Earthquake in Granada (M5.2), as seen from my seismograph. The planet never stops moving.
Guillermo Ruiz
@iaasgeek.bsky.social
Sr. Specialist Solutions Architect, Efficient Compute @AWScloud - vBeard, Arm & Grafana Champion, ex-Hashicorp Ambassador, vExpert, CiscoChampion/ Ötillö dreamer / *Tweets my own*
Anthropic announced it will watermark anything it generates (even if you’re just passing your own text for a grammar check). This is how I see things going once every AI model starts doing the same. 😂
What happens when a VM running its own hypervisor finds a way to reach the host? I spent some time digging into Zapscape (CVE-2026-64561), a KVM vulnerability that turns a Shadow MMU bug into a VM escape. Here my notes... www.linkedin.com/pulse/zapsca...
Zapscape: When a Virtual Machine Stops Behaving Like a Virtual Machine
Virtualization is one of the foundational technologies behind modern cloud computing. Whether you're deploying applications on-prem, running OpenStack, or launching VMs in the cloud, virtualization re...
linkedin.com
I used my phone’s barometer to reconstruct a flight’s cabin pressure profile. Sometimes the best observability tool is already in your pocket. www.linkedin.com/pulse/readin...
Reading a Flight Through Cabin Pressure
This project started as a personal research question. After my medical incident, I became curious about something many passengers never think about: what altitude is my body actually experiencing up t...
linkedin.com
Nitro Enclaves are secure by design. Prowler helps verify they’re configured as intended with 11 new security checks. prowler.com/blog/your-ll...
Your LLM Runs in a Nitro Enclave. Who’s Checking the Enclave?
11 new Prowler checks bring the first CSPM posture coverage to AWS Nitro Enclaves — from KMS attestation policies to detecting enclaves left in debug mode, without ever touching the instance.
prowler.com
When I was a kid my dad told me that one day California will break away from the continent and slide into the ocean. Of course, that's not how plate tectonics works, but when you are a kid, that image stays with you. And that story turned into curiosity… www.linkedin.com/pulse/next-s...
Is the next San Andreas earthquake waiting beneath the Salton Sea?
Lately, it feels like earthquakes have been appearing in the news more often than usual. With all the news on media it brought me back to one of the things that has fascinated (or traumatized) me sinc...
linkedin.com
The recent Mw 7.5 earthquake in Venezuela is a reminder that we still can't predict earthquakes. The only thing we can do is detect them almost immediately and use the first seismic waves to warn people before the strongest shaking arrives. www.linkedin.com/pulse/chasin...
Chasing Seismic Waves: From a Mw 7.5 Earthquake to Earthquake Early Warning
The recent Mw 7.5 earthquake in Venezuela is a reminder that we still can't predict earthquakes.
linkedin.com
🚨 AWS just enabled nested virtualization on standard EC2 instances. No more bare metal needed to run microVMs, Kata, Firecracker, or emulator-heavy CI. Simpler sandboxes, lower cost, more flexibility! github.com/aws/aws-sdk-...
Release 2026-02-12 · aws/aws-sdk-go-v2@3dca5e4
github.com
Everyone’s talking about EU cloud sovereignty and replacing US tech. Real question: could a fully sovereign EU cloud even exist across the whole stack, from hardware to AI and control planes? www.linkedin.com/pulse/contro...
Control Beats Origin: Rethinking EU Sovereignty in the Cloud Era
Lately I’ve been reading a lot of people complaining about EU Sovereignty Clouds and hyperscalers deploying “sovereign regions” in Europe. The latest announcements from the European Commission seem to...
linkedin.com
After sharing the ZeroLeaks report (big shout to Lucas Valbuena), I got few DMs asking about OpenClaw, AWS, and whether it can actually be secured. Here’s my take on it...and how I’d approach it. www.linkedin.com/pulse/harden...
Hardening OpenClaw on AWS? A paranoid engineer’s short guide
After sharing a short post with the ZeroLeaks report, I started getting few DMs about OpenClaw and the risks of deploying it. Most questions boiled down to two things: Can it run on AWS? And can it ac...
linkedin.com
Ever feel busy all day but still struggle to get real work done? We’ve spent years optimizing the cloud: cost, performance, efficiency. But we rarely apply the same rigor to how we actually work. This article is about why that happens… www.linkedin.com/pulse/effici...
The Efficiency We Forgot About
Most discussions about efficiency in our day-to-day starts in the same place: costs, performance, utilization, cloud bills. Not always in that order, but close enough.
linkedin.com
2025 was a wild ride I never saw coming. Now it’s time for Stop-endo: stopping the setbacks and letting the endo-stoppable energy begin. Happy New Year to everyone!
Starting a new series on Embedded AI and efficient compute. Before we talk satellites or Amazon Leo, we need to understand what really happens inside attention. So I ran the experiments on a MacBook. Turns out attention is way sparser than we think. www.linkedin.com/pulse/why-ai...
Efficient AI at the Edge, Part 1: Why AI Models Get Faster When You Look Closer
A couple of weeks ago, Karol Piątek shared an article on sparse logarithmic attention. It stuck in my mind, and over the past few days I’ve been exploring a simple question: Is attention really sparse...
linkedin.com
Are you at re:Invent? I’m not there physically this year, but you can still grab some limited-edition Buildercards... there are a few surprises waiting for you. Just swing by the #Buildercards space in the Caesars Forum and make sure you don’t miss yours! #AWSreInvent
My friend Christian is swimming 80 km from Tenerife to La Gomera and back. Not for glory, but to support disadvantaged families in Nepal with the Vicente Ferrer Foundation. Every stroke is hope. Proud of his courage, purpose and heart. Follow live: share.garmin.com/BrazadasSoli...
MetalNES emulates the Nintendo NES transistor-by-transistor. Same dev who built the fastest NES emulator in the 90s… came back to build the slowest on purpose. This is not nostalgia. It’s hardware archaeology. (And yes, we show how to run it.) www.linkedin.com/pulse/metaln...
MetalNES: The Most Accurate NES Emulator Ever Built (And Why Your CPU Still Cries Running It)
Most emulators try to pretend to be the original console well enough to get the job done. They skip the messy details, the physics and the silicon.
linkedin.com
For decades, Python said: “Sure, I’ll run all your threads... one at a time.” 😅 Now, with 3.14t (no-GIL), it finally means it. Parallel threads. Multi-core scaling. Real performance gains. www.linkedin.com/pulse/python...
Python 3.14 finally speaks fluent multi-core
For over 30 years, Python developers have lived with the same frustration: no matter how many threads you spawn, only one CPU core is really doing the work. This was due to something deep within the i...
linkedin.com
A sad day. Jane Goodall shaped my passion for nature, along with Jacques Cousteau. In 1987 she even helped my father’s expedition in DRC, investing all she earned back into primate conservation. One of a kind. Impossible not to love her. RIP Jane.
Qualcomm used to be “the Snapdragon company", where phones and mobile dominance were its flagship. But the story has changed. From clusters in the cloud to glasses on your face, Qualcomm is trying to build a continuum of compute. www.linkedin.com/pulse/qualco...
Qualcomm’s Next Act: From Mobile Powerhouse to Data Center Challenger
Qualcomm ruled the smartphone chip market for years with Snapdragon. But the company has been quietly evolving, from expanding its CPU design ambitions to preparing for a future in data center GPUs an...
linkedin.com
Cloud isn’t just racks and APIs anymore. With AI, it comes down to chips, memory, and networks. I put together a blog on how AWS Trainium2 impacts the game, from scale-up fabrics to memory bandwidth that actually keeps the math engines busy. www.linkedin.com/pulse/silico...
Silicon at Amazon, why it matters, and what Trainium2 really is
Cloud used to be about racks and APIs. In the AI era it's about chips, memory, and networks.
linkedin.com
Last night my eldest had some friends over. When I got home, I found them all deep into the Oculus 2 and 3… future metaverse engineers in training 😂
Reading about Oxide’s $100M Series B took me back to the early vBlock days, when we tried to bring cloud-like experiences on-prem. Oxide feels like the version we always wanted. Not just servers, but a cloud rack with an engineer’s soul. Worth watching. www.linkedin.com/pulse/reinve...
Reinventing the Rack: Why Oxide Might Be the Most Important Infrastructure Startup You Haven’t Looked At (Yet)
More than a decade ago, when AWS was still something we debated in boardrooms, I was on the other side of the fence, helping enterprises build their own private clouds with VCE vBlocks. These were mas...
linkedin.com
Before hyperscale cloud ruled everything, the story was different. I lived part of it, designing and deploying enterprise cloud regions at CSC. In this blog, I revisit the lessons from those early days, with anecdotes and headaches from the field. www.linkedin.com/pulse/tracin...
Tracing the Roots of Cloud Computing
Today, cloud is almost synonymous with AWS. But in the early 2000s, things looked very different.
linkedin.com
Programming languages don’t just shape applications - they shape infrastructure revolutions. From Ruby on Rails to Rust, here is how they’ve helped in the cloud era and now in the AI space. www.linkedin.com/pulse/from-r...
From Ruby to Rust: How Programming Languages Spark Infrastructure Revolutions
Some programming languages don’t just help build apps - they define entire eras of infrastructure. In the mid‑2000s, that language was Ruby, driven by Ruby on Rails, which helped accelerate the adopti...
linkedin.com
Not all AI needs a GPU. While the spotlight’s on giant models + GPU farms, real-world AI often runs just fine on CPUs. Docs generators, copilots, lightweight inference on Graviton… efficient, affordable, and overlooked. www.linkedin.com/pulse/all-ai...
Not All AI Needs a GPU
Every headline in AI seems to orbit the same gravity well: larger models, faster tokens, and ever more powerful GPUs. From xAI’s Colossus clusters to OpenAI’s 4.
linkedin.com
Tech stocks dragged the Nasdaq down today — Nvidia, AMD and Marvell all hit hard. But Intel surprised everyone, soaring nearly 7% thanks to SoftBank’s $2B bet and possible U.S. government backing. Sometimes the market flips the script.
No more fixed GPU assignments. With Dynamic Resource Allocation on #AmazonEKS, your AI workloads get the GPU power they need, when they need it — boosting efficiency & cutting costs. www.linkedin.com/pulse/squeez... #Kubernetes #EKS #GPU #AI #CloudComputing
Squeezing Every FLOP: The Untold Story of AI GPU Utilization
Building AI superinfrastructure isn’t just about raw scale — it’s about turning every GPU cycle into useful work. In our ongoing series on next‑gen AI infrastructure, we started with how AWS engineers...
linkedin.com