liveBook · Manning

Probabilistic model

As explained in Vibe Engineering

A probabilistic model represents uncertainty by assigning probabilities to possible outputs or states.

A probabilistic model predicts likely next tokens rather than objective facts, so its outputs are non-deterministic and have no built-in model of truth. section 1.3

Model settings and added retrieval shift tradeoffs among accuracy, creativity, safety, cost, and latency; no configuration maximizes all five. Treat generated code as uncertain: fluency and a plausible first result do not guarantee correctness. section 1.3 section 1.4

Repeated ISBN validation prompts

Repeated prompts produced different Python ISBN-13 validators. One implementation rejected hyphenated input; another referenced an undefined variable and crashed. Tests for hyphens, checksums, length, and non-numeric characters expose failures that a plausible-looking answer can hide. section 1.4

How Vibe Engineering covers it

How deep it goes
Defines statistical next-token prediction and explains non-determinism and tradeoffs, without detailing model internals.
Examples
Compares repeated ISBN-13 validator outputs and tests for formatting, checksum, length, and non-numeric input.
What you can do after
Evaluate generated code by rerunning prompts and testing cases that expose failures beyond the happy path.
How it connects
Connects probabilistic models to accuracy, creativity, safety, cost, latency, retrieval, and software verification.

Also explained in

Compare how 5 books explain probabilistic model →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Vibe Engineering. The section links point to the parts of the book it draws on.