On May 21, 2026, Jessica Lau published a how‑to listing 10 concrete prompting tips for ChatGPT, arguing that explicit instructions, well‑scoped context, and clear output constraints produce more usable, voice‑accurate results.
Jessica Lau published a how‑to on May 21, 2026 that lays out 10 practical techniques for getting better outputs from ChatGPT. She argues that AI will not infer a user’s intent: prompts must make “the quiet part” explicit so outputs reflect a desired voice, constraints, or role. That clarity matters because it reduces downstream editing and makes results more repeatable for teams that rely on consistent model behavior.
The guide offers a reusable prompt skeleton: define the role (for example, editor or analyst), explain the task, provide relevant context, and specify output format and length. Lau illustrates the skeleton with a concrete example: Role: You’re a B2B marketing editor. Task: Turn these rough notes into a LinkedIn post [paste notes or upload file]. Context: Audience is ops leaders (persona attached); avoid hype. Output: Under 200 words, one hook, three bullets. End with a soft CTA.
Lau groups the ten tips into fundamentals and situational techniques. The fundamentals focus on providing context, including examples, and explicitly defining output format; situational techniques cover model selection, chaining prompts (splitting complex jobs into steps), and a set of short phrases that reliably nudge tone or behavior. She also notes the same prompting patterns translate to other chatbots such as Claude, Gemini, and Copilot. citing research, Lau warns that giving a model more text is not always better — very long instructions can be weighted unevenly and make performance less reliable, so builders should be conservative with context and iterate to find the right balance.
The piece lists several actionable practices for product teams and engineers: include both good and bad examples in prompts to teach style, attach persona documents or concise background notes, and explicitly define format and length. Chaining — breaking a task into ordered steps — helps manage complex generations and improves consistency. According to Lau, these methods reduce rework and increase the likelihood that models deliver production‑ready outputs without heavy manual edits.
Beyond individual prompts, Lau connects prompting practice to automation and product development. She points readers toward workflows that link chat models to external tools so teams can deploy repeatable prompts into production. For builders, that means treating prompts as part of the product surface: versioning, testing, and documenting prompt skeletons to preserve consistency as models and requirements evolve.
Sources
Replies (0)
No replies in this topic yet.
