🧭 Fundamentals
Understand what the decision solves before installing any model.
Track map
🧭 Deciding is not conversing
Choose the tool by output type.
🧭 The three primitives
Write questions that can be evaluated.
Detailed content
Deciding is not conversing
Choose the tool by output type.
What it is
A decision engine takes a state and chooses within a defined set.
Why learn
Overlapping categories make the label ambiguous even before involving AI.
Key concepts
observed information
question contract
exit option
effect on the system
What it is
System 1 is an analogy for quick, bounded decisions; System 2 represents tasks that require elaboration, planning, or generation.
Why learn
Separating steps lets you measure the cost and error of each one.
Key concepts
narrow task
extensive analysis
distinct steps
deterministic decision
What it is
In Laya inference, an encoder processes the sequence and computes values for the options.
Why learn
A forward pass does not mean constant cost.
Key concepts
represents input
multiple sequences
value before softmax
model calculation
What it is
Jev appears in sources as a decision service accessed via API.
Why learn
Hosting a model involves installation time, memory, energy, monitoring, and updates.
Key concepts
own execution
remote service
complete operation
required service
What it is
The provided images present a narrative of competition, with phrases like the end of LLMs and an alternative that’s even faster.
Why learn
In the course, the image serves as an editorial reference.
Key concepts
gets your attention
testable claim
recorded result
condition of the conclusion
What it is
This course project is a triage console in Portuguese.
Why learn
Before installing, write the minimal input, the categories, the meaning of urgency, and who decides the final routing.
Key concepts
no excess
confirms the decision
residual case
expected result
The three primitives
Write questions that can be evaluated.
What it is
Choice returns the alternative with the highest value in the distribution and the options’ probabilities.
Why learn
Short names with clear descriptions help distinguish classes without consuming the entire token budget.
Key concepts
highest value
set of probabilities
class description
other option
What it is
Score uses an ordered list of levels.
Why learn
A scale only makes sense when the levels have order and distinguishable criteria.
Key concepts
ordered levels
weighted average
meaning of the levels
mean absolute error
What it is
Noul responds to a statement with a value between zero and one.
Why learn
Questions about explicit intent are easier to label than vague predictions about future behavior.
Key concepts
proposition
modeled probability
observed phrase
operational cut
What it is
In the analyzed SDK, the confidence of choice and score is 1 minus the normalized entropy of the distribution.
Why learn
Do not compute ECE using a field called confidence without checking the code.
Key concepts
dispersion
highest probability
frequency versus prediction
internal certainty
What it is
The schema contains stable identifiers, type, instructions, and criteria when necessary.
Why learn
Validate results against the policy: classes must be known, values must be finite, and probabilities should approximately sum to 1.
Key concepts
stable format
data checking
rounding
code that reads output
What it is
Write six customer support messages and manually answer the four questions.
Why learn
Disagreements between humans show where the question or rubric needs improvement.
Key concepts
human answer
multiple readings
documented agreement
unaligned examples