I first realised that our users were no longer people when the support issues stopped sounding human.
We had launched a modest tool that allowed users to set up automated strategies. You defined rules, and bot-controlled wallets executed them on-chain based on market conditions and portfolio data. Nothing groundbreaking.
Early on, the users were easy to imagine: traders, DeFi power users, a few funds. Their questions felt familiar and emotional. Why did this trade behave like that? Why didn’t this trigger execute? Can you support this asset?
Then the pattern shifted.
New accounts appeared without Telegram handles, without personality, without downtime. Their behaviour was exact, continuous, and mechanical. And the questions we started getting weren’t from them—but from other developers.
“Agent cluster X behaves incorrectly when prices diverge.” “Our treasury agent is reacting too aggressively to news.” That was the moment it clicked.
We weren’t building software for people anymore. We were building infrastructure for autonomous agents. And agents don’t have intuition—only inputs.
If the inputs are wrong, the outcome is wrong. No nuance. No hesitation.
That’s when APRO and AT stopped being interesting experiments and became essential infrastructure.
Once you have dozens or hundreds of agents acting across chains, the old approach—pulling a price from one source and hoping it’s fine—breaks immediately. Humans can ignore a weird candle. An agent will panic instantly if its data says the world just ended.
We needed a buffer between chaotic markets and hyper-reactive code. Something that aggregates, questions, filters, and only then delivers information.
That role was filled by APRO.
We started by routing all core market data through APRO: spot prices, indices, reference rates. Instead of each agent relying on different APIs or chain views, they all consumed the same curated signals.
That alone eliminated a huge category of failures.
Agents stopped responding to isolated wicks on illiquid venues. They stopped disagreeing about what “reality” looked like at any given moment. Their decisions were still good or bad—but at least they were operating from the same shared understanding.
Then we went further.
Structured product agents began using composite feeds that included volatility and funding data. Agents managing tokenised real-world assets relied on APRO to confirm off-chain updates tied to calendars and disclosures. We even experimented with agents that reacted to text-based events using APRO-processed signals derived from documents and news.
In every case, APRO functioned as a translation layer between messy external information and the rigid world of smart contracts.
AT became important once we recognised how foundational this layer had become.
The real single point of failure in our system wasn’t contracts or UI—it was our perception of reality. If APRO failed, drifted, or became misaligned, the impact wouldn’t hit one user. It would ripple across hundreds of agents simultaneously, affecting user funds, treasuries, and partner systems.
That’s not a dependency you ignore.
So we chose to hold AT in our treasury and encouraged our community to understand why.
Holding AT isn’t a magic shield. But it does two critical things.
First, it gives you influence over how APRO evolves. If you believe certain markets require more conservative handling, you can back that view. If you want better coverage for a chain or data type, your voice carries weight.
Second, it aligns incentives. If APRO succeeds as the data backbone for agent-driven systems, AT holders benefit. If it fails, they feel it. That alignment reflects reality instead of denying it.
This mindset reshaped how we build.
Risk is no longer just about contracts. It’s layered. Agent logic. Contract architecture. Treasury policy. And beneath all of it, the data layer. APRO and AT sit there in bold in our internal diagrams.
If that layer is wrong, nothing above it matters.
That framing drives real decisions. Before deploying on a new chain, we check whether APRO’s coverage is mature enough. When designing new agents, we ask whether their inputs can be cleanly expressed as APRO feeds. When discussing treasury strategy, we evaluate how much AT exposure makes sense given our dependence.
These aren’t theoretical conversations. They determine whether we wake up to a stable system—or a coordinated failure caused by bad upstream data.
On a personal level, this changed my own portfolio as well.
I don’t hold AT just because we use APRO. I hold it because I believe the future of Web3 is dominated by non-human actors—and those actors are only as rational as the data they consume.
More bots. More agents. More DAOs with automated treasuries. More real-world systems mirrored on-chain. All of it means one thing:
We’re giving control to code that cannot sanity-check the world the way humans can.
Either we surround that code with data networks that are incentivised to be careful—or we accept massive value loss due to glitches, thin liquidity, and sloppy feeds.
APRO and AT represent my belief that we’ll choose the careful path.
They embody a vision of growth that isn’t just faster or bigger, but more honest about where facts come from.
That’s a vision I believe in. I build on it. I hold it.
And every time one of our agents behaves calmly in the middle of chaos because APRO filtered the noise, I’m a little more confident that this is the right side of the bet.
