LiquidMetal AI

A GPU Cluster Is a Financial Instrument with Cooling Fans.
Every 5% yield gap on a 1GW campus is $15B/year in unrealized revenue. We help improve that yield.

[LiquidMetal AI @LiquidMetalAI]

Your GPU Cluster Is a Financial Instrument with Cooling Fans. Every 5% yield gap on a 1GW campus is $15B/year in unrealized revenue. Most operators don't know how much they're losing.

Find your yield gap.

We'll quantify it in dollars ($/GW)

[LiquidMetal AI @LiquidMetalAI]

Read more :
https://t.co/uqp8YY8q0T

[LiquidMetal AI @LiquidMetalAI]

Being AI Native - Part Two : Design-First Development: What It Actually Looks Like

Translating from theory to practice can be a challenge. To help others clear this hurdle, we’re going to demonstrate Design-First development by first creating a CI scheduler, and then fixing a bug in it.

Read more 👇

[LiquidMetal AI @LiquidMetalAI]

@pepsi #stop

[Netlify @Netlify]

Hackathon winners drop! 🏆

Congrats to everyone who built in the @LiquidMetalAI hackathon collab.

Here are the Grand Prize projects:

✨ Hakivo: https://t.co/Rc1MIh3aj9

✨ AI Compliance Automation: https://t.co/mCRWoQmCMC

✨ Project Sentinel: https://t.co/u6w4MURLPu

[Devpost @devpost]

Let's celebrate the winning projects in the @LiquidMetalAI AI Champion Ship! 🚢💨🏆

This mission? Pushing @LiquidMetalAI beyond the ordinary.

The results? Elite execution across the board—from the "Ghost Protocol" defense in Project Sentinel to high-impact literacy tools like TeddyTales.

Check out the full fleet of champions:
🔗https://t.co/YmYl6SNCzM

[LiquidMetal AI @LiquidMetalAI]

@csaba_kissi Netlify.

[LiquidMetal AI @LiquidMetalAI]

@samgoodwin89 on Fridays...

[LiquidMetal AI @LiquidMetalAI]

@gunta85 @rauchg @vercel @alchemy_run Oh, you aren't the only weird one. Ha. Self-driving infra is everything you said Gunther, plus branching/versioning of apps/data (think Neon-like branching of apps/data) + auto full-fidelity trace + agent swarm communications (aka Annotations), etc.

https://t.co/qrJHgUdwuZ

[LiquidMetal AI @LiquidMetalAI]

Agreed. Full-fidelity trace and logs must be built into everything you can create, build, and deploy. We do this on our Raindrop AI infra platform for backends. You also need versioning of apps and data too otherwise your AGENT is fixing in prod vs just tagging a section of prod to be your canary and rolling forward or rolling back as needed.
https://t.co/KVASd4rC0X

[LiquidMetal AI @LiquidMetalAI]

@levie Yes, but it is not just with coding repos but also with data, repositories, files, buckets, tables, databases.. We call it "Annotations" and every Agentic platform needs to support it or it will never get up to speed.

https://t.co/s4vtWe3Ifh

[LiquidMetal AI @LiquidMetalAI]

Try it here: https://t.co/vmC3P9NhZu

[LiquidMetal AI @LiquidMetalAI]

58 modules. Zero infra tickets.
Most stacks can’t say that.

As systems grow, every new agent, job, or workflow usually comes with:
new services, new dashboards, new on-call surface area.

You end up paying a “platform tax” just to keep the thing running.

Raindrop is built to make that trade-off disappear.

In our AI ChampionSHIP program, one team decomposed their system into 58 modules because that’s what the product needed, not what the infra allowed.

They described those modules in Raindrop manifest.
The platform handled:
→ packaging and deployment
→ routing between modules
→ autoscaling and health
→ observability out of the box

No custom Kubernetes setup.
No separate deployment pipelines.
No infra management on their side.

That’s the point of Raindrop:
design the architecture your product deserves,
without turning your team into an infrastructure org.

[LiquidMetal AI @LiquidMetalAI]

Meet AuditGuardX, #2 Grand Prize Winner in AI ChampionSHIP

It turns enterprise compliance into an AI-native workflow.

It reads policies across dozens of frameworks,
spots gaps, and regenerates compliant documents

All that in minutes instead of months
so teams can ship and stay audit-ready at the same time.

Under the hood, AuditGuardX runs as a serverless on Raindrop + Vultr: Raindrop’s SmartMemory, SmartBuckets, and SmartInference orchestrate document analysis, voice chat, and semantic search.

AuditGuardX is a sharp example of what happens when one builder treats Raindrop as the backend—and focuses all their energy on real enterprise impact.

Congrats @patsinfotech

https://t.co/IWRPrzJcPG

[LiquidMetal AI @LiquidMetalAI]

@janwilmake https://t.co/gESW0yYV1D - MCP support, build an API or agent and deploy globally. AI infused building blocks like RAG obj store or NLP SQL. A lot of CF under the hood except inference.

[LiquidMetal AI @LiquidMetalAI]

@Bhavani_00007 hosting what?

[LiquidMetal AI @LiquidMetalAI]

Hakivo Demo Video : https://t.co/gLq9RtYSO2

[LiquidMetal AI @LiquidMetalAI]

Meet the first Grand Prize winner of the AI ChampionSHIP: Hakivo.
Hakivo turns Congress into an AI product.

It tracks bills, summarizes them in plain language, and delivers NPR-style audio briefings so anyone can actually follow what’s happening in government.

Under the hood, it runs 58 Raindrop modules:
agents, services, tasks, queues;
without managing a single piece of infrastructure.

SmartBuckets power semantic search over thousands of bills.

SmartMemory keeps long-running conversations grounded.

Hakivo is our kind of winner: ambitious, real-world, and built to go beyond the demo.

Congrats @tarikjmoody

Learn more - Link below !!

@Vultr @cerebras @elevenlabs @CloudflareDev @stripe @WorkOS are the best sponsors of The AI ChampionSHIP.

[LiquidMetal AI @LiquidMetalAI]

2968 Participants.
3 Grand Prizes:

Congrats to:

#1: Tarik Moody - Hakivo - @tarikjmoody
#2: Patrick Ejelle-Ndille - AI Compliance Automation - @broadcomms
#3: Saicharan Ramineni - Project Sentinel - https://t.co/oDubtr2kuU

Check out the full gallery of all 15 winners selected for 6 categories:

https://t.co/YNjKWMyF2U

[LiquidMetal AI @LiquidMetalAI]

You shouldn’t be reinventing conversational memory.

In the AI ChampionSHIP program, one team took a different route.

They needed an agent that stayed context-aware across sessions. Instead of building custom storage and retrieval for conversation history, they defined a SmartMemory in their Raindrop manifest and plugged it into the agent.

Raindrop handled:
→storing and indexing the right parts of every conversation
→retrieving relevant context on the next interaction
→keeping state consistent across long-lived sessions

No bespoke “memory service.”
No manual history management or one-off databases to keep alive.

SmartMemory turns cross-session context
from an infra project into a configuration decision.

For builders and engineers, that means you spend less time worrying about how to remember, and more time deciding what your agent should remember to actually be useful.

Try it out: https://t.co/rC4lanpwtu