gane5h

The Latency Crisis is Real

Over a year ago, we kicked off an internal moonshot initiative at @Covalent_HQ to reimagine blockchain infrastructure from first principles.

We called it our Manhattan Project - like the wartime R&D effort that brought together the brightest minds to solve an urgent problem.

But ours isn't about physics. It's about blockchain data.

Why?

Because something has fundamentally changed.

Chains now produce blocks in under a second.

Yet most apps still rely on APIs that update every 5–30 seconds.

Legacy indexers. Polling loops. Stale snapshots.

We call this the Latency Crisis - and it’s not just a developer inconvenience.

It’s a systemic failure at the foundation of Web3.

↠ Oracles are outdated by the time they post.
↠ Block explorers lag behind actual chain state.
↠ Dashboards feel broken.
↠ Bots miss profitable trades.
↠ Wallets display stale balances and confuse users.

This isn’t an edge case. It’s the default behavior across the ecosystem.

After 5+ years building Covalent, I can say this with confidence:

The architecture that got us here won’t take us further.

The problem has outgrown the tooling.

So we stopped patching legacy code. Stopping recommending short-term hacks to our 40K+ devs on the @goldrushdev API.

We rebuilt the stack - from raw block ingestion to sub-second data delivery. End to end.

Over the next couple of days, I’ll share what we’ve learned, what we’ve built, and why the next wave of blockchain apps won’t be possible without solving latency at the root.

I'm taking a different direction on X instead of the usually yapping and KOL shilling. I'm going deep.

This is Day 1.

#SpeedRunTheChain

🧵 Follow along.

We are shopping around for a smart account / wallet stack for an upcoming project and I'm a bit surprised at how the current market is laid out:

Seems like @safe, followed by @Alchemy, @zerodev_app , Pimlico, Biconomy lead in historical deployments - but if you study growth rates, it's actually Zerodev and Pimlico that's growing over the last 3 months.

When it comes to actual transactions from these deployments - it's seems like Zerodev/Pimlico holding steady.

What makes Zerodev stand out here? Not too familiar with this market.

As an EVM builder, I've started to notice something uncomfortable - the new wave of devs on these faster blockchains like Solana, MegaETH, Monad, Sonic amongst others are already operating in a different paradigm - streaming-first, low-latency and built for real-time UX.

Meanwhile, many of us in the Ethereum world are still stuck thinking in blocks and polling intervals. We're trying to graft speed onto systems that were never designed for it, and it shows.

As chains like @base move to a sub-second blocktime, EVM devs will need to level up or be left behind.

Here are four examples that hit home for me:

1️⃣ Wallet price feeds
If there's a swap every block, then prices change every block. A polling strategy just doesn't work to reflect the latest prices.

2️⃣ Onchain gaming
Ethereum games are mostly off-chain simulations with onchain settlement. Newer games will stream onchain events (like user moves, enemy actions, or loot drops) with no perceptible lag. EVM games today are mostly turn-based games.

3️⃣ AI Agents
While building out our AI Agent SDK, we wanted to build an AI trading copilot that ingests live market data every 400ms, detects patterns like spoofing or whale activity, streams real-time insights directly into the UI - suggesting pre-filled orders or adjusting to risk. Let's just say that this was hard to build when the data is delayed, batched and had to be polled out of RPCs and indexers.

4️⃣ Composable data streams
The data infrastructure of the future will be streaming state changes across protocols - like syncing interest rate changes directly into yield aggregators. In the EVM world, we stitch APIs together and pray for consistency.

This isn't just theoretical, this is a problem today with apps on testnets for Monad and MegaETH. We see these issues first-hand speaking to devs working with @Covalent_HQ.

We are building a solution - stay tuned! 👀

Every builder will need to upskill to prepare for the new world of fast, high-throughput chains. If blocks are streaming out at ~250ms, here's your cheat sheet for latency:

🔄 L1 cache: 0.5 ns
🔄 RAM: 100 ns
🔄 SSD read: 100 µs
🔄 Disk seek: 10 ms

Within a data center in a single availability zone:

↔️ 5ms

Global Round-Trip-Time from the West coast:

↔️ NY: 70 ms
↔️ London: 160 ms
↔️ Tokyo: 120 ms
↔️ Singapore: 200 ms
↔️ Dubai: 280 ms

💡 In a subsecond chain world, every ms counts. Speed of light is your new bottleneck. Optimize accordingly.

TIL

Pectra

the name of the latest Ethereum upgrade, is a portmanteau of:

• Prague - the name for the execution layer track
• Electra - the name for the consensus layer track

Together: Pectra = Prague + Electra

Pectra represents the joint coordination of EL and CL changes in a single hard fork - continuing Ethereum’s post-merge upgrade cadence.