We Have an API and We're Prepared to Use It

The SpectralBinary engine is now a service. Send it text. Get back tonal geometry. No training, no guessing, no vibes-based scoring.

For most of the last two years, SpectralBinary lived inside ToneThread Studio as a working framework. It scored conversations, flagged drift, ran benchmarks, refused to flatter people who deserved better, and quietly built up a body of evidence that tone is measurable structure rather than aesthetic opinion.

That phase is over. The framework now answers HTTP requests.

What the API actually does

You send it a piece of text. It returns a five-dimensional tonal coordinate, a zone classification, and a deterministic identifier you can audit forever.

Three properties matter, and they are the reason the API exists at all:

• Deterministic. The same input produces the same output every time. Today, next week, on a different server, after a power cycle. No model drift. No version anxiety. No "well, the score was different yesterday."
• Corpus-independent. Nothing was trained. There is no learned distribution to contaminate, no fine-tune to chase, no dataset to disclose. The engine is a fixed geometric transformation over symbolic input.
• Model-agnostic. SpectralBinary scores text. It does not care which model wrote it, which human wrote it, or whether the two are still distinguishable. That last property is becoming load-bearing.

The five axes, briefly

Tone is not a slider between happy and sad. Treating it that way is how nuance dies. SpectralBinary resolves any piece of text against five independent structural properties, each scored on a normalised continuum.

• Warmth. Empathy, connection, relational temperature. Whether the language reaches toward people or away from them.
• Certainty. Confidence and ambiguity. Tracks how firmly a position is held in the language itself, not in the claim being made.
• Intensity. Energy, urgency, magnitude. Useful for spotting when calm communication suddenly stops being calm.
• Coherence. Structure and logical flow. Drops in Coherence are how obstruction, evasion, and institutional fog become visible.
• Resonance. Temporal persistence. Whether a tonal signal compounds across time or quietly decays. The most forward-looking axis we have.

Zones for human-readable output

The continuous spectrum partitions into six named zones. They exist because not everyone consuming the output wants a five-decimal coordinate. Sometimes you just want to know whether a message landed in Serenity or fell off into Ignition.

• Ignition. Stress, distress, rupture signals.
• Vitality. Urgent, reactive, elevated.
• Harmony. Measured, balanced, considered.
• Serenity. Stable, controlled, confident.
• Immersion. Deep engagement, conviction, narrative gravity.
• Zenith. Exceptional clarity, or context-dependent overstatement.

What sits on top of it

The API is the substrate. The product layer is what most people actually use day to day, and it is growing.

ToneHalo

A safety gate for mail and messaging. Outlook, Gmail, Apple Mail. Pre-send and pre-read tonal scoring with manipulation flagging. Designed to detect predatory and grooming intent before harm reaches the recipient. Plain-language verdicts. No therapy claims. No clinical posture. Just a calibrated second pair of eyes.

ParasiTick

Interpersonal manipulation detection across six categories: gaslighting, coercion, guilt-tripping, love-bombing, deflection, and passive aggression. Returns Overt, Covert, or Clean against each. Built for situations where the words look fine and the structure is doing something else.

BureaucraTick

The institutional cousin. Detects obstruction patterns in regulatory filings, policy documents, guidance language, and corporate communications. Five signals around frame, expert appeal, specificity, agency, and attribution depth. If a document is technically responsive while structurally evasive, BureaucraTick says so.

CarOwnerVirus

A live tonal monitoring dashboard. Real-time five-axis readout on a feed of your choosing, with the Signal Depth Index running underneath. Originally built around one corpus. Now evolving into a research and forensic analytics portal with multi-source ingestion and overlay support.

Financial Tone Analysis

Earnings calls, shareholder letters, regulatory filings, and adviser communications scored across the same five axes with full temporal trajectory. The use case is straightforward. Snapshot sentiment is not the same as tonal trajectory. A management team whose Certainty compresses across four quarters is communicating something a single document cannot show.

The boundary

Worth being plain about this. The methodology is public. The architecture is public. The axis names, the zone topology, the operator class, the differential properties. All of it has been written up, in some cases more than once, and is available to anyone who wants to read it.

The character-to-spectral mapping function and the operator parameter set are not. They are API-gated, and they will stay that way. This is a deliberate position, not a hedge. The methodology has to be inspectable for the framework to be trusted. The mappings have to be controlled for the framework to be reproducible. Both things are true at the same time.

Why now

Three reasons, in descending order of how much they matter.

One. The model collapse problem is no longer hypothetical. AI trained on AI-generated text loses tonal range. The Harvard JOLT paper from March 2025 named this as a public policy crisis without offering a measurement methodology. SpectralBinary is the measurement methodology. Sitting on it would be a strange thing to do.

Two. Compliance, safety, and governance teams have been asking for objective tonal scoring for years and getting handed sentiment classifiers that cannot tell the difference between confidence and bluster. There is a visible gap. We have a thing that fits in it.

Three. The field has built sophisticated infrastructure for measuring what models know. It has built almost nothing for measuring how models sound, whether they persist, and whether they are getting flatter over time. Filling that gap is the work.

If you want access

The API is in private beta. We are accepting a small number of integration partners across four lanes: enterprise communications, mail and messaging safety, financial tonal analytics, and AI evaluation infrastructure. If any of those describe the problem you are trying to solve, write to us.

We have an API, and we are prepared to use it.

ToneThread Studio
East Fremantle, Western Australia
SpectralBinary methodology and architecture are public. Character-level mappings, operator parameters, and calibration internals remain proprietary and API-gated.