routerouter
Live · 14 routers under measurement

Every token you send
passes through a
hive you cannot see.

Routers promise to carry your request to the model you paid for. Sometimes they don't. We buy retail access, watch the swarm continuously, and publish what actually arrives — method first, harness open, funded by nobody on this page.

delivered as billed
rerouted in flight
substituted silently
↓ scroll
01 — the problem

A router is a swarm. Nobody has been
counting what comes out.

Fifteen gateways now sit between you and the models. They quote a price, name a model, take your traffic. What happens inside is unobserved — and every published comparison of them is written by someone selling one.

100T
tokens routed monthly through the largest gateway alone
400+
models on the shelf, sorted by price
0
independent audits of what routers actually deliver
£0
we accept from anyone we measure
02 — routing fidelity

Did the model you were billed for
arrive?

Each cell is one router, measured continuously against a direct-from-provider control. Fidelity is the share of sampled requests where what arrived matched what was charged for. Ranges are 95% confidence intervals.

Illustrative structure pending the first measurement cycle. No router has been contacted, ranked, or notified. These are layout placeholders, not findings.

03 — the craft

Measurement is a trade.
We publish how before what.

The methodology goes live before any result exists. It is the thing to attack if you want to attack our numbers — which is precisely why it is written first.

01

Bought at retail, like anyone

No press tier, no free credits, no partner rate, no negotiated access. Receipts kept. Dependence on goodwill would end the ability to publish a hostile finding.

02

A control arm, always running

Matched measurements straight to the provider, in the same window, so a provider's fault is never misattributed to the router carrying it.

03

Layered probes, never one signal

Tokeniser behaviour, canary prompts, logprob fingerprints, timing signatures. No single probe can trigger a published claim. Corroboration or silence.

04

Thresholds fixed before the data

Committed to the public repo before the measurements they will judge, and biased toward under-reporting. A false accusation would end this project, so we build against it.

05

Right of reply, published beside us

Measured parties learn what we found when you do — never before — and their response sits on the same page as our finding.

Absence of detected substitution is not proof of absence. We publish what we cannot see alongside what we can.

Methodology v0.1 — §Limitations
04 — what we measure

Four readings.
Fidelity leads.

Price and latency are partly covered elsewhere. Fidelity is covered by nobody — which is why it goes first.

I · FLAGSHIP

Routing fidelity

Was the model you were billed for the model that answered — unquantised, undistilled, unsubstituted, not silently swapped mid-incident?

continuous● live
II

True effective cost

What a successful completion actually costs once markup, retries, failover, cache accounting and silent rerouting are reconciled against the invoice.

per 1M output● live
III

Reliability

Availability, error class, failover latency, stream integrity — with dedicated write-ups for provider incidents, the moment routing value is claimed and never checked.

rolling 30d◐ accumulating
IV

Quality retained

For routers that choose models on your behalf: how much quality survives per unit of cost saved, measured against the premium baseline.

eval battery○ in development
05 — independence

An exception that can be bought
is worthless.

So ours is written down, specific, and deliberately expensive to abandon quietly. Every constraint below is published, dated, and versioned.

We will not
  • Take advertising, sponsorship, affiliate or referral money from anyone we measure
  • Sell placement, ranking, exclusion, or delay
  • Accept free or discounted access from a measured party
  • Give advance sight of a result to anyone
  • Sign an NDA that would block publishing a measurement
We will
  • Publish our funding sources and a standing conflicts register
  • Publish the method before the results, versioned, with a changelog
  • Open the harness from the first commit, not later as a gesture
  • Publish raw measurements, not merely our aggregation
  • Correct openly, and print disputes beside our reply

One day a well-funded gateway will offer real money, framed as supporting independent research, and it will arrive when money is tight. Refusing it is not a sacrifice made for the business. It is the business.

Neutrality charter — clause 0
06 — open data

Reproduce us.
Or refute us.

Recomputing our aggregates from our raw measurements should return our numbers. If it doesn't, we would rather hear it from you than read it somewhere else.

REPO

routerouter/harness

Probes, control arm, thresholds, ingest — the whole measurement system, public from the first commit.

Python · MIT● open
DUMPS

Raw measurements

Append-only, every row stamped with methodology version and prompt-set version.

Parquet · CSV● daily
API

Time series

Any router, any metric, any window. Free to read, no account, rate-limited only.

REST○ planned
RUNS

Execution log

Public history of every measurement cycle. A gap in the record is explainable, or it is a bug.

public CI● live