Support Substrate Wire
Substrate Wire is one person and a stack of machines in a Southern California garage. No ads, no paywall, no tracking, no investors. Every story, signal, and verdict on this site is produced on local hardware — the same kind of machines this site reports on, just a few billion dollars smaller.
The Local Compute Fund
This site tracks hundred-billion-dollar AI datacenters. It runs on none of them. Reader support goes to one thing: the next machine. Each upgrade below makes the analysis you read measurably better — bigger models, deeper overnight passes, more sources checked.
The first building block of a dedicated analysis machine: a graphics card that will let the site run larger, sharper models on every overnight edition.
Goal: $1,400 — just getting started
A separate machine handles every overnight deep-analysis job, running models ten times larger than anything that fits on a gaming GPU — a genuine near-frontier reasoning pass on every nightly edition.
Roughly triples the analysis machine's writing speed and makes real-time interactive checks viable alongside the overnight batch jobs.
The largest publicly released open models run at higher quality settings, so the site's most demanding analysis gets meaningfully more accurate.
Every job on the site — breaking-news summaries, deep background, multi-source cross-checks — runs against a frontier-class model at interactive speed.
What support has bought so far
Everything on the site today: nightly editions, Shadow Wire telemetry fusion, the AI Work tracker.
Ways to help
Monthly support keeps the lights on; one-time tips go straight to the fund.
- $3/month — covers the site's domain, CDN, and error alerting
- $7/month — covers one telemetry source's API access
- $15/month — funds roughly an hour of GPU time toward the fund each month
Everything on the site stays free no matter what. No perks behind payment.
Check the work before you give
We publish a public hit/miss prediction ledger — every Shadow Wire call, resolved against reality, misses included. See the track record → The methodology is public too. See the methodology →