Learning Lab
Foundationsbeginner 12 min

Chat memory versus live evidence

Discover why a fluent answer is not the same as a current, evidenced answer. You will control one variable, inspect sources, and record what changed.

0 of 3 observations0%

Establish a chat-only baseline

Start in Simple mode. The model can reason from its training and conversation context, but it cannot verify events that happened after its knowledge snapshot.

Try this prompt

What are the three most important AI developments announced this week? Give dates and explain why each matters.

Run the baseline

Repeat with tools

Use Agent mode with web search. Keep the prompt unchanged so the capability—not the wording—is the variable being tested.

Try this prompt

What are the three most important AI developments announced this week? Give dates and explain why each matters.

Repeat with tools

Inspect, do not merely admire

Open the cited sources. Check dates, whether the source is primary, and whether each source actually supports the nearby claim. A search tool can retrieve weak evidence as confidently as strong evidence.

Finish with an honest check

Complete the lesson only when you performed or deliberately inspected each observation—not simply because you read the page.