kW- Gavin

Leading @impactrealm @spacemetaverse | Buidl in web3, living in metaverse

[kW- Gavin @kWgavin]

3/ I think we’re quietly building something new.

Not just AI assistants.

External cognitive architecture.

Still thinking through the implications, but I suspect this shift is more significant than most people realize.

[kW- Gavin @kWgavin]

Something I’ve noticed recently:

Every persistent thought I have now seems to end up with an AI agent attached to it.

If I’m researching a topic, there’s an agent collecting sources.

If I’m building a business system, there’s an agent helping me think through architecture.

[kW- Gavin @kWgavin]

2/ If I’m creating content, there’s an agent helping me identify patterns and refine ideas.

A year ago, all of that cognitive work lived entirely inside my head.

Now it lives across a growing network of tools, agents, notes, prompts, and workflows.

[kW- Gavin @kWgavin]

12/ I have found that improving the evidence trail creates more reliability than another validation layer. The more important the decision, the more evidence the workflow should require.

What do you think?

[kW- Gavin @kWgavin]

I used to think AI hallucinations were a model problem. After building a few workflows, I am starting to think they are a workflow design problem. And the fix is not making the AI more self-aware.

[kW- Gavin @kWgavin]

11/ Good AI systems do not rely on self-awareness. They rely on process design. Before adding another validation layer, ask: does this workflow trust confidence, or does it require proof?

[Andrej @0xdrej]

Crypto is not Web3 - but it can save it. Why AI is the next iteration of the internet and crypto is a critical part of keeping it free.

The internet began as a place of total freedom, where web 1.0 users existed completely outside of society’s restrictions. Anyone could post a website, and there was virtually no oversight of what got posted. It was a place where you could look for anything. Now it’s a place where things look for you.

What happened? Two things: smartphones and social media.

As billions of people rushed in over the past ten years, they’ve joined a fundamentally different internet. Today, the web no longer takes place on unique URLs. It takes place on centralized platforms, and it takes place on apps. This happened over the course of several years, but if you blinked you would’ve missed it.

So far, the consequences have been dismal. Despite its early promises, more users spending more hours on fewer platforms has ultimately meant one thing: much more data in much fewer places. This data translated into an unprecedented amount of power, and the platforms that promised to fix the world began abusing it.

What followed was a small group of digital actors getting wholesale access to our digital lives. They use our data for profit while maintaining a system of total surveillance and programmed content. Centralized companies now control what we see and observe what we do, and the egalitarian utopia that was promised to us has turned into a digital feudal state.

So ask yourself:

If Web 2.0 happened when the internet became more closed off, when free apps were offered to users whose data was then captured and sold – what is the next evolution of these trends?

It’s not crypto. It’s AI.

AI in its current state is the ultimate example of a closed internet built on extracted data. And whether we like it or not, Web3 began the moment AI had its iPhone moment two years ago. It began with the release of ChatGPT.

The difference is, the apps of Web 2.0 controlled our view of the internet. Web3 is on track to control our perception of reality itself.

There are less than five companies in the world right now who can achieve an internet scale web crawl. These companies basically control the internet, and by transitivity, the future of AI development.

Today, Grass has over 2 million users. At its current scale, Grass can scrape enough data for fine-tuning specific AI models or informing certain types of real-time inference. When it reaches 25 million users, it will be able to scrape a pre-training corpus large enough to train ChatGPT from scratch on a weekly basis.

Grass's vision extends beyond merely scraping data for model training; we want to fundamentally transform the internet. With 100 million users, Grass will be large enough to achieve an internet-scale web crawl, owned and controlled by its users.

If we can give users control of the rails by which data itself is acquired for AI, we can begin to seize power back from the small handful of companies who took over during Web 2.0.

We stand at a pivotal time in history. Web2 did not need (or necessarily want) to become evil; we just didn’t have the tools to prevent it. The system at the time fell victim of the market incentives that compelled it to act nefariously. Before the advent of blockchain coordination mechanisms, it was impossible to include users in the equation.

We have a chance to do things properly this time.

[kW- Gavin @kWgavin]

@matty_ Dm’ed your TG ser

[Darren Lau @Darrenlautf]

coingecko, etherscan, aptos, chromia, aevo, drift , jupiter, kaito, c4 , chasm , not3lau , virtuals , altlayer , layerzero , mantle , tangent, magic eden, pendle, pengu, abstract

Wait the malaysian cabal actually quite goated

[kW- Gavin @kWgavin]

@clwnwrldnjyr Sir any kang tou? Work to earn sounds good

[kW- Gavin @kWgavin]

@OX_DAO Everything everywhere all at once