LEARNMMXXVI

Prompt engineering,
the durable basics

Prompt engineering is the practice of writing inputs that consistently get useful outputs from an LLM. The hype peaked in 2023 with thousand-line system prompts and elaborate jailbreak chains. The durable craft is much simpler: be specific, give examples, ask for structure, and iterate. This page covers the basics that hold across frontier models in 2026, including the eight inside Namulai.

Try Namulai free30-day free trial · €19.80/month after · cancel anytime
01 / BE SPECIFIC

Vague prompt, vague answer

Write a blog post about marketing produces a generic blog post about marketing. Write a 600-word blog post for B2B SaaS founders about pricing-page A/B testing, with a sceptical tone and three concrete examples produces something usable.

The rule: state the audience, the format, the length, the tone, and any constraints up front. Every minute spent specifying the prompt saves five minutes of revision later.

02 / EXAMPLES

Show, do not just tell

If you want output in a specific shape (a particular kind of list, a particular voice, a particular formatting), give the model one or two examples in the prompt. This is called few-shot prompting and it consistently outperforms long descriptions of what you want.

One good example is worth two paragraphs of instructions. Two examples are usually enough. Beyond three, returns diminish quickly.

03 / STRUCTURE

Ask for structure, get structure

Ask for the answer as a numbered list and you will get a numbered list. Ask for it as a JSON object with these specific keys and most frontier models will produce valid JSON.

For anything that downstream tools or humans need to parse, ask for structure explicitly. For reasoning-heavy questions, asking the model to think step by step before answering still helps, even on models that already have built-in reasoning modes.

04 / IN PRACTICE

Prompting across Namulai's eight models

Most prompt-engineering advice transfers between models, but each has quirks. Claude rewards polite, well-structured prompts and resists adversarial framing. ChatGPT is the most forgiving of casual phrasing. Gemini benefits from explicit role descriptions. DeepSeek prefers concise, technical prompts.

The Namulai chat lets you re-route the same prompt to a different model in one click, which is the fastest way to learn each model's character. Try the same hard question on three of them and the differences become obvious.

05 / FAQ

learn.prompt-engineering.faqTitle

learn.prompt-engineering.faq.q1

learn.prompt-engineering.faq.a1

learn.prompt-engineering.faq.q2

learn.prompt-engineering.faq.a2

learn.prompt-engineering.faq.q3

learn.prompt-engineering.faq.a3

learn.prompt-engineering.faq.q4

learn.prompt-engineering.faq.a4

Practise prompts across eight models in one chat

Try Namulai free

30-day free trial · €19.80/month after · cancel anytime