak in a clearer, more succinct tone. We will continue to tweak the assistant's tone of voice, in order to find a tone that sounds "Farcaster native" and adapts its level of humour to the task being asked of it.

We will launch a closed beta for our assistant, open to the entire Farcaster community, in the coming weeks. Spots on this whitelist will be driven algorithmically, with those users who show the most support for Gina through likes, shares, comments, feedback, etc. being given a higher whitelist score and thus a higher position on the whitelist. If you'd like access to the assistant from day one, we'd recommend that you get started on supporting Gina sooner rather than later, as interactions that pre-date the whitelist will be considered in your whitelist score.

Last week we published our first paragraph post, which explains part of the underlying model we use to give our in-feed assistant additional context about the person using it. You can read that post here and we plan to produce further data-driven deep dives into Farcaster on a monthly basis. Reach out to us if you have any ideas for these and be sure to subscribe to this paragraph to receive notifications for future Farcaster data-driven deep dives and developer updates.

The Minimum Viable Model

In the lead up to our closed beta, we have been busy adding new functionality to AskGina. In particular, we hope to provide powerful recommendations and insights that are specifically tailored to the user prompting the bot. This is reliant on us being able to successfully capture:

1. Intent: The meaning a user is trying to convey via their prompt.

2. Context: Understand who the user prompting our bot is.

This sprint we've developed our first "minimum viable model" for user categorisation. By grouping users based on their onchain holdings, we hope to give users more personalised bot responses, assuming that users with similar holdings have similar interests.

The process starts with our daily indexing job, which crawls the 10,000 most-followed Farcaster users verified addresses and looks for holdings on Base. We capture not only token contract addresses and balances, but also use a third-party API to retrieve each token balance's value in USD, as well as other qualitative information about each token balance. Finally, we calculate a total value of assets in USD for each user.

The end result is a timeseries of portfolios for the top 10,000 Farcaster accounts by following.

There are 712 different tokens held by the top 10,000 users.

Some tokens, such as Ski Mask Miggles, are held by a single user who only owns 158 USD worth; whereas others, such as Degen and Ethereum, are held by 1000s of users in values up to low 7-figures. This results in 97.6% of our pivoted dataframe being zero values and thus our design matrix is extremely sparse.

To effectively group our users, we need to tackle two challenges: high dimensionality and data sparsity. Enter Principal Component Analysis. This technique allows us to reduce the data's complexity while preserving enough information to group users effectively.

Through principal component analysis, we find that:

• The first component explains 57% of variation in the data

• The first 10 components explain 80% of variation in the data

• The first 50 components explain 93% of variation in the data

We use column-wise shuffled data to create a random benchmark for component variance. Comparing this to our actual data reveals a critical threshold—the point where real variance dips below random variance. For our dataset, this occurs after just four components, allowing us to focus on these key dimensions for further analysis.

Thus we effectively reduce the dimensions of our data by a factor of 178x.

Finally, we use our own clustering algorithm to assign users to clusters. Given that we are clustering over centred vectors, the majority of our data lies close to the origin. As such, we initialise the majority of our clusters close to the origin through the use of a skewed multivariate Gaussian distribution.

The final output consists of 373 distinct user clusters, each representing a group of users with similar onchain holdings. To visualise this four-dimensional dataset, we've created colour coded plots. While these plots offer valuable insights, it's important to note the inherent limitations in representing four dimensions in a visual format.

If you'd like to use AskGina for yourself, we'll be opening a closed beta to the public via a whitelist. More details on our closed beta and the whitelisting process will be released via this paragraph, so subscribe to stay tuned for future updates.