2 min lesson
Which model, and how to prompt it
Use "Use the model as a thought partner, often" to say what a strong answer must include.
Step 1 of 3
Which model, and how to prompt ita daily driver, told to self-validate
On model choice, be a daily driver rather than a constant switcher. Staying on one model for a while - a high-reasoning Anthropic model has been a solid daily driver for security work - builds an intuition for how it approaches problems and where it gets stuck, and it thinks well about security when you tell it to. Evaluate new models when they ship. Some GPT models (the GPT-5.x line among them) are very good at hard problems but slower, so reach for those when you want in-depth work. This is one practitioner's read, not the Cursor opinion.
Interactive widget. Tab through its controls; the result updates in the panel below as you change them.
A daily driver builds intuition; reach for a slower, deeper model on hard problems.
Learn more
Full explanation
Full explanation
If they ask how you'd use the agent itself to do security review, give them the loop: prime sub-agents per vuln class, demand a table plus a working PoC for each finding, then run the "I don't believe you" validation loop until model and human reach ground truth. It shows you treat the agent as a fast but fallible partner - the same suspicion the threat model is built on, turned into a workflow.