Author: @knimkar Translation: Vernacular Blockchain

We seem to be entering the intersection of AI and encryption Use case experiments for the Cambrian explosion phase. I'm very excited about what's coming out of this energy and wanted to share some of the exciting new opportunities we're seeing in the ecosystem at @SolanaFndn.

1) Promote the most dynamic agent-driven economy on Solana. Truth Terminal demonstrates for the first time what is possible when AI agents are able to interact on-chain. Achievement. We look forward to seeing experiments that safely push the boundaries of what agents can do on-chain. The potential in this area is huge, and we haven’t even begun to explore the design space within it. This is already proving to be one of the most unexpected and explosive areas of convergence between crypto and AI, and it’s just getting started.

2) Let the large language model (LLM) perform better in writing Solana code, empowering Solana developers. The large language model has already performed better in writing code. Pretty good, and they're going to get even more powerful. We hope to leverage these capabilities to make Solana developers 2 to 10 times more productive. In the short term, we will create high-quality benchmarks to measure LLM's ability to understand Solana and write Solana code (more on that below), and these tests will help us understand LLM's potential impact on the Solana ecosystem. We look forward to supporting teams making high-quality progress in fine-tuning their models (and we validate the quality of these models by performing well on benchmarks!).

3) Support open and decentralized AI technology stack. What we call "open and decentralized AI technology stack" refers to the ability to promote access An open and decentralized protocol for the following resources: data for training, computing resources (for training and inference), model weights, and the ability to verify model output ("verifiable computation"). This open AI technology stack is important because it:

Accelerates experimentation and innovation in the model development process

Provide a way out for those who may be forced to use untrusted AI (such as approved AI)

We want to support teams and products building at all levels of this technology stack. If you are doing work related to these key areas, you can contact the author of the original article!

Below, we will explain in more detail why we are excited about these three pillars and what we want to see built.

1) Promote the most dynamic agent-driven economy

Why do we care about this? There has been a lot of discussion about Truth Terminal and GOAT, and I won’t repeat it here, but what can be clearly said is that all kinds of crazy functions that are possible when AI agents interact on the chain have irreversibly entered reality (and In this case, the agent hasn't even taken action directly on-chain yet).

We can confidently say that at this time we cannot know with certainty the future of agent behavior on the chain What that will look like, but to give you a sense of how vast this design space is, here are some of the things that are already happening on Solana:

Like Truth Terminal of AI leaders are trying to foster a new era with memecoins like $GOAT ;

At the same time, applications like @HoloworldAI, @vvaifudotfun, @TopHat_One, @real_alethea allow users to easily create and launch agents and related Tokens.

By training AI fund managers to act as personalized agents for various well-known crypto investors, Make investment decisions and fuel their portfolios. For example, the rapid rise of @ai16zdao at @daosdotfun has created a new metaverse of AI funds + agent cheerers.

There are also some agent-centric games, such as @ParallelColony. In these games, players ask the agent to take actions through instructions, which often results in unexpected consequences. result.

Possible next directions:

Agency management of multi-faceted projects that require economic coordination among all parties. For example, the agency could be tasked with tasks such as "finding a compound that can cure [X] disease" is such a complex task. Agents can do the following:

Raise funds via tokens on @pumpdotscience;

p>

Use the funds raised to pay for access to relevant paid research and to pay decentralized computing networks (such as @kuzco_xyz, @rendernetwork, @ionet etc.) to run simulations on various compounds;

Use bounty platforms like @gib_work to recruit humans to perform tasks that actually work (For example, run an experiment to verify/improve the simulation results);

Or perform a simple task, such as helping you build a website, or create an AI work of art (e.g. @0xzerebro).

There are many other possibilities.

Why does it make more sense to have agents perform financial activities on-chain (rather than in the traditional financial system)? Agents are fully capable of leveraging both the traditional financial system and cryptocurrencies. Here are a few reasons why cryptocurrencies are particularly well-suited for certain aspects:

Micropayment scenarios - Solana excels in this area, and apps like Drip have demonstrated its potential.

Speed ​​– Instant settlement can be crucial for agents, especially if you want them to be optimal in terms of capital efficiency.

Access to capital markets via DeFi – The advantages of cryptocurrencies become particularly clear once agents begin to conduct financial activities beyond strictly payments. This is perhaps the strongest reason for agents to participate in the crypto economy. Agents can seamlessly mint assets, conduct transactions, invest, borrow money, use leverage and other operations.

Solana is uniquely suited to support this type of capital markets activity becauseThere is already a wealth of top-tier DeFi infrastructure for the Solana mainnet.

Finally, technology is often path-dependent. The key is not which product is the best, but the first one to reach critical mass and become the default path. If we see more agents creating significant wealth through cryptocurrencies, this could solidify crypto connectivity as an important capability for agents.

What we want to see

A proxy combined with a wallet to be able to perform operations on the chain A bold experiment. We do not give an overly specific definition here because the possibilities are very broad and we expect that the most interesting and valuable agent application scenarios will be those that we cannot predict. However, we are particularly interested in the exploration and infrastructure construction in the following directions:

At least in the prototype stage on the testnet (preferably on the mainnet)

2) Let LLM be good at writing Solana code and empower Solana developers

Why do we care about this? LLM already has strong capabilities and is improving rapidly. But writing code is a direction that deserves special attention in the application field of LLM because it is a task that can be objectively evaluated. As explained in the post below, “Programming has a unique advantage: by ‘self-playing,’ superhuman data scaling can be achieved. Models can write code, then run it, or write code, write tests, and check for self-consistency. ”

Limiting the negative impact of hallucinations - current models are very powerful, but still far from Not perfect. Agents cannot be given complete freedom to perform actions.

Promote non-speculative use cases - for example, let you buy tickets through @xpticket, optimize returns for a stablecoin portfolio, or buy food on DoorDash, etc. .

Currently, although LLMs are still far from perfect at writing code and have some obvious shortcomings (for example, they are poor at finding vulnerabilities), Like Github CTools like opilot and the AI-native code editor Cursor have fundamentally changed software development (and even changed the way companies recruit talent). Given the projected rapid progress, these models are likely to revolutionize software development. We hope to leverage this progress to make developer productivity on Solana an order of magnitude faster.

However, there are currently some challenges that hinder LLM's performance in understanding Solana:

No Sufficient high-quality raw data for LLM training;

Lack of sufficient verified builds;

On sites like Stack Overflow In such a place, there is a lack of sufficient exchange of high-value information;

Solana infrastructure is developing rapidly, which means that even code written 6 months ago may not be Perfectly suited to current needs;

There is no way to evaluate how well a model understands Solana.

What we want to see

Help us publish better Solana data on the Internet !

More teams release verified builds.

I hope more people in the ecosystem can actively participate in Stack Exchange and ask good questions and provide high-quality answers;

Create high-quality benchmarks to assess LLM's understanding of Solana (RFP to be released soon);

Create an LLM that scores high on the above benchmarks Fine-tune releases and, more importantly, speed up the work of Solana developers. Once we have a high-quality benchmark, we may offer rewards for the first model to achieve the benchmark score - stay tuned.

The final achievement here will beA high-quality, differentiated Solana validator client created entirely by AI.

3) Support open and decentralized AI technology stack

Why do we care about this a little? It’s unclear how power in AI will be balanced between open source and closed source AI in the long term. There are good arguments made as to why closed source entities will remain at the cutting edge of technology and capture most of the value from the underlying model. Right now, the simplest expectation is that the status quo will continue - large companies like OpenAI and Anthropic pushing the technology frontier, while open source models will quickly follow suit and eventually have uniquely powerful fine-tuned versions for certain use cases. We hope Solana can be closely integrated and support the open source AI ecosystem. Specifically, this means facilitating access to: data for training, computing power for training and inference, weights for the resulting models, and the ability to validate model outputs. The specific reasons why we think this is important are:

A. Open source models help accelerate model development, debugging and innovation. How open source communities can quickly refine and fine-tune models like Llama Such an open source model shows how the community can effectively complement the efforts of large AI companies in advancing the frontiers of AI capabilities (even Google researchers pointed out last year that "we have no moat, and neither does OpenAI" regarding open source). We believe that a thriving open source AI technology stack is critical to accelerating the rate of progress in the field.

B. Provide an exit for those who may be forced to use AI they do not trust (such as recognized AI) that may now be dictators or authoritarian regimes The most powerful tool in your arsenal. Approved models provide a sanctioned version of the truth and become a tremendous means of control. Highly authoritarian regimes may also have better models because they are willing to ignore the privacy of their citizens to train their AI. The question of AI being used as a control tool is when, not if, and we want to prepare for this possibility by supporting open source AI technology stacks wherever possible.

Solana is already home to many projects supporting open source AI technology stacks:

Grass and Synesis One is facilitating data collection;

@kuzco_xyz, @rendernetwork, @ionet, @theblessnetwork, @nosana_ai, etc. are providing a large number of decentralized computing resources.

Teams like @NousResearch and @PrimeIntellect are working on developing frameworks to enable Centralized training is possible (see below).

Us What I hope to see is the development of more products at all levels of the open source AI technology stack:

Decentralized data collection, such as @getgrass_io, @usedatahive、@synesis_one

On-chain identity authentication: includes protocols that allow wallets to prove they are human, as well as protocols that verify AI API responses so consumers can confirm they are interacting with LLM

Decentralized training: such as @exolabs, @NousResearch and @PrimeIntellect

Intellectual property infrastructure: enabling AI to be licensed (and paid for) The content they exploit