It just sits next to the node, codes the gossip with RLNC,
and makes blocks travel ~6x faster.
150ms average. Months of real testnet data.
No client changes required.
Sometimes the biggest unlocks don’t look flashy.
They just remove the silent ceiling everyone was hitting.
While everyone was arguing about L2s and blobs,
Ethereum’s networking layer quietly became the bottleneck.
Blocks got heavier.
Validators got more global
Propagation latency started capping everyone’s performance.
Optimum’s mump2p doesn’t try to replace everything
@get_optimum
mump2p is the upgrade that Ethereum's networking layer has been waiting for, and don't just take our word for it...
Top validators like @Nodemonste have experienced its speed firsthand over the course of our testnet.
plus a conversation with someone from Boston Dynamics.
Next:
more products through Q3 and Q4,
new ways to earn,
egocentric data capture.
Catch them at IROS Pittsburgh, Booth 1122, end of September.
Quiet work. Real progress.
@PrismaXai
Just finished PrismaX’s Office Hours thread.
July and August were about locking the infrastructure.
Collection, Verify Quality, and the validator network now run as one system.
Cornerstone customers are already in.
They’re not flooding the market with junk data.
@PrismaXai
Our monthly Office Hours ran last Friday. Tons of questions from the community, most of them about the same thing: what PrismaX is focused on and where the platform goes next.
Here is what we told them 🧵
Consistency second.
Volume last.
The First 100 was the starting group, not a closed club.
They’ve already shipped a lot:
Verify Quality is open,
robot fleet expanded to Piper, TOK2 and YAM,
Teleop Cup launched,
robotics data standards series started,
@PrismaXai’s Verify Quality isn’t a feature — it’s the necessary consensus layer that turns raw teleop into usable training signal. Without it, more data just means more noise
The bottleneck was never parameter count. It was trust in the data!
You can’t scale a trillion-parameter robot model the way we scaled LLMs.
Text models drown in clean, reversible data. Robots live in irreversible physics. One bad trajectory permanently poisons the prior.
That’s why volume alone is a trap
@PrismaXai
Underlying infra iterations fly under the radar.
Team quote: Can’t believe nobody has figured these out yet.
Numbers don’t lie.
Keep this infra project on your watchlist.
#Optimum#DePIN#Eth