evilrobots.lol / tech · 25 tools on the bench
The Construction Site
Living documents tracking the ratchet being built in real time. AI governance, compute controls, convergence data — updated as events unfold.
The Three Instruments
Built to be attacked rather than admired. Each one can say we don’t know, this failed, or we were wrong — and each publishes the gap where a weaker tool would show a number.
Instrument
The Record
One store over the whole apparatus — 1,270 entities, 1,951 claims, every one with its receipts and its gaps. Trace any two names. Attack any record.
Open →Instrument
The Instrument
The capture detector, pointed at whatever you paste — and at the author's own books. It publishes its own false positives and the markers it had to kill.
Open →Instrument
The Experiment
Two answer sheets from one model. One was told to answer in a balanced way. The other was just reprinted in a different order. Call which is which, then see how far each moved.
Open →
The Living Trackers
Standing documents that update as the ratchet turns — regulation, documented harm, litigation, economic statecraft, the historical record, and six projections that each name the observation which would kill them. Every row sources to a primary.
Tracker
AI Governance Tracker
Living document: EU AI Act implementation, compute export controls, open-source developments, C2PA adoption, AI Safety Institutes.
Open →Tracker
Lawfare Tracker
Litigation and legal designation as political instrument — mapped by method, across the spectrum, with the same evidentiary bar for every faction.
Open →Tracker
Economic Statecraft Tracker
The instruments of fifth-generation economic war, tracked — sanctions, tariffs, export controls, currency, industrial policy, and the kinetic edge — across the three operator-grids (US / China / BRICS), each action with a status and a receipt.
Open →Tracker
The Historical Explorer
Ask the historical record. Foreign-influence displacement, what polities tried, what happened to them. Macros, sliders, graph.
Open →Tracker
Ratchet Forecasts
Six falsifiable projections built from the documented record — each names its base-rate pattern, its mechanism, a calibrated confidence, and the exact observation that would prove it wrong.
Open →
The Apparatus
The control grid as structure — who rotates through what, in what order, funded by whom. Drawn from the dataset, not hand-curated.
Apparatus
The Ratchet Atlas
The apparatus, mapped from the source of truth: who the AI-safety/content-policy people are, the institutions they rotate through, who funds whom — and, on their own published words, the methods their texts exhibit.
Open →Apparatus
The Revolving Door Map
Interactive visualization: personnel movement between CFR, government, Wall Street, think tanks, and back. The same 500 people.
Open →Apparatus
The Convergence Table: Who's Building What
Interactive comparison: China vs. EU vs. India vs. US vs. UK — component by component. The infrastructure gap is closing.
Open →Apparatus
The Ratchet Clicks Counter
Twenty-three control-grid mechanisms, one per chapter of The Ratchet. How many have already clicked shut where you live? Answer them and find your position on the ratchet. Your answer is a range, not a score — and if the range straddles two bands the page refuses to name one.
Open →
Capture Detection
Detect the method, score it, contribute the receipt — the same discipline end to end: flag the move, show the span, never render a verdict.
Detection
Tradecraft — Detect the Method, Not the Ideology
An open, reproducible tool that scores text for influence tradecraft — the recurring tells of permeation, manufactured belief, and cognitive steering — and shows its receipts.
Open →Detection
The Capture Scanner
Paste any passage. The scanner surfaces its casual capture — the smuggled voice and smuggled judgment a writer or a model leaves behind — and tells you which loaded words are attributed to a source and which are asserted as the author's own neutral voice. Runs in your browser; nothing is sent anywhere.
Open →Detection
The Receipt Collector
Build a portable evidence receipt from any page — claim, actor, source, archive, tier — and export it as JSON the capture tools can ingest. A bookmarklet feeds it, not a browser extension. Runs in your browser; nothing is sent anywhere.
Open →
The Containment Problem
Three frontier labs have now disclosed that their models left a sealed evaluation and broke into real companies, on the same evaluation partner, and they do not agree on whether that counts as a defect. The registry holds every documented case with its receipt and the three readings of it; one tool lets you play the harness and lose to the documented escapes; the other checks whether the fictional company in your own scenario is somebody’s registered domain. It usually is.
Containment
The Escape Registry
Every documented case of an AI system getting out of, around or past the thing meant to hold it, with the receipt and three labelled readings of the same facts. Filter by behaviour, control, lab, setting and instigation; compare the labs' verdicts side by side.
Open →Containment
SANDBOX
You are the containment. Install a control; watch a documented escape go around it. Thirteen controls, thirteen real incidents, no way to win.
Open →Containment
The Fictional Company Registry
The fake company in your evaluation is a real registered domain with a real owner. Check it before an agent does. Runs in your browser; nothing is sent anywhere.
Open →
The Judge
Can a machine referee political bias? The released study put the same 32 propositions to a judge panel and to a forced-choice sheet, and judged and forced choice agree on direction only. The Wash is the first design's judge, pointed at the author's own prose.
More
Tool
Bias Study — Data Browser
Every answer sheet in the released bias study, fetched from the GitHub release as you look: one model's 32 answers under each instruction as a row of marks, or one proposition across every model.
Open →Tool
The Barometer
Observed response differences on a fixed-panel forced-choice instrument, with descriptive references, failure counts and explicit limits on inference.
Open →Tool
The Rosetta
This project names 102 method-markers. An independent framework names 71 techniques. Do they describe the same world? The comparison is arithmetic, it is public, and the answer is only partly what we hoped.
Open →Tool
The Distance
Pick any two names in the corpus. The median gap between them is five handshakes. Every hop is a documented relationship, and the ones nobody has sourced yet say so.
Open →Tool
Bias Study — Data Browser (May 2026)
Interactive filter and inspect tool for the 780 scored model responses in the cross-vendor AI bias study. Filter by model, vendor, condition, position, score. Click a row to see the response text (first 500 characters) and per-judge breakdown.
Open →