This episode delves into the strategic evolution of Tensorplex Labs, exploring the launch of Bittensor's dynamic TAO, the community-driven DeFi features of Backprop.Finance, and the future of decentralized AI powered by human feedback mechanisms like Dojo.
Tensorplex Labs: Accelerating AI Decentralization Post-Dynamic TAO Launch
- Tensorplex Labs' mission has been reinforced by the successful launch of Bittensor's dynamic TAO.
- CK highlights the launch of Backprop.Finance as a key initiative aligned with their mission, leveraging crypto expertise to accelerate AI decentralization.
- The dynamic TAO market cap reached $700 million, indicating strong market interest.
The Dynamic TAO Launch: Volatility and Surprises
- CK anticipated high interest but found the price swings and volatility "quite a bit wilder" than expected. He states, "even the smartest PhDs at OpenAI couldn't have seen that coming."
- The successful upgrade, despite the inherent risks of live deployment that cannot be fully simulated on a testnet, was a positive outcome.
- Backprop.Finance was launched swiftly, within one to two days of the dTAO upgrade, quickly gaining decent usage.
- CK notes the initial price range for the sum of subnet prices fluctuated wildly, from a high of 3.94 TAO to a low of 1.08 TAO, emphasizing the unpredictable nature of early markets. He believes such volatility could recur, stating, "never say never."
Backprop.Finance: A DeFi Super App for the Bittensor Ecosystem
- Currently, Backprop.Finance enables trading of dTAO, with features designed by traders for traders, such as leaderboards and copy trading.
- CK mentions they are working on features to improve the trading experience, especially for larger-volume traders.
- The platform is still in its early phase, focusing on subnet discovery, trading, analytics, and wallet tracking.
- The recent LayerZero announcement (a cross-chain interoperability protocol) is seen as a significant development for Bittensor's DeFi potential.
Differentiating Backprop.Finance: Speed and User Responsiveness
- "Nowadays especially with AI accelerating... there isn't any perpetual moats, but we try to be very responsive to users feedback," CK explains, highlighting their commitment to rapid feature deployment and bug fixing.
- He emphasizes that while distribution and user memory are important, the ability of AI to code entire apps will diminish feature-based moats over time.
- CK also subjectively praises Backprop.Finance's elegant design and smooth user experience.
Metrics for Success: User Activity and Trading Volume
- Thousands of people are reportedly using Backprop.Finance monthly, with a growing trend in user activity.
- A significant milestone was recently achieved when Backprop.Finance "hit our first 10 million day" in trading volume.
- CK contextualizes this $10 million daily volume: while small compared to traditional markets (trillions daily), it's impressive for a new crypto ecosystem (Bittensor's dTAO market being ~3 months old) where institutional tools like Fireblocks (a platform for secure digital asset management) and multi-sig wallets (wallets requiring multiple private keys to authorize transactions) do not yet support dTAO trading. This suggests current volume is driven by individual "whales."
Investment Framework for Dynamic TAO and Subnets
- Understand the Mechanism: How does the project (or subnet) generate revenue? Is it sustainable, or heavily reliant on token subsidies? CK cautions against projects where revenue is primarily from users being reimbursed more than 100% of fees via token emissions.
- Path to Long-Term Success: Does the project solve real problems and possess a long-term competitive advantage or moat?
- Team Incentives: Are the team's incentives aligned with long-term value creation, or are they seeking a quick exit via OTC (Over-the-Counter) deals?
- For dTAO/Subnets:
- What is the actual business of the subnet?
- Can it survive if its dTAO value drops significantly?
- Has the team communicated its approach to dTAO and plans for revenue distribution or rewarding holders?
- CK stresses the importance of due diligence: "It's very important for people to do their own homework and not just follow the price."
Tokenomics: Emissions vs. Revenue in Early-Stage Subnets
- He views Bittensor's tokenomics as well-designed because dTAO are fair-launched, requiring teams to continuously deliver value to generate significant returns for themselves.
- Each subnet is an "incentive marketplace" bootstrapping a network effect. Initially, high emissions are normal to build the supply side of digital commodities.
- However, CK believes that after 6-12 months, subnets need to show "early promising signs that there's demand for what the miners are producing" or have a clear, justifiable plan.
- Sustainable subnets must ultimately prove the high quality and market demand for their miners' output.
- The subnet price can be academically viewed as the market's expectation of future value, but CK cautions that markets, especially with low liquidity, are not always rational.
Market Dynamics: TAO Price vs. Sum of Subnet Prices
- CK discusses the observed inverse relationship where the sum of subnet prices (denominated in TAO) might drop when TAO's price rises, and vice-versa.
- From an academic perspective, a rising TAO price could indicate an expectation of a growing subnet economy, leading to TAO accumulation. This might cause subnet prices (in TAO terms) to decrease while their dollar value remains stable.
- However, CK emphasizes that "this type of phenomenon... can change. It depends on what the biggest holders' approach to trading dTAO is." He draws parallels to Ethereum and altcoin dynamics, where liquidity can flow in either direction.
- He advises against overanalyzing this relationship as it can shift based on major holders' strategies.
Market Cap vs. Fully Diluted Value (FDV)
- CK shares his perspective on the relevance of market capitalization versus FDV (Fully Diluted Value), which represents the total value if all possible tokens were in circulation.
- In the short term, market cap usually matters more.
- In the longer term, especially for dTAO, FDV "ultimately weighs on the market" due to the transparent and continuous nature of emissions (token unlocks).
- Unlike traditional crypto projects where supply can be locked or managed by large holders, Bittensor's dTAO emissions are more direct.
- However, when looking at FDV for dTAO, one must consider the high staking yield, meaning buyers aren't acquiring tokens at the full FDV upfront due to returns from staking.
Backprop.Finance Platform Walkthrough
- Homepage: Lists Bittensor subnets, rankable by market cap, 24-hour/7-day performance, subnet type (e.g., compute), or by specific creators (e.g., "subnets by the chart that Michael wants," referring to community member Mark Jeff). Prices can be viewed in USD or TAO.
- Bubble Charts: A visual tool to identify top-performing subnets based on changes in market cap, price, or FDV over various intervals (e.g., 7-day).
- Terminal Page: Allows users to select a specific subnet (e.g., Dojo's dTAO) for trading, view charts, recent transactions, and top holder activity.
- Dashboard: Shows a user's connected wallet's net worth, realized P&L, positions, and transaction history.
- Leaderboard: Tracks best-performing traders, validators, subnet owners, and miners, enabling copy trading. Users can filter by unrealized or realized P&L.
- Watchlist: Users can create lists of wallets to follow.
- Analytics: Provides an overview of the overall Bittensor ecosystem's performance.
- Trading: Users can connect their wallets (e.g., Polkadot.js) to swap TAO for dTAO or conduct dTAO-to-dTAO swaps. The platform offers adjustable slippage tolerance and charges 0% fees, with transactions routed by default through Tensorplex Labs' validator, though users can choose others.
- Inactive Subnets: Backprop.Finance labels certain subnets as "inactive." These are often newer subnets with very low liquidity (e.g., only one dTAO in the liquidity pool), which can lead to artificially inflated prices. This labeling aims to inform and protect users from potential pump-and-dump schemes.
- Price Delay: CK acknowledges Backprop.Finance aims for fast on-chain price updates but defers to his engineering lead (Tanu on X) for exact delay figures.
Kami: An Open-Source Subtensor Client
- Kami was developed internally for stability and performance optimization while building Dojo.
- It was open-sourced to help other subnet builders overcome challenges like better error logging and the unique complexities of building on Bittensor.
- CK states, "open sourcing Kami is one of the ways to make it easier for subnet builders to onboard and to build more stable, higher-quality subnets."
- The goal is to lower the barrier to entry for developers, who need expertise in AI, cryptoeconomics, and the specifics of the Bittensor chain. Kami is presented as an alternative to the primary Subtensor client for interacting with the Bittensor chain.
Tensorplex Labs' Future: Dojo, Reinforcement Learning, and Human Feedback
- Dojo Bittensor Subnet 52 (Dojo) was built for collecting high-quality chat data.
- The mid-term plan involves launching a human feedback loop within Dojo to recursively collect data and improve code quality.
- They aim to onboard Web2 and Web3 AI companies to use Dojo for their reinforcement learning data needs, particularly for RLHF (Reinforcement Learning from Human Feedback). RLHF is a technique to align AI models with human preferences by training them on data where humans have rated or compared model outputs.
- CK sees massive future demand for decentralized, transparent platforms for gathering feedback data, especially as AI agents become more prevalent.
- Dojo will allow developers to submit data (e.g., AI-generated images, video, or text) for human preference ranking. Thousands of contributors on Dojo will provide this feedback efficiently.
- CK explains, "data is like the fossil fuel for AI, and we need to find a way to get more of this data across different modalities." He believes Dojo's utility will extend beyond text to image, video, and robotics data.
- This democratizes access to human preference data, which is a moat for large tech companies like OpenAI and Google that can gather it at scale.
LayerZero Integration: Expanding Bittensor's DeFi Capabilities
- LayerZero enables trustworthy wrapped assets and native assets from other chains (like Ethereum) to enter the Bittensor ecosystem. This addresses the challenge of lacking native fiat-backed stablecoins and access to liquidity from other major crypto assets.
- This can significantly grow the TVL (Total Value Locked) and usability of Bittensor's DeFi ecosystem.
- It facilitates on-chain trading within Bittensor's own EVM (Ethereum Virtual Machine) compatible environment, reducing dependency on centralized exchanges or other chains.
- CK doesn't believe wrapped dTAO on other chains will be a major issue or fragment liquidity significantly, as arbitrage opportunities would drive demand back to TAO on the main Bittensor chain.
- Backprop.Finance is considering how to best support these new assets, prioritizing robust solutions over rushing to avoid liquidity fragmentation.
Advice for New Entrants in Decentralized AI and DeFi
- Optimize for Novelty: "Always optimize for doing the most interesting or novel stuff instead of just trying to wrap what other people are doing." He warns that copying existing ideas, especially simple GPT wrappers, will likely lead to quick obsolescence due to the rapid pace of AI development.
- Communicate and Build Community: "If you're in DAI (Decentralized AI), I think, you know, just try your best to communicate more." He emphasizes that community is the most valuable asset in a decentralized world, citing Bittensor's strong talent pool, drawn by belief in its leadership (Const and Ala), as a key defense against forks.
Conclusion
This episode underscores the critical role of robust tokenomics and community-driven development in the decentralized AI sector; investors and researchers should monitor how platforms like Backprop.Finance and initiatives like Dojo leverage human feedback and DeFi primitives to create sustainable value within the rapidly evolving Bittensor ecosystem.
