Every day, somebody's architecture bills them for standing still. We read the news, find the wreckage, and rebuild it flat.
Somebody is always paying for compute they did not need. We find the receipts in the day's news, then rebuild the thing flat: open formats, precomputed answers, static delivery, and no server standing around waiting to be billed for.
A community project of airbrx, measured against the flat-stack manifesto. Know a story that belongs here? Send it on GitHub.
-
Datacenter Buildout
How to Fill a Gigawatt
Seven projects want more power than The Dalles Dam can make. Some will get built. What matters next is whether those megawatts do real work or answer queries nobody needed.
re: Airbrx
-
Datacenter Buildout
Seven Projects, 5,450 Megawatts, and the Machines That Didn't Need to Be On
Seven data center projects are lined up along a forty-mile stretch of the Columbia Gorge, and between them they've asked for roughly five and a half gigawatts. That's nearly three times what The Dalles Dam can generate. Nobody in Klickitat County is going to stop the buildout with a blog post, and I'm not going to pretend otherwise. But every megawatt in those interconnection queues is a guess about future demand, and a lot of that demand is servers sitting awake waiting for traffic that never comes. The flat stack won't cancel these projects. It can make sure the ones that get built are full of work instead of idle machines, and that the ones that don't need to get built… don't.
re: Columbia Gorge News
-
Attack Surface
92% Left the Door Open. The Problem Is How Many Doors They Actually Built
IBM's 2026 breach report says nearly every company whose AI got attacked never controlled who could reach it. That's not an AI problem. That's what happens when adding a door is free and nobody counts the doors.
re: Cybersecurity Dive
-
Bill Shock
The Agent Only Knows One Direction: Up
Every story about a rogue AI agent is about the thing it deleted. The more expensive stories are about the things it added, kept, and wouldn't let go of. Agents are trained on production architectures, and production architectures only ever grow. Nobody taught them how to scale to zero... re: Lan Tian, InfoQ, GitClear
re: InfoQ
-
Repatriation
Exit Velocity: They Have to Give It Back. They Don’t Have to Make It Readable
Portability is a key to our Flat Stack Manifest, and while Brussels cut the price leash, your vendor can comply perfectly while handing you forty terabytes you can’t read. Sometimes Portability isn't portable after all...
re: Revenue Wizards
-
Bill Shock
The Library That Already Believes In This. And Still Needs $25 a Month From Each of Us
The Internet Archive is one of the least bloated infrastructures on the internet; it's self-built, self-hosted, no ad tech, no vendor tower. And it is still asking donors to triple their money in September just to keep the servers on. We use the Flat Stack Manifesto to support their cause. Solid scorecard means the money you donate to them goes further.
re: The Internet Archive
-
Datacenter Buildout
The Cheapest Data Center Is the One We Don't Have to Build
The cloud has always encouraged us to imagine compute as infinitely available but the physical world is beginning to remind us that it isn't. By 2030, data centers could consume 945 TWh of electricity every year, and building that out costs roughly $11 million per megawatt before you even buy the servers. Wasted compute isn't just waste on a cloud bill. It is electricity, water, land, and billions of dollars of infrastructure being committed to machines that didn't need to be running in the first place. The most important compute optimization may be the compute we never run.
re: Stewart Hicks
-
Bill Shock
Visibility doesn't mean "Problem Solved" -- the trap of good FinOps reporting
The Harness report found that organizations take an average of 31 days to identify and eliminate cloud waste and then It takes about 25 days to detect and rightsize overprovisioned resources. That means it could take TWO MONTHS for you to see cost reduction... an then you find another one...
re: Harness
-
Bill Shock
FinOps Shouldn't Start With the Bill
I think there’s something fundamentally wrong with the idea that we look at reports and try to scale down. As with most of our tech stuff, it’s not really about something going rogue, we usually build up services that we need for the architecture we built.
re: Cloudzero
-
Attack Surface
400 Packages Compromised for the want of 20 lines of code
A maintainer account fell at 09:00 UTC and 400+ packages across the keyv and cacheable ecosystem shipped credential-stealing malware. A cache library runs with the privileges of your process, not the privileges of its job.
re: Wiz Research
-
Attack Surface
Probe all you want, there's no compute here...
A brand-new domain gets found in seconds, because your TLS certificate is published in a log anyone can subscribe to. What arrives next is an alphabet of probes for .env files and forgotten CGI scripts. At the end of that alphabet is a Lambda function running someone else's scanner on your bill.
re: Unit 42, Palo Alto Networks
-
Cascade Failure
GitHub Didn't Break Today. Its Model Vendor Did.
Copilot lost Grok for two hours and fifty-four minutes because of a provider GitHub doesn't own. That phrase — "upstream model provider" — has appeared on their status page twelve times in seven weeks.
re: GitHub Status