The political economy of power

Who owns
the machine?

Wireframe News tracks the machinery of American power: the compute, the money, the guns, the cages, the law. One lens runs through all of it. Does it distribute, or does it concentrate? We name the mechanism, draw it to scale, and bring the receipts.

5–6entities hold frontier compute
≈2000:1access ratio, top to base
1lens · distribution vs concentration
The master lens

Distribution compounds.
Concentration extracts.

Every story runs through one test. Not left or right, not hero or villain: one structural question, applied to every mechanism we can document. Distribute a consequential resource, whether capital, compute, training, data, or decision-making power, and its returns compound across everyone who holds it. Concentrate it and the risk compounds instead, on the same clock. That's a structural claim, not a moral one, and every dispatch is a test of it.

FIG.01 · the five roles, drawn in the order the argument builds. Every story we run stacks from these layers
Context · the base
Slate green. The broad distributed public: the many small holders a capability starts with.
Institutional · the structure
Steel blue. The state, the load-bearing infrastructure the base builds and feeds.
Hidden · the concentration
Muted violet. Where power pools: the few nodes that set terms for everyone upstream. Usually the argument's center.
Stress · the mechanism
Amber. The pressure that moves it: capital, contracts, control of the chips. The cost.
Collapse · what breaks
Deep red. Failure, irreversible loss: what the concentration costs the base.

Color is never chosen by taste; it is bound to the role. Max three accent roles per frame, usually one or two. Learn the code once and every diagram reads on sight.

The working vocabulary

When a mechanism doesn't have a name, we coin one. Six handles from these pages:

Structured Friction

Democracy's slowness is the design, not the flaw. The friction is what prevents capture.

The Compute Barons

The handful of entities who own frontier compute, and who set the terms for everyone downstream.

Silicon Immigrants

AI as an arriving labor force: no wages, no rights, no ceiling on headcount.

The Signal

The machine-readable version of you, and the algorithmic ecosystem that trades on it.

Crypto-Kleptocracy

The apparatus that converts public office into private tokens, payment rails included.

The Extraction Ledger

The standing account of what gets taken, by whom, and who pays. Updated as the receipts land.

Recent dispatches

The receipts, drawn to scale.

Each dispatch names one mechanism and shows the evidence. Color marks the role at the center of the argument.

Latest dispatch

Payer of Last Resort

The president called FIFA about a red card, and a senator said thank you. The grift was never hidden, it was filed. The pattern, the paperwork, and the repair manual.

Jul 7, 2026 Read
From the record

Beat China, Become China?

Watch you always, watch us never. The tech and finance elite welcome total surveillance because they've already decided they'll be the watchers, not the watched.

Jul 2, 2026 Read
The receipts

Who Gets the Future

New global order: AI CEOs as heads of nation-states at G7

Jun 22, 2026 Read
How we work

Forensic, receipts-first.

No house voice, no throat-clearing. Four rules run every story, in order.

01 State the claim

Lead with the answer. The finding sits in the first line, never buried three paragraphs down.

02 Name the mechanism

Not feelings, not vibes. The contract, the chip, the transfer that actually does the work.

03 Bring the receipts

Every claim is sourced and drawn to scale. If we can't show it, we don't run it.

04 One lens

Distribution versus concentration. Learn the code once, read every diagram on sight.

How we build

The receipts are the product.

AI runs through the whole shop: research fan-outs, drafting passes, pipeline machinery. The gates are human. Anti-low-effort, not anti-AI.

01

Score the field.

A daily pipeline reads hundreds of sources and scores them against the lens. Most die there. What survives has already earned attention.

02

File the receipts.

Everything lands in a research wiki: hundreds of cross-linked pages where every entity, claim, and contradiction is filed. Receipts accumulate across months before a story runs on them.

03

Verify adversarially.

Every load-bearing figure is traced to a primary document before publish. AI runs the fan-outs; a claim that can't survive the check gets a confidence flag or gets cut.

04

Draw it.

The diagrams are vector work, drawn to scale from the verified numbers. Never generated text-images; generators garble exactly the details that matter.

05

Gate the publish.

A voice lint, an editorial grading pass against our best work, a named-civilian requirement, and day-of re-verification of every live story. Karl is the anchor: editor, curator, and the only byline.

FIG.02 · the research wiki: every page a node, every claim linked to its receipts
Where we stand

We build with the thing we cover.

Covering AI power while using AI daily is not a contradiction. It is the qualification. Three terms first, because the argument depends on them:

Narrow AI

Technology with a job: a defined purpose, a bounded domain, an off switch. The kind we use, and the kind we back.

AGI

Artificial general intelligence: matches humans across most cognitive work. The labs' stated goal, on their own timelines.

ASI

Artificial superintelligence: exceeds us at everything, including improving itself. Nobody has explained why we should want this.

Narrow AI, by conviction.

We believe in technology with a job, and we run this shop on it. AGI is the labs' ambition, not a public mandate. ASI is nobody's legitimate need.

Augment, not replace.

The human stays in the loop: in our shop, in every shop. AI should widen who gets to make things, not narrow who gets paid for them.

Good tools, badly aimed.

Technology can be good and used badly. The same feed that connects a family can be tuned to divide a country. We wrote The Signal about exactly that.

The race is the risk.

The AGI race runs on a prisoner's dilemma: every lab builds what it fears because the others will anyway. And the finish line is a species smarter than us. That stopped being science fiction, so thinking about it stopped being optional. We wrote Gods or Ashes about the race.

The political axis is the one that matters.

The engineering debates are mostly landing. The open question is who ends up holding it. The state did not nationalize the labs; the labs are capturing the state. That is the story we cover.

Safety is infrastructure.

The voluntary commitments were the financial regulations of this era, and we watched the repeal happen in six weeks. Friction is not inefficiency. It is the load rating.

The anchor

Every dispatch carries a name.

Karl Herbst
Karl Herbst
Founder · systems thinker · pipeline designer

Karl Herbst is a systems thinker who learns by building: architectural drawings, then visual-effects pipelines, now the reporting machinery this publication runs on. The trade was always the same move. Take a system too complex to hold in your head and break it into parts that read on sight. And he spent decades watching algorithms turn the impossible photoreal, so he knew what a synthetic image could do to an audience long before it had a politics.

“I created Wireframe News to point that literacy at power. The systems deciding what you see are now the story, and I read them from the inside.”

One accountable byline. No house voice. Five roles, five colors, never more than three in one frame: if a frame needs a fourth to make its point, the point isn’t clear yet.