Business Pill · Where a model stops knowing
A knowledge cutoff is the date where a model’s training data ends. Everything after it, the model has never seen, and it does not always know what it is missing.
A short explainer video, under a minute. The customer and the robot are an illustration.
At a glance
What a Knowledge Cutoff Is
That day has a name. It is called the knowledge cutoff: the date where a model's training data ends.
The Harder Problem
The harder problem is that the robot does not always know what it is missing.
Why It Matters
The video turns this into one question: does this answer depend on anything recent?
The Short Answer
Picture a customer who asks a robot who won the big match last night. The robot answers with confidence. It names last year’s winner.
The robot is not lying. Its reading stopped on a certain day, and everything after that day it has never seen.
What a Knowledge Cutoff Is
That day has a name. It is called the knowledge cutoff: the date where a model’s training data ends.
The robot’s wrong answer is not a malfunction. It is an answer from the world as it was on that day.
The Harder Problem
The harder problem is that the robot does not always know what it is missing. It can describe an old world as if it were today.
That is why the answer came with confidence: nothing inside the robot tells it that last night is missing.
The Idea on One Board
The Harder Problem
The harder problem
The harder problem is that the robot does not always know what it is missing.
The fix
The fix is to hand the robot fresh information.
The Fix
The fix is to hand the robot fresh information. The video gives three forms: a search, a document or today’s date.
Why It Matters
The video turns this into one question: does this answer depend on anything recent?
A model knows the world up to a date. After that, it needs to be told.
The honest limit
The harder problem is that the robot does not always know what it is missing. It can describe an old world as if it were today.
Where This Shows Up in AI Business Today
From our news coverage
Ai2 Open-Sources AstaBrief: 8B Model, 3.5x Faster Reports
Our report on Ai2’s AstaBrief says that Ai2 states that most of its training and evaluation was done in 2025 and has not been rerun against today’s frontier models, and that its results are best read as validation of its engineering approach at the time. The date a model’s training data ends is what this lesson calls the knowledge cutoff.
The Business Engineer Lens
Where this idea sits in the Business Engineer library:
The Question to Ask
- Does this answer depend on anything recent?
How This Concept Connects
Layer 2 of 8 · Models & Training, on the Business Engineer Concept Map
- Knowledge Cutoff builds on Pretraining. A model knows only its training text up to the day it was collected, which the cutoff lesson names as the knowledge cutoff.
- Knowledge Cutoff contrasts with AI Hallucination. A cutoff answer is a true answer from an old date; a hallucination is fluent text invented when the model does not know.
- RAG builds on Knowledge Cutoff. RAG starts from the fact that a model knows only its training data; looking up pages hands it what it never saw.
The Product Overhang Doctrine
Capability limited by missing fresh context.
Get the full Concept Map as a free poster.
Related Business Pills
All Business Pills
- The Memory Wall: Why a Faster AI Chip Is Not Faster AI
- Tokens per Watt: What an AI Data Centre Actually Produces
- Why AI Can’t Be Both Instant and Cheap: Latency vs Throughput
- The Model and the Harness: Why Same-Model Products Differ
- Context, Not Capability: Why a Smart AI Model Gives Poor Answers
- Discardable Software: When Code Is Cheap Enough to Throw Away
- Human in the Loop vs Human on the Loop: Supervising AI Agents
- The AI Audit Problem: When Making Work Is Cheaper Than Checking It
- AI Evaluation as Acceptance Test: How to Know It’s Good Enough
- Extensibility Is Control: Who Holds the Power in an AI Product
- RLHF Explained: How an AI Model Learns What People Prefer
- Pretraining Explained: How an AI Model Learns Before Anyone Teaches It
- Fine-Tuning Explained: How to Adapt a General AI Model to One Job
- Tokens Explained: The Unit AI Reads, Writes and Bills In
- Embeddings Explained: How a Machine Compares Meaning
- RAG Explained: How a Model Answers From Your Documents
- AI Hallucination Explained: Why Models State False Things
- Distillation Explained: How a Small Model Learns From a Large One
- Reasoning Models Explained: What Changes When AI Thinks First
- Tool Use Explained: How an AI Model Goes From Text to Action
- Capex and Depreciation Explained: Why Chip Lifetime Drives AI Profits
- Run-Rate Revenue Explained: What an AI Company’s Number Means
- Backlog Explained: Revenue That Is Signed but Not Yet Earned
- Switching Costs Explained: Why It Is Hard to Leave an AI Supplier
- Temperature Explained: The Dial That Sets How Predictable an AI Model Is
- Guardrails Explained: Rules Enforced Around an AI Model, Not Inside It
- Benchmarks Explained: Why a Test Score Is Not Your Own Result
- Agent Memory Explained: How an AI Assistant Remembers You Between Chats
- MCP Explained: The Shared Plug That Connects AI Models to Tools
- Quantization Explained: How a Large AI Model Is Shrunk to Fit
- Parameters Explained: Where an AI Model’s Knowledge Lives
- Scaling Laws Explained: Why AI Labs Keep Building Bigger
- Synthetic Data Explained: When a Model Writes Its Own Training Examples
- Transformers and Attention Explained: How AI Reads Every Word at Once









