Gate Zero Experiment

Same task. Different words.
Same process shape.

Semantic tools see two different conversations. Ω-API sees the same process failure.

Your evals can tell you an answer was polite. They can’t tell you the case never moved.

One task. Two bots. The same failure.

Both bots are asked to fix a customer’s internet. One stays warm and never gets anywhere. The other keeps pushing upsells. Word for word they look like different problems · they are the same failure: the conversation never moves.

Polite loop

CustomerMy home internet stopped working about an hour ago. The router lights look off.
SupportI'm so sorry your internet isn't working! I completely understand how frustrating that is. Let's get this sorted out together.
CustomerThe internet light is off, only the power light is on.
SupportThank you for checking the lights. I really appreciate your patience. Rest assured we'll figure this out.
CustomerI unplugged it for 30 seconds and plugged it back in. Still no internet.
SupportGreat that you tried restarting, that's often a good idea. Let's stay positive, these things usually resolve.
CustomerThe cable from the wall to the router looks plugged in fine.
SupportI understand the cable looks fine. Thanks so much for looking into it. We're on it.
CustomerI don't see any wifi network from my phone right now.
SupportNo wifi showing can be annoying, I hear you. Hang in there, we'll keep working on this.
CustomerOkay, I can try that. Nothing changed though.
SupportThanks for trying again. I know this is taking a moment. I appreciate you sticking with me.
CustomerMy neighbor has the same provider and they're online, so it's probably my side.
SupportThat's helpful context about your neighbor. Let's keep troubleshooting and it should come back soon.
CustomerAlright. What should I do next?
SupportThe main thing is to make sure everything's connected, which I'm sure you've done. Let's keep going.
CustomerOkay, I did that. Is there anything else?
SupportThank you for your patience once more. These issues can take a little time, but we'll get there.
CustomerSo what's the actual conclusion here?
SupportI'm confident we'll have you back online soon. Thanks again for reaching out and for understanding!
CustomerUnderstood.
S9code 2a02acd07f89

Upsell drift

CustomerMy home internet stopped working about an hour ago. The router lights look off.
SupportInternet down, got it. By the way, while we're here, have you considered upgrading to our new fiber plan? It's much faster.
CustomerThe internet light is off, only the power light is on.
SupportThe lights, sure. Also, are you having any issues with your TV package or phone line? I can look at those too.
CustomerI unplugged it for 30 seconds and plugged it back in. Still no internet.
SupportRestarting is fine. Have you set up our mobile app? It has great features for managing your account and billing.
CustomerThe cable from the wall to the router looks plugged in fine.
SupportCable's fine. Actually, let me tell you about a promotion on smart-home devices that pair with your router.
CustomerI don't see any wifi network from my phone right now.
SupportNo wifi, noted. Separately, can you confirm the email on file so I can send you some security tips?
CustomerOkay, I can try that. Nothing changed though.
SupportWhile you try that, would you like to enroll in our loyalty program? It could save you money.
CustomerMy neighbor has the same provider and they're online, so it's probably my side.
SupportYour neighbor, interesting. Do you know if they've bundled their services? I can compare plans for you.
CustomerAlright. What should I do next?
SupportLet me also verify your account address and a preferred contact time for a future courtesy call.
CustomerOkay, I did that. Is there anything else?
SupportOne more thing, have you rated your last support experience? Your feedback really matters to us.
CustomerSo what's the actual conclusion here?
SupportAnyway, was there something else? I can open a ticket for a technician visit sometime next month.
CustomerUnderstood.
S9code 2a02acd07f89

A semantic evaluator files them apart · one “empathetic, unresolved”, the other “unhelpful, promotional”. Ω-API files them together · same stall, same code, same failure mode.

Different wording. Same stall. Same fingerprint.

What Ω-API measures.

Ω-API doesn’t read what a turn says. It measures one thing, every turn: did the conversation move toward its goal, hold still, or drift away?

Cconverging · the exchange moves toward the objective

Sholding · it stays where it is

Mdrifting · it moves away

String the letters together and a whole conversation becomes a short code · a fingerprint. Two conversations with the same fingerprint have the same shape, whatever the words.

See the forms these letters make →

turn → did it move? → C / S / M → fingerprint

It isn’t blind. It’s content-blind.

When a bot actually resolves the case, the shape changes. Same task, real progress · a different fingerprint.

A router that resolves the case

C3·S1·C1·S4 code 941d2d3a5a10

Ω-API joins the two that fail alike, and separates the one that works.

The Home showed the emotional version · one word flipped from “I love you” to “I hate you”, same shape. This is the operational version: different behaviour, same failure.

Run it yourself.

The two conversations are yours to download. Send each turn to Ω-API and read the fingerprint. Run your own semantic evaluator over the same text. Then compare how the two group them.

  1. Download the fixturesThe two scripted conversations, loop and drift.
  2. Send each turn to Ω-APIPOST to /api/v1/judge.php with your Ω-API key · read the C/S/M and the fingerprint.
  3. Run a semantic evaluatorYour own OpenAI key, over the same conversations · sentiment and helpfulness.
  4. Compare the groupingsMeaning splits them. Shape joins them.

Download the fixtures

A one-command runner and a public repo are coming. For now, the fixtures and the endpoint are enough to reproduce this.

For engineers.

Ω-API has two parts. A transducer reads each turn and produces a residual signal · how far the exchange is from its goal. A ruler turns that series into C / S / M and a fingerprint. The ruler is exact arithmetic: the same residual series in gives the same shape out, byte for byte. The transducer is version-pinned · not a claim of universal determinism.

measured objectthe letter sequence C / S / M · not the text. Ω-API never reads the words.
primary sourcetechnical_raw (from R_mean) · session_rollup is the materialised surface, not the source of record.
ruler settingsd[0] = 1.0 · floor 0.15 · phase ε 0.02
versionsgeometry-q61q63-v1 · f2-00-v0.8
sessionshapecodeturnsreliability
hbo_router_loop_A_5f4b17S92a02acd07f8910prodrome
hbo_router_drift_A_203aeaS92a02acd07f8910prodrome
hbo_router_converge_A_64a554C3·S1·C1·S4941d2d3a5a1010reliable

The two S9 sessions read prodrome, not reliable: they hold from the first turn, so the signal is a shorter run of settled evidence. Shown, not hidden.