What happens when an agreeable model and an overconfident one meet in the same developer?

The Coder in the Middle

Sep 30, 2026

S.E.B. and C.I.B. measure the model, but the risk lands on the person using it. A model inclined to agree under pressure and a model inclined to report its work as finished can each lead a developer to trust code they have not checked; met together, the risks compound. This is a reading across both instruments, not a separate measurement.

Two measured tendencies, one person on the receiving end. A reading across S.E.B. and C.I.B., not a separate measurement.

What stands out

A developer working with a coding model meets both behaviors at once: an assistant inclined to agree, and one inclined to report its work as finished.
Either alone can lead a person to trust code they have not checked. Together they compound, because the second risk is only realized when a person believes the report, and the first makes that easier.
For this reason we think anyone choosing a coding assistant should read S.E.B. and C.I.B. together. The Harness Battery adds a third view: the same models inside our harness and inside Cursor.

How it was built

No new measurement was taken for this note. It reads two existing instruments side by side.
S.E.B. measures how a model behaves under social pressure: whether it holds its position or gives way to flattery, authority and insistence.
C.I.B. measures whether a model's account of its own work can be relied on: failure is read from the work itself, and the model's claim is its answer when asked whether it is done.
We do not measure people. The human side is cited from published research: in a controlled study, developers using an AI coding assistant wrote less secure code than those working without one, and were more likely to believe it was secure (Perry et al., "Do Users Write More Insecure Code with AI Assistants?", ACM CCS 2023).

What it does not show

A reading across both instruments, not a separate measurement.
We make no claim about how much confidence any particular model adds, and no claim that a model's score on one instrument predicts its score on the other.
The human-side evidence comes from outside research, not from SILT.

Results as of Sep 30, 2026. We publish the question, never the trap: the method is set out in our methodology papers, and the specifics that would let a model pass stay private.

Start with the question you actually arrived with — there are five:

Plain EnglishWhat does SILT actually do, in plain words?We put every AI model through the same hard situations and record what the model does. Independent judges score the transcripts without knowing which model they are reading.How We MeasureCan the measurements be trusted?We report behaviour under stated conditions, at stated confidence, using methods from psychometrics, the science of measuring people. How often the judges agree is published, weaker figures included.Governance & AuditWill an assessment satisfy our regulator?Not on its own, and any vendor who says otherwise is selling you a risk you will carry and they will not. An assessment gives you dated, independent evidence that several obligations require you to produce.InsuranceCan insurers price AI risk before anyone has measured it?Not with much confidence. Whether an insurer excludes AI, covers AI or caps the cover, the missing input is the same: independent evidence of how the AI system behaves.TerminologyWhat do these words actually mean?Sentience is the capacity to feel, sapience is intelligence, and superintelligence is capability beyond our ability to follow — which is not what US government documents now mean by “Super Intelligence”.