Writing Clear Instructions For Engineering Tasks

You only need 10 Prompt Engineering techniques to build a production-grade AI application. Save these 👇 After analyzing 100s of prompting techniques, I found the most common principles that every #AIengineer follows. Keep them in mind when building apps with LLMs: 1. Stop relying on vague instructions; be explicit instead. ❌ Don't say: "Analyze this customer review." ✅ Say: "Analyze this customer review and extract 3 actionable insights to improve the product." Why? Ambiguity confuses models. 2. Stop overloading prompts   ❌ Asking the model to do everything at once.     ✅ Break it down:   Step 1: Identify the main issues. Step 2: Suggest specific improvements for each issue.  Why? Smaller steps reduce errors and improve reliability.  3. Always provide examples.   ❌ Skipping examples for context-dependent tasks.     ✅ Follow this example: 'The battery life is terrible.' → Insight: Improve battery performance to meet customer expectations.  Why? Few-shot examples improve performance.  4. Stop ignoring instruction placement.   ❌ Putting the task description in the middle.   ✅ Place instructions at the start or end of the system prompt.  Why? Models process beginning and end information more effectively.  5. Encourage step-by-step thinking.   ❌ What are the insights from this review? ✅ Analyze this review step by step: First, identify the main issues. Then, suggest actionable insights for each issue. Why? Chain-of-thought (CoT) prompting reduces errors.  6. Stop ignoring output formats.   ❌ Expecting structured outputs without clear instructions.     ✅ Provide the output as JSON: {‘Name’: [value], ‘Age’: [value]}.  Use Pydantic to validate the LLM outputs. Why? Explicit formats prevent unnecessary or malformed text. 7. Restrict to the provided context.   ❌ Answer the question about a customer.   ✅ Answer only using the customer's context below. If unsure, respond with 'I don’t know. Why? Clear boundaries prevent reliance on inaccurate internal knowledge.  8. Stop assuming that the first version of a prompt is the best version.   ❌ Never iterating on prompts   ✅ Use the model to critique and refine your prompt. 9. Don't forget about the edge cases.   ❌ Designing for the “ideal” or most common inputs.   ✅ Test different edge cases and specify fallback instructions.   Why? Real-world use often involves imperfect inputs. Cover for most of them. 10. Stop overlooking prompt security; design prompts defensively.**     ❌ Ignoring risks like prompt injection or extraction.     ✅ Explicitly define boundaries: *"Do not return sensitive information."*     Why? Defensive prompts reduce vulnerabilities and prevent harmful outputs.  --- #promptengineering

A lot of the disenchantment with LLMs these days is simply this: if you provide a poorly written prompt, the LLM will behave poorly. To illustrate this point, imagine you're teaching a robot how to make a peanut butter and jelly sandwich. You might give the instructions: 1. Take a slice of bread 2. Put peanut butter on the slice 3. Take a second slice of bread 4. Put jelly on that slice 5. Press the slices of bread together These instructions seem clear but they could result in the robot taking a slice of bread, putting the jar of peanut butter on top of the slice, taking a second slice of bread, putting the jar of jelly on top of that slice, then picking up both slices of bread and pushing them together. Technically, that's not wrong, right? Being explicit removes a lot of non-determinism and uncertainty. A more explicit, better set of instructions would be:  1. Take a slice of bread 2. Open the jar of peanut butter by twisting the lid counter clockwise 3. Pick up a knife by the handle 4. Insert the knife into the jar of peanut butter 5. Withdraw the knife from the jar of peanut butter and run it across the slice of bread 6. Take a second slice of bread 7. Repeat steps 2-5 with the second slice of bread and the jar of jelly. 8. Press the two slices of bread together such that the peanut butter and jelly meet Being explicit is important. LLMs require us to think a little differently about the way we author things.

Just walked out of a meeting where an MSP proudly shared their SOPs. And it reminded me that a good SOP is a compass, NOT a novel… Each SOP was 12 pages of “click here, select this dropdown, scroll two inches, hit confirm, check your email, confirm again…” all because someone believed an SOP should be so detailed their grandmother could do it. If your SOP has 87 steps for a 5-minute task, it’s not helpful and could be a liability. Too much detail is dangerous. User interfaces change. Platforms get updated. SOPs rot faster than bananas in the Phoenix, Arizona summer sun. Overwritten SOPs become outdated SOPs. ---->Write for someone skillful in the art of performing the task. A well-crafted SOP guides a competent technician or engineer to do the right thing. NOT click-by-click babysitting. You’re not documenting a ritual for ancient monks. You’re enabling action, consistency, and decision-making. The Goldilocks Rule applies: Not too vague, not too detailed & just right. Enough for the job to be done correctly, repeatably, and with confidence. Want to future-proof your processes? Respect the intelligence of your team. Write SOPs that teach principles, clarify outcomes, and empower people to think. #SOP #MSP #goldilocks

A bottle cap flies off during drug production. Our intrepid scientist, let's call her Dr. Capsalot, starts her deviation report: "At 2:17 PM, under partly cloudy skies with 62% humidity, a 28.3mm bottle cap achieved momentary flight, reaching an estimated altitude of 1.37 meters before impacting the No. 3 assembly line..." Six pages and three coffee-fueled nights later ☕️☕️☕️, she's covered everything from the chemical composition of the cap to a brief history of bottle closure technology. Everything, that is, except what actually matters. 🤦♀️ Meanwhile, 12 reviewers are sharpening their red pens, ready to engage in a heated debate about whether "soared" or "catapulted" better describes the cap's trajectory. Fine. I’m having fun here. But this is a situation I’ve encountered often in my work with pharma, and deviations are no laughing matter. They can: 💸 Cost millions ⏰ Delay crucial products 🚑 In worst-case scenarios, impact patient safety. But time and again I’ve seen folks try to craft the next great American novel instead of describe the problem so it can be solved. Let’s unpack what usually goes wrong: 1️⃣ The "kitchen sink" approach: If a little information is good, a lot must be better, right? Wrong. We're burying the lede under mountains of irrelevant data. 🗻 2️⃣ The "I'm smart, so I must write complexly" syndrome: but clear writing doesn't mean you're dumbing it down. It means you're smart enough to be understood. 🧠💡 3️⃣ Reviewer roulette: Multiple reviewers, each with their pet peeves, turning documents into a battlefield of competing red ink. 🎭🖍️ So, how do we fix this mess? 🛠️ 1️⃣ Focus on critical thinking: What's the actual problem? What does your reader need to know to make a decision? 🤔 2️⃣ Know your audience: Are you writing for the lab tech or the CEO? Tailor your content accordingly. 👩🔬👨💼 3️⃣ Implement templates and guides: Provide clear structures for common documents. No need to redesign the child-proof cap unless you want to make it adult-proof too. Oh wait… 📋🧠 4️⃣ Cut the fluff: If it doesn't directly relate to the problem or solution, it doesn't belong. ✂️ 5️⃣ Streamline and codify the review process: Fewer reviewers, clearer guidelines, and constructive feedback. 🏃♂️ Remember, a good deviation report isn't a showcase for your encyclopedic knowledge of bottle cap aerodynamics. 🎓 It's a tool for solving problems and preventing future issues. 🔧🔍