MO§ES™ vs Canditech

Simulate the job or measure the job?

Native analytics measure usage. Skills systems measure predefined capability. Engineering systems measure engineering work. MO§ES™ measures the operator operating technology and builds upward from that object.

Canditech is a skills system that uses job simulations. The operator is placed in a simulated work environment — a constructed scenario that mimics a real job task — and evaluated on how they perform. For AI skills evaluation, Canditech monitors ChatGPT sessions during the simulation, observing how the operator uses AI to complete the simulated task.

MO§ES™ is a measurement system that measures real workflows. The operator does their actual work — not a simulation of it — and the system measures the operating behavior that emerges from that work. The canonical telemetry (INPUT, OUTPUT, CACHE READ, CACHE WRITE) is captured from the AI systems the operator uses in their real job, not from a monitored ChatGPT session in a simulated environment.

The difference is simulation vs reality. A simulation is a controlled environment that approximates real work. Reality is real work. Simulations are useful for screening — they tell you whether a candidate can probably do the job. Measurement tells you whether the operator actually does the job well, day after day, across the full range of their real work.

Comparison at a glance.

A simulation is not the job.

Canditech's job simulations are carefully constructed. They mimic real tasks, use realistic data, and create time pressure similar to real work. For hiring and screening, this is valuable — a simulation is a better predictor of job performance than a knowledge test or a self-report.

But a simulation is still a simulation. It removes the context that makes real work real: the specific tools the operator uses, the specific workflows they follow, the specific collaborators they work with, the specific constraints they operate under, the specific AI systems they have access to. A simulation that monitors ChatGPT tells you how the operator uses ChatGPT in a simulated scenario. It does not tell you how the operator uses their actual AI tools in their actual work.

MO§ES™ measures real work. The operator uses their actual AI systems — not just ChatGPT, but whatever systems they use in their job. In the demo dataset, 50 operators produced 1,668 observations across 5 AI providers. The telemetry came from real sessions, not simulated ones. The workflows were the operator's actual workflows, not constructed approximations.

The advantage of real-work measurement is ecological validity. What you measure is what actually happens. The disadvantage is deployment: you need access to the operator's real AI system telemetry, which requires integration. Canditech needs only its own simulation platform.

One tool vs all tools.

Canditech's AI skills evaluation monitors ChatGPT sessions. The operator uses ChatGPT during the simulation, and Canditech observes how they use it. This is a focused measurement: one AI tool, one session, one simulated task.

MO§ES™ measures across all AI systems the operator uses. The canonical telemetry — INPUT, OUTPUT, CACHE READ, CACHE WRITE — is produced by every token-based AI system. In the demo dataset, 5 AI providers and 27 MCP tools were measured. The operator's performance profile aggregates across all the systems they use, not just one.

This matters because operators use multiple AI systems. An engineer might use a coding agent for implementation, a chat model for design review, and a specialized tool for documentation. A simulation that monitors ChatGPT captures none of that. Canonical telemetry captures all of it — because the same four signals are present in every system.

The 5 canonical derived metrics — Leverage, Yield, Token SNR, Log Leverage, Construction — are computed the same way regardless of which AI system produced the telemetry. This means an operator's performance profile is comparable across systems, across workflows, and across roles. A simulation score is comparable only within the simulation's framework.

Screening vs continuous evaluation.

Canditech's primary use case is hiring and screening. Job simulations help employers assess whether a candidate can do the job before hiring them. This is a well-established use case for simulation-based assessment, and Canditech serves it well.

MO§ES™'s primary use case is ongoing enterprise evaluation. The system measures operators continuously, benchmarks them against cohorts, diagnoses performance patterns, tests interventions, and re-evaluates. In the demo dataset, 50 operators were measured over a 30-day window, with 12 interventions tested against target metrics and 13 benchmark classes for comparison.

These are different use cases with different requirements. Hiring needs a quick, controlled, comparable assessment. Ongoing evaluation needs continuous, contextual, governance-guardrailed measurement. An enterprise might use Canditech to screen candidates and MO§ES™ to evaluate and develop the operators it hires.

When to choose which.

Go deeper.