enter the veldt

Compared to a more general chat application like Perplexity, the veldt will have more specific and recent information specific to ai and crypto.

In addition to specific chats, users can scroll through the feed of podcasts and research and use quick ai prompts to get more details, generate a bull/bear case, and more. See the feed below containing a Delphi podcast where the bottom-left "ai" button in the post is pressed showing a few quick action prompts such as get "more details" and "Bull v Bear". Pressing "more details" auto prompts the ai copilot on the right-hand side which answers you with both content from the podcast and related posts.

Beyond consuming the default feed, you can have the ai generate a custom feed of posts or video podcast snippets for you. For example, below we selected the video generation menu button and used the default "market trends" prompt to generate a feed of podcast clips with market predictions. You can then watch the ai-selected clips while skimming the auto-generated summarization of the new feed.

While there are more features, the gist of the app is to synthesize recent crypto and ai information for you. The veldt ai can help you explore trends, research protocols, write content, and more.

The app's monetization (ai ain't free) works off of credits that are consumed as you message the ai and use the ai-related features such as generating a content feed. You get 50 free credits when you sign up which should last a few days to a week. From there, you can purchase more credits with the embedded Solana wallet (Privy) by transferring in SOL to your addresses.

Coming out of the beta phase, we will add more features around user profiles, content creation, creator rewards/monetization, and payment methods (stripe). Beta testers (i.e. users who purchase credits and use the app somewhat regularly) will be critical for the future iterations of the product and thus I'm deeply appreciative of anyone who takes an earnest interest in using the beta product.

short - I hate scrolling twitter (especially the new ‘for you’ page. jfc). All I want is an ai to help me understand and build upon what y’all cook up without all the bs injected. The veldt is simply me designing a tool I want and opening it up for others.

long - build something ambitious that articulates my thesis on how ai changes communication paradigms and subsequently content platform models.

[warning - there's far more bs below than above]

Across time communication paradigms from writing, to telegraph, radio, phone, all the way to the internet, have a) increased the number of people a person can draw information from and b) decreased the time to communicate with said person. Plotting the data transmission rate over time with different technologies on a log-axis plot illustrates the impact of communication technology.

While the trend is obvious -- technology makes it faster to communicate with more people over time -- the impacts on organizational structures, information supply chains, and businesses across those paradigms are not.

Among a litany of examples, a simple one is the business consolidation phase of the 1980s and 90s where advanced information speeds (shown above) changed the optimal organization structure.

Communication technology such as phone, fax, and email enabled managers to receive and send more information to wider audiences leading to more employees per manager and thus operational cost efficiency in consolidation. The resulting lower unit cost structures from faster, flatter organizations were used to undercut less-scaled competition.

Effectively, multi-order-of-magnitude changes in communication efficiency change how information flows and thus how humans organize. And, in the periods of communication technological bursts (gilded age, internet age, etc.) we have seen aggressive companies/armies harness the technology to form new organizational structures with inherent competitive advantages over less technologically agile incumbents.

However, while we've seen noticeable communication improvements in business organizations, it pales in comparison to cultural information dissemination via social media distribution platforms. Modern businesses still have roughly span and controls of 15-20 (i.e. a 1 to 20 peer-to-peer information ratio) while social media platforms effectively have a peer connection ratio of one to a billion where each individual user can both theoretically reach billions of people and consume content from them. This 1 to 1 billion information ratio leads to a dramatically fast information supply chain for cultural events and news.

The disparity between intraorganizational information speed flows and societal information flows across social media platforms arises from a few key assumptions in the social media platform model which leads to inefficiencies at smaller scales such as an organization.

1. Full Post Relavence: posts are shown in their entirety to the consuming user putting the burden on the consumer to find the relevant points within. Furthermore, to synthesize across a topic, users must consume multiple full posts. This constraint leads to shorter, less in-depth content versus longer, more detailed and nuanced content utilized in organizations.

2. Advertisement Monetization Model: advertisement models rely on time on platform and, in order to achieve predictability the user will stay on the platform after algorithmically putting a post next in the feed, the platform requires content predictability. This leads to a) water-down mass-market content with generic appeal to wider audiences and b) highly concentrated creator dynamics since larger, known creator brands offer lower consumption friction, more data, and thus higher algorithm predictability coming at the detriment of small to medium creator accounts (particularly dense information creators).

3. Consumption vs Action Oriented: social content platforms are biased towards entertainment consumption where the consuming user isn't expected to repackage or take meaningful action on the content.

4. Viewable Content: people need to see a post's content before they can contribute to, comment on, or share. This constraint means only open, public content can benefit from current internet scale models.

These assumptions and constraints of traditional social content platforms make the model ill-suited for organization information flows since organizations rely on dense, industry-specific information synthesis to create (often) proprietary content with audience sizes too small for advertisement models.

However, with llms and crypto, we see a new platform model and information supply chain possible that will scale organizational information ratios to social media platform levels.

We'll call this model peer-ai-peers (paps) where creator users connect or upload notes, documents, voice chats, etc. to the ai engine which retrieves them piecemeal to generate dynamic UIs for the consuming user. So instead of consuming the content from the thousands of people you follow, you actively use bite-sized pieces of content from people you follow to create new, specific documents.

In effect, this model enables millions of people to contribute to content without having to view the new content. If you imagine prior to the rise of llms, information disbursement was near instantaneously at near zero marginal cost, making the bottleneck in the information supply chain distillation of the information volume.

The paps model changes this and we can envision how it might look in the context of the veldt -- post beta ofc :)

This is possible because llms and crypto enable:

• Synthesization at Scale: llms and retrieval systems can rationally stitch together content from thousands of sources and languages to synthesize meaning for an end user or task.

• Generative UI: llms can generate custom UI components specific to a user's task or request turning content products away from consumption-based to creation-based.

• Open Financialization: Crypto enables business models beyond ad or subscription models by capturing financialization value leaked to the traditional financial industry such as exchanges, brokers, etc. Basically, token issuers can program their assets to capture fees instead of traditional intermediaries. Users benefit by receiving programmable payment streams for contributions to ai responses and these payment streams can further be absorbed into downstream financial products. Overall, crypto offers a more expressive monetization surface area all while having less payment friction and costs for all parties.

• Contribution without Viewing: llms can aggregate a corpus of a person's documents and opinions to synthesize contributions to content without them actually having to view the content. For example, if you have all my analysis on a particular stock, I don't need to read your proprietary investment memo for your llm to use my analysis in your memo.

In a nutshell, it's now possible to create a platform that incentivizes creators to create deep, in-depth content that can be absorbed into thousands of organizations all at once.

Say a portfolio manager investing in semiconductor stocks wants to reassess the landscape via a report.

Instead of drawing information from the few members of their team and their organization, they should directly be able to generate a report with an ai copilot which draws upon the knowledge of the expert's tokens they hold.

For example, they would likely want to subscribe to a few financial analysts covering various stocks in the sector, regional political experts, technology specialists, supply chain analysts, data center analysts, macro analysts, and various other micro analysts regularly providing insights into the space.

When generating the analysis, your ai copilot should be able to retrieve pieces of content across all of the creators you've subscribed to in order to synthesize an advanced analysis than possible with a base LLM trained on open data every so often.

Subscriptions to the various content analysts can be priced on a bonding curve (or similar monetization mechanism) enabling them to earn market rates for their services versus them trying to guess the market value of their content in a monthly subscription fee. People and organizations in demand of the creator's content within their ai systems would bid on access thus allowing the creator to effectively 'work' for multiple firms at once while receiving market rates for their labor.

Going a step further, because content creators' earning history and holders' buying patterns are open onchain, one could build protocols to a) wrap the earnings from a creator's tokens in a smart contract to then b) utilize its future earnings as collateral in loans. From there, thousands of individual loans can be packaged into products (regional regulations permitting) where investors can buy and sell future knowledge interest in specific, granular industries or technologies.

For example, say an investor was bullish on knowledge demand for a specific fabrication method of gpu chips versus knowledge demand for a different type of chip, they could buy interests one while shorting the other. This mechanism enables financial markets at the labor demand level in a granular fashion versus at the aggregated company level effectively creating a global information marketplace.

Another downstream affect of having access to deep content creators tokenized is that labor becomes a real, measurable asset. An organization might hold $10 million worth of creator tokens which a) provides a recognizable balance sheet asset for labor assets held and b) opens the door for financial operations against this asset, now that it's measurable.

Additionally, this will likely lead to firms that specialize in aggregating labor assets as a capital asset and leasing out access as an operating expense to less cash-rich companies.

The current product is in a premature beta phase with a primitive ai copilot specific for crypto and ai industries. Users are not yet able to connect or upload their content nor are they able to follow other users yet. Effectively, all you can do in the current product is use the copilot to chat, write short content, and explore crypto-related content.

However, this will change as we come out of the beta phase. Look for more content creation, token monetization models, and social features in the coming weeks/months.

For now, we appreciate people willing to try out the primitive product and new features as they are released as it meaningfully helps us scale the system and calibrate the product features.

see you in the veldt

ps - the word 'veldt' refers to an open rural landscape in southern Africa which, in ways, embolizes Darwinian competition. Post the rise of llms, labor markets will become competitive forcing humans to Darwinian competition with both themselves and ai. That competition will produce lions and dik-diks -- you have decided which its going to be. Additionally, the short story by Ray Bradbury articulates well the result of fully abdicating in favor of ai.

<100 subscribers