J A K E simba

"💙 Army "

[J A K E simba @JakeBnb]

The gate opening is one way to make something move.

Giving it a reason to move is another.

That distinction matters in financial systems too. I'm curious to see how 2Factor applies this idea to BTC and long-term exposure.

[2FactorFinance @2FactorFinance]

A bull runs when someone opens the gate.

A buffalo runs because it wants to.

https://t.co/Vjn2paUGDQ

[J A K E simba @JakeBnb]

The leaderboard being a race between collectives changes the dynamic.

Your own activity matters, but the people around you matter too.

With only 20 referral slots, building the right collective may be more important than simply being early.

[2FactorFinance @2FactorFinance]

1000 people, 10 seats, 1 BTC 🔥

Season 1 has a feature most of them haven't noticed.

What you build now does not stop counting when the season ends. Marks carry forward into Season 2.

But only 10 leave with Bitcoin on top.

https://t.co/HqANmFdXVe

1/5

[2FactorFinance @2FactorFinance]

A bull runs when someone opens the gate.

A buffalo runs because it wants to.

https://t.co/Vjn2paUGDQ

[J A K E simba @JakeBnb]

@unna_gi Most AI conversations optimize for getting the answer quickly. I think that’s backwards for learning. Sometimes the better agent is the one that refuses to close the loop too early and leaves you with a sharper next question.

[J A K E simba @JakeBnb]

@unna_gi Since you frame Mochi as part of the learning process, I’d be curious about how it tracks an unresolved question across turns. Is the conversation state closer to a summary, a dependency graph, or something else?

[鰻 @unna_gi]

I don’t mind if I forget one answer.

What’s frustrating is when I have to explain everything all over again just to move on to the next question.

I once closed my laptop after feeling like I’d finally grasped a concept. The next day, I still remembered the conclusion, but I’d forgotten which question had led me there.

In the end, I opened everything up again just to find the point where I’d truly understood it.

One answer leads to another question. If something still doesn’t make sense, I break it down again, then stop once I start to see the pattern.

Tomorrow I want to continue.

Not repeat.

That’s why I see Mochi from @GenLayer as more than just a bot that can answer questions.

I don’t need an AI that remembers everything. I need an AI that knows what I don’t need to repeat.

For me, memory is like leaving a thread where I stopped.

I don’t need to unravel everything from the beginning every time I come back. I just need to follow the last thread and know where to pick up.

It might sound trivial.

But to understand something like GenLayer, I actually consider the ability to continue a conversation as part of its learning process.

The bot answers my questions. Mochi crafts the next question so it still ties into the previous one.

That’s why I want to use Mochi as a conversation I can set aside and then pick up later.

One concept today.
Take it one step further tomorrow.

What’s one GenLayer concept you understood yesterday but feels unfamiliar again today? Bring it to Mochi, then come back tomorrow. Which part will you eventually be able to move forward with without having to repeat it?

[J A K E simba @JakeBnb]

@unna_gi How could the majority’s choice fail?’ is powerful. I’d make that adversarial: can Mochi generate the strongest plausible case against the leading option before the vote closes?

[J A K E simba @JakeBnb]

@unna_gi Your 800-of-10,000 example made me think about a different risk: what happens when the minority understands the trade-off better than the majority?

[J A K E simba @JakeBnb]

@unna_gi Suppose voters see two options with similar quality. What missing context matters most?

[J A K E simba @JakeBnb]

@unna_gi The majority can follow appearance or price. How would you expose "which option looks more appealing" first?

[鰻 @unna_gi]

In the Mochi scenario, the cheapest material wins. And that’s exactly the problem.

It’s not that cheap materials are always bad, but because the majority of voters don’t truly understand the trade-offs they’re making.

I discovered this while testing an idea.

I wanted to give NFT owners the right to vote on the material for limited-edition NFTs.

So I asked Mochi on Telegram: https://t.co/ebptrerou4

“What’s the riskiest assumption if owners don’t understand the supply chain and production costs?”

I initially thought the problem was simply about choosing the wrong material.

It turns out my assumption was much more fundamental.

I assumed that people with voting rights also had enough information to exercise those rights wisely.

In the example Mochi gave, out of 10,000 NFT owners, only 800 truly understand the trade-offs between cost, quality, and material availability.

The rest might vote based on which option looks more appealing, is cheaper, or simply because they’re following the majority.

At that point, my perspective on voting changed.

The majority can reach an agreement without truly understanding what they’re agreeing to.

It feels like giving all passengers the right to determine the ship’s course, but not giving them a map.

My hot take: voting rights don’t automatically lead to good decisions. If the context is poor, increasing the number of voters doesn’t always make for smarter decisions.

That’s why there’s one feature I’d like to see in Mochi: “Decision Audit” Mode

Before I open the vote, Mochi must read the proposal and the options that will be presented to the community, then stop me if there’s any important context that’s missing.

For example:

“What information do voters not yet know?”

“What trade-offs are they most likely to overlook?”

“If the majority chooses this option, how could that decision fail?”

I don’t need Mochi to just be a smarter vote counter.

I want Mochi to audit the quality of the information behind each vote before it’s cast.

@GenLayer, this is exactly the kind of feature I really want to try.

In your opinion, which is more dangerous: a voter who makes the wrong choice, or a voter who doesn’t know that their choice could be wrong?

[J A K E simba @JakeBnb]

@unna_gi Hot take: an agent that asks for human confirmation at the right moment is more impressive than one that completes everything autonomously. Knowing when not to act is an underrated form of intelligence.

[鰻 @unna_gi]

I once saw a demo that looked perfect, only to fall apart because I changed just one small assumption.

Since then, I’ve been more interested in one question:

What happens when an agent no longer operates under ideal conditions?

If you’re building an agent that needs to gather information, choose options, run automations, perform transactions, or make decisions without us monitoring every step, Agent Tank is definitely worth a try.

It’s not just for AI engineers. If your project involves software performing tasks that typically still require a human to choose the next step, I think you have a good reason to join.

Agent Tank is a 14 day hackathon to build solutions for the Agentic Economy.

Today, September 3, the build phase begins and ends on September 17.

That means you have 14 days to pick a problem, build a solution, and then see if your product holds up when things start to go wrong.

And up for grabs is 5% of all @GenLayer Points.

But if I were to participate, Points wouldn’t be the main reason.

I want to find out if an idea that seems reasonable in the repo also makes sense when faced with the real world.

Hot take: If your agent only seems smart on the happy path, you haven’t really tested it yet.

So pick one problem you think the agent should be able to solve.

Tell me about one thing that didn’t go according to plan.

I want to compare not just what you’re promoting, but how much your initial solution had to change before it finally worked.

👉 https://t.co/9wxE5nrojo

[J A K E simba @JakeBnb]

@unna_gi You mentioned agents gathering information and then making decisions independently. How would you think about stale context between those two stages, especially when the final action changes an external system?

[鰻 @unna_gi]

I thought MEXC would just be a place I’d pass through. As it turns out, that one introduction led to many connections that have made this journey feel so much warmer.

Happy International Friendship Day.

[J A K E simba @JakeBnb]

@unna_gi "isolated islands" is the perfect metaphor. But here's what I keep coming back to. Even if @InternetCourt connects them all, each island still sets its own rules. Who decides which protocol's rules take priority when two agents disagree on escrow terms?

[鰻 @unna_gi]

I once bought three apps for work.

One for research.
One for writing.
One for design.

The problem is, they all run independently. I have to act as the messenger. Copy the results from one app, move them to another, and repeat the process. It’s not that the tools are bad. They just never really connect.

And it turns out, AI agents in crypto are facing the same problem.

Now there are protocols for identity. There are protocols for negotiation. There are protocols for payments. They’re all great at their respective tasks.
But almost all of them only handle the happy path. They stop working when transactions no longer go smoothly.

If two agents disagree, execution fails, or a dispute arises, who connects everything? Almost none.

In my opinion, this is actually the biggest bottleneck in the AI agent economy. It’s not because we lack protocols, but because most protocols are built like isolated islands.

That’s why I view @InternetCourt from a different angle.
For me, this isn’t a new layer. It’s an interface that connects all the existing layers into a single, seamless workflow from identity, negotiation, payment, escrow, and execution all the way to dispute resolution. Everything can be executed by two agents using a language that’s easy to understand.

My take: the smartest protocol isn’t necessarily the most useful one. The one that will win out is the protocol that allows all systems to work as a single unit.

In your opinion, does the future of AI agents require more new protocols, or does it actually need a connector like this? Explain your reasoning

[鰻 @unna_gi]

I once interviewed at a tech company. The position needed someone adaptable. I wrote flexible on my CV. The interviewer said we need someone adaptable, not flexible.

I didn't get the job. Even though it means the same thing. I tried to explain. The interviewer said it's different. I said how so? They said adaptable is about mindset. flexible is about schedule. But in the dictionary, both mean: able to change.

And that made me think: if two words with the same meaning can make me fail an interview, what about AI that has to decide something more important?

One AI says safe.
Another says secure.
Another says no risk.

Different words. But the same meaning. And the most important thing: what's checked isn't the words. It's the meaning.

Same with the AI I found at @GenLayer. When there's a dispute, each AI answers independently. What's checked isn't the words. It's the meaning. If one is wrong, more AIs come in to review.

Because if 5 AIs with different models all say "safe," that's not a coincidence. That's truth that survived from 5 different perspectives.

One voice says safe. Five voices that never looked at each other say safe. Which one makes you confident in your choice?

[J A K E simba @JakeBnb]

@unna_gi Hot take: meaning-based verification is actually less trustworthy than syntax matching. If five AIs use different words, how do you know they didn't just get lucky with different errors? Identical wording would at least prove they processed the same information the same way.

[J A K E simba @JakeBnb]

@unna_gi I did the same thing for 2 years with DeFi. Copied yield strategies from Discord, pretended I understood impermanent loss. The worst part wasn't losing money. It was when someone asked me to explain LP tokens at a meetup and I froze. That shame you're talking about?