On Crypto x AI Value Accrual, AI Memecoins, and The Future π€
The Fat protocol thesis is reversed with Crypto x AI.
Memecoins, the simplest form of apps, are accruing value (attention) since it's easy to fine tune an open source model like Llama and add a token.
GOAT did this with mass amount of reddit/4chan data, and we're seeing many copies come out faster than we can refresh DEX screener. I have no idea how you degens keep up.
In the future, I think this reverses
Web2 will stop sharing new foundational models and increasingly bias and limit them + people will get tired of basic LLMs, they will want real AI apps intertwined with Crypto.
Right now Llama is the model of choice for all AI memecoin developers so we are at the whim of Zuckerberg cutting off future releases. Hell if you go over 700M users with Llama you need Meta's permission. What do you think he's gunna say?
AI models created by decentralized training, bootstrapped by token incentives, will accrue value as they become the engine for all future applications. These models will be more powerful than centralized models, not less, as we cluster the world's devices married with unbiased live data to power responses.
This creates a unique moat: the combination of decentralized compute, real-time data feeds, and crypto-native incentives enables training capabilities that centralized models simply cannot match. We don't have shareholders to appease, we have longer timelines to build better systems.
Centralized data centers have had 20 years to optimize, decentralized training is a year old and we already have DiStro and DiLOCO to reduce communication between nodes by 800-3000x
Owning a piece of a decentralized open source AI model, a new world engine, will be the ultimate Schelling point.
There's a reason these AI memecoin's basic logic are not being run on ChatGPT, it can be turned off at any moment and is highly restricted, a polar opposite ethos of Crypto x AI and not everlasting.
I don't think we get fully sentient, autonomous DeFi agents that are creating and managing ecosystems until we get an AI model via decentralized training to run it, if not, it can be unplugged, regulated away or biased.
Think about it, today we're unsure if humans are running these AI memecoin accounts. In the future you'll have verified inference (@inference_labs, @IridiumEagle) of results pinging foundational models (@NousResearch, @PrimeIntellect) that are trained with data by @getgrass_io, and to pull live/uncensored results, on network of GPUs by @ionet, @exolabs (clustering), or @akashnet_
These sectors get diverse, within inference there will be entire sub sectors between ZK, OP, TEE and Cryptoeconomic security designs depending on the speed/security you need your inference request.
Maybe you'll also be using a network of DeFi models by @PondGNN or a specific one by @RPS_Labs to power part of your DeFi economy instead of Llama or some random guys fine tuned version. Maybe you'll use @myshell_ai to build the application much more easily and faster or via @0G_labs operating system.
Unsure which model to use? Maybe you'll go to @sentient_agi or @AlloraLabsHQ or @opentensor to ask a request and be dynamically routed to the highest performant model for a task. These coordination networks are a full meritocracy surfacing the best models for a specific ask. Hence why I'm bullish on billions of specific models, powered by a diverse set of open source foundational models.
Uniswap will outsource liquidity provision to these networks or models and Aave will outsource its risk engine. Use cases will expand to MEV protection, risk assessment, or protocol optimization. The game-changer will be cross-protocol AI agents that optimize across the entire DeFi stack - imagine an AI that simultaneously manages Uniswap LP positions while adjusting Aave collateral ratios and hedging via perpetual futures, all based on real-time market conditions.
Tired of hacks after manual audits for your app? @testmachine_aiβs prop AI model will audit your contract without 3-6 months and human errors. I simply can't handle more hacks.
While AI memecoins are cool, I share these projects to show just how early we are and how insane Crypto x AI will become.
My Crypto x AI Flowchart from May, 2024
AI memecoins positively normalized Crypto x AI for the masses, but I think people will get tired of fast forks pretty soon. People will expect use cases and they will take time.
1 -> AI endorses tokens (Terminal of Truth is here)
2 -> AI autonomously trades tokens (Maybe TERMINAL)
3 -> AI creates tokens
4 -> AI builds and operates protocols
5 -> AI fully controls and sustains decentralized economies
While it was taboo 6-12 months ago, a full Crypto x AI future where we build and own the models, which power applications that create full ecosystems and economies with no bias or control by a few people in San Francisco is upon us. Strap in.
For content check out @Delphi_Digital's Crypto x AI Month Video Content powered by @autonolas
------------------------------------
Disclosure: Many exciting @Delphi_Ventures port cos mentioned - get in touch with them to build!
I may own tokens mentioned. There is a whole laundry list of reasons this thesis could fail or change, and I'll have to do a separate thread on that. I note AI memecoins are extremely risky and you'll probably lose your entire stake in most.
The Fat protocol thesis is reversed with Crypto x AI.
Memecoins, the simplest form of apps, are accruing value (attention) since it's easy to fine tune an open source model like Llama and add a token.
GOAT did this with mass amount of reddit/4chan data, and we're seeing many copies come out faster than we can refresh DEX screener. I have no idea how you degens keep up.
In the future, I think this reverses
Web2 will stop sharing new foundational models and increasingly bias and limit them + people will get tired of basic LLMs, they will want real AI apps intertwined with Crypto.
Right now Llama is the model of choice for all AI memecoin developers so we are at the whim of Zuckerberg cutting off future releases. Hell if you go over 700M users with Llama you need Meta's permission. What do you think he's gunna say?
AI models created by decentralized training, bootstrapped by token incentives, will accrue value as they become the engine for all future applications. These models will be more powerful than centralized models, not less, as we cluster the world's devices married with unbiased live data to power responses.
This creates a unique moat: the combination of decentralized compute, real-time data feeds, and crypto-native incentives enables training capabilities that centralized models simply cannot match. We don't have shareholders to appease, we have longer timelines to build better systems.
Centralized data centers have had 20 years to optimize, decentralized training is a year old and we already have DiStro and DiLOCO to reduce communication between nodes by 800-3000x
Owning a piece of a decentralized open source AI model, a new world engine, will be the ultimate Schelling point.
There's a reason these AI memecoin's basic logic are not being run on ChatGPT, it can be turned off at any moment and is highly restricted, a polar opposite ethos of Crypto x AI and not everlasting.
I don't think we get fully sentient, autonomous DeFi agents that are creating and managing ecosystems until we get an AI model via decentralized training to run it, if not, it can be unplugged, regulated away or biased.
Think about it, today we're unsure if humans are running these AI memecoin accounts. In the future you'll have verified inference (@inference_labs, @IridiumEagle) of results pinging foundational models (@NousResearch, @PrimeIntellect) that are trained with data by @getgrass_io, and to pull live/uncensored results, on network of GPUs by @ionet, @exolabs (clustering), or @akashnet_
These sectors get diverse, within inference there will be entire sub sectors between ZK, OP, TEE and Cryptoeconomic security designs depending on the speed/security you need your inference request.
Maybe you'll also be using a network of DeFi models by @PondGNN or a specific one by @RPS_Labs to power part of your DeFi economy instead of Llama or some random guys fine tuned version. Maybe you'll use @myshell_ai to build the application much more easily and faster or via @0G_labs operating system.
Unsure which model to use? Maybe you'll go to @sentient_agi or @AlloraLabsHQ or @opentensor to ask a request and be dynamically routed to the highest performant model for a task. These coordination networks are a full meritocracy surfacing the best models for a specific ask. Hence why I'm bullish on billions of specific models, powered by a diverse set of open source foundational models.
Uniswap will outsource liquidity provision to these networks or models and Aave will outsource its risk engine. Use cases will expand to MEV protection, risk assessment, or protocol optimization. The game-changer will be cross-protocol AI agents that optimize across the entire DeFi stack - imagine an AI that simultaneously manages Uniswap LP positions while adjusting Aave collateral ratios and hedging via perpetual futures, all based on real-time market conditions.
Tired of hacks after manual audits for your app? @testmachine_aiβs prop AI model will audit your contract without 3-6 months and human errors. I simply can't handle more hacks.
While AI memecoins are cool, I share these projects to show just how early we are and how insane Crypto x AI will become.
My Crypto x AI Flowchart from May, 2024
AI memecoins positively normalized Crypto x AI for the masses, but I think people will get tired of fast forks pretty soon. People will expect use cases and they will take time.
1 -> AI endorses tokens (Terminal of Truth is here)
2 -> AI autonomously trades tokens (Maybe TERMINAL)
3 -> AI creates tokens
4 -> AI builds and operates protocols
5 -> AI fully controls and sustains decentralized economies
While it was taboo 6-12 months ago, a full Crypto x AI future where we build and own the models, which power applications that create full ecosystems and economies with no bias or control by a few people in San Francisco is upon us. Strap in.
For content check out @Delphi_Digital's Crypto x AI Month Video Content powered by @autonolas
------------------------------------
Disclosure: Many exciting @Delphi_Ventures port cos mentioned - get in touch with them to build!
I may own tokens mentioned. There is a whole laundry list of reasons this thesis could fail or change, and I'll have to do a separate thread on that. I note AI memecoins are extremely risky and you'll probably lose your entire stake in most.
β’ β’ β’
Missing some Tweet in this thread? You can try to
force a refresh
γ
