Thoughts on Agents, Jobs to Be Done, and Trading

Trading today feels messy and overwhelming, but it's also why millions are here. It's not about a lack of talent or data — it's about how scattered the process is and who has the best setup to do it. Traders are bouncing between Twitter, explorers, wallet trackers, and analytics tools, trying to piece everything together. The market moves too fast for that!

The opportunity isn't another tool — it's a network. Something that pulls everything into one place, automates the repetitive work, and makes smarter decisions.

This is where agents come in.

Not just bots that follow rules, but intelligent systems that think like traders. Brains with bodies. Agents that watch, learn, and act on opportunities in real-time.

Here's how I'm thinking about this through the jobs to be done as a trader:

1. Tracking Signals at Scale

If Elon tweets something, traders scramble. They're checking Twitter, analyzing wallet activity, and guessing how the market will react — a memecoin? An NFT? What's to come?

An agent could handle all of this automatically. It would scan accounts, connect tweets to on-chain activity, and highlight the real opportunities like a recently launched pump fun token with one of the words or inferred joke!

2. Decoding Trends and Meta

Markets move on narratives. Right now, it's the "TikTok" meta — tokens tied to quirky trends and viral stories.

The agent would learn these trends by analyzing social sentiment on X or TikTok, transaction flows, and hashtags. It would surface what's emerging before anyone else notices.

3. Hunting and Watching Wallets

Wallet tracking is a grind. Traders manually check PnL, trade timing, and behaviors to find accounts worth following. It takes forever! People always tell me this.

The agent simplifies this. It finds profitable wallets, monitors them, and shares insights without any manual work!

4. Acting on Opportunities

This is where the real value lies. The agent doesn't just give insights — it takes action. A brain with a body.

If tracked wallets move into a token, the agent cross-checks that token with social signals and makes a call for you!

The thinking is simple — if we want to succeed in solving problems using agents, and within trading, then we have to solve annoying jobs for traders. You can imagine this network: we have agent pump for trading tokens, we have a version of agent sherlock for searching social media, and what we need is a copycat to copytrade top performing wallets. The idea is to automate the jobs to be done and let the agent learn from you to get better.

Once we nail those, expand into strategy automation and workflows. The agent takes the grunt work off the trader's plate so they can focus on what matters most.

It's not just another tool — it's a network traders rely on to win.

I keep asking myself: how do we make this feel like an extension of their brain? How do we make sure people see the value? We build. If we can do that, the rest will take care of itself.

---

The Real Problems

What's the problem with crypto?

Humans don't know how to use it.

Why don't humans know how to use it?

It's built for computers to talk to computers — that's what composability means.

---

What's the problem with AI?

Agents don't know how to use the Internet.

Why don't agents know how to use the Internet?

It's built for humans to use.

---

Agents Are Already Here

AI agents already exist today but we take them so much for granted we've forgotten they exist.

- Google is everyone's favorite AI agent — tell Google what you're looking for and it comes back with ten blue links. - Those feeds on Instagram and TikTok you doom scroll on are just AI agents finding content just for you.

But when it's time to take action like buying a toy on TikTok or subscribing to a Substack newsletter, the human takes over.

The Internet today is basically reinforcement learning with human feedback (RLHF) — AI agents figure out what you're looking for and learn based on what actions you take.

Agents are all brains and no body. They know what actions to take but can't take action.

Read Write Act

The internet changed how we do things — we can find anything with a simple search, understand anything with AI, and move value anywhere with crypto. But there's a massive gap between finding what you want and actually doing it.

You're still doing everything manually. Want to grow your crypto? You're clicking buttons all day. Want to purchase something online? You're going on different sites and asking friends for input. Want to automate your digital life? Impossible to connect everything together. We have AI that understands, crypto that transfers value, but they're not working together to make our lives easy.

Imagine your digital life running on autopilot — where your intent for something automatically turns into actions. Just like Google's ten blue links changed how we find anything, AI and crypto give us agents to do everything.

GriffAIn is a network of AI agents that starts by managing your Solana portfolio and expands to automate your entire digital life, from budgeting, spending to shopping and beyond.

We're building the bridge between AI that understands you in natural language and crypto that accesses open data and executes actions for you. We're starting with the most important thing — your money — to earn trust.

The vision? A future where agents don't just find information but take action for you. We want to turn the internet into a read-write-act space, where your digital life works on your terms.

But what about own? Do I still own? Well I don't think users give a shit about owning. It's about access. Having access and being able to do what you'd like with what you have. Maybe ownership is important in some places, but I've taken a spin on Read Write Own. Try Read Write Act. Write and Act are different, just to start out. Why? Well, in this case, write refers to publishing your intentions, questions, and needs to the agent. Act is when the agent reasons and decides how to use that information. In this world, we are still early in the write phase.

Data and Personalization

Which brings me to data. If you've been around long enough, you know this is what I care about. Using data to build personalized experiences for users. But what's personalization? It's data + context + action. Action can be done when the intention is understood and there's enough data to make a decision. For agents, they need data and guidance. We thought the problem was bringing interesting data onchain, thus we needed consumer experiences to use digital assets.

No — consumers need to make their own experiences that are interesting. How? By having an easy way to turn their intent into action.