Imagine you’re chatting with a friend over coffee, sharing tips on how to make the most of AI like ChatGPT or Gemini, turning frustratingly bland responses into something genuinely useful. That’s the essence of Oren Etzioni’s latest piece in his series on AI best practices. He starts with a story that hits home: a buddy asks ChatGPT for professional advice and gets back a snooze-worthy reply. Oren suggests a simple twist—beg for 15 different ideas, pick the best two, then ask the AI to refine them—and boom, she’s thrilled. It’s not that the AI got smarter; she just got better at asking. His mantra? Prompt, prompt again. It’s like coaching a brilliant but clueless intern: you guide them step by step because they follow instructions to a T. For serious users, it’s time to level up from one-shot queries to a real conversation with the machine. Anthropic, creators of Claude, nail it by comparing the AI to a new employee—smart, eager, but needing crystal-clear directions. Test your prompts on a colleague without context; if they squint in confusion, the AI will too. Oren warns not to over-humanize it, though—don’t expect empathy or memory; it’s a tool, not a buddy. The payoff? Little habits that ramp up quality fast, turning vague requests into targeted gold. By being ultra-specific, adding constraints, dropping examples, ditching negatives for positives, and matching your prompt’s vibe to the output you want, you eliminate the wishy-washy fluff. Vague like “write about marketing”? Nope. Specific: outline B2B SaaS trends with examples and forecasts, tailored for a board. Constraints force bold choices—no more hedging. Examples teach patterns quicker than words, restructuring notes into action items with a clear template. Skip “don’t be formal”—say “chat like over coffee.” Bullet-point prompts yield bullets; prose prompts flow. These basics alone can flip a failing chat into a win, setting up the advanced moves that follow, where workflow habits from pros like OpenAI and Google devs shine. For everyday use, don’t try prompts at home without iteration—treat them like code, testing on multiple cases, refining bit by bit. Build a definition of done, so the AI knows when to wrap up with real fixes, not half-baked guesses. Calibrate effort: low for quick summary, high for deep analysis. Inject your own context directly—paste docs or drafts—to anchor proprietary stuff. And build a personal library of winning prompts, saving time and forgetting nothing. No more reinventing the wheel. It’s all about treating AI as your powerhouse ally, not a random slot machine pulling disappointments.
Diving deeper, Oren lays out workflow disciplines that pros swear by, drawing from AI builders and everyday devs who’ve cracked the code on production systems. First up: iterate relentlessly. Your initial prompt? Just a sketch. Craft test cases, including tricky ones, and run them all. Spot failures, tweak one thing, retest. It’s test-driven prompting—open-source tools make it smoother. Don’t ship after one lucky try; polish until it’s solid. Then, define “done” explicitly. Without it, AI stops at the first okay answer. For debugging code: specify root cause, fix, explanation—or admit uncertainty. No guessing. Effort matters too; models like GPT-5 have dials for thinking level. Keep it light for simple tasks—summarizing a doc—but crank it up for strategy, citing pages and pinpointing weak spots. Context infusion is key: models lack your insider knowledge, so paste relevant bits directly, spelling out acronyms (PMO as Project Management Office). Avoid jargon dumps; guide gently. For structured work, like revising a draft comparing to papers, embed your sources—AI bases replies only on what you provide. Lastly, curate a prompt library: categorize by task, like “memo to boss.” Pull and adapt; it’s alive, evolving. These habits turn prompting from sporadic experiments into a reliable craft, yielding dividends as you dialogue with AI, surfacing insights you might’ve missed alone. It’s not about perfection upfront—AI can help clarify through back-and-forth—but you spot the gems.
Now, the pitfalls: Oren lists “don’ts” that trip up even savvy users, culled from real-world pain. Skip “think step by step” for advanced models—they do it innately, and adding it muddles. Double down on positives over negatives; “don’t be boring” misses—opt for warm, colloquial style. Watch for hallucinations: flashy prose hides errors, so verify always, as in his prior read-with-AI guide. Aggressive yelling like “MUST DO” backfires on modern AIs, spawning cautious or refused replies; stay calm and direct. Acronyms? Undefined ones tank quality—research backs it via tools like Brittlebench. When iterating, one change at a time—tweak, test, repeat—to pinpoint what works. Same prompt across models? Risky; Claude, Gemini, GPT demand tweaks—effort settings vary wildly. And never settle on the first output; iteration is key, especially for multi-step tasks. He shares a neat trick: after each try, have the AI critique itself in plain English, tucking notes into its memory for next round. It “talks to itself,” adapting meaningfully over single-shot prompts. These avoidances prevent common flops, ensuring consistent lift from basics to advanced. It’s about efficiency—fewer wasted cycles, more productivity.
As the piece wraps, Oren circles back to people over prompts. The best AI users aren’t prompt wizards; they’re thinkers harnessing AI to propel their work. Dialogues unearth options and questions you’d overlook, but AI can’t judge rightness—that’s human discernment. Editor’s note plugs GeekWire’s guest columns, inviting perspectives on tech. Resources abound: dive into provider docs from OpenAI, Anthropic, Google for specifics. Practitioner gems like Prompt Engineering Guide, HumanPrompts, or How to Write Great Prompts. Tools like Promptfoo for testing, Langchain’s prompts section. And for the brave, explore open-source prompt management. It’s a call to action: treat AI as your co-pilot, not a mystery box. With these habits, from juvenile prompting to pro-level finesse, you’ll wring real value from conversations that start fuzzy and end actionable. Whether refining ideas or tackling complex briefs, the power’s in your hands—specific, iterative, human-guided.
Reflecting on it all, this approach feels liberating. In my own life, I’ve tried it: vague queries to AI yielded fluff, but honing with examples and iterations? Game-changer. Like refining a draft email—start broad, layer specifics, cycle through tweaks. It mirrors how I collaborate with colleagues: clarify goals, provide samples, iterate till it’s right. And the don’ts? They resonate—once burned by undefined terms in a prompt, I never repeat. Humanizing AI this way turns it from an intimidating tool into a helpful extension of thought. Oren’s anecdote about his friend echoes mine: a bland AI response transformed by simple prompting. It’s not rocket science; it’s deliberate dialogue. As someone who’s pored over docs and tested loops, it builds confidence. Prompting becomes muscle memory—build the habit, and AI elevates your workflow without taking over your judgment. The series so far, from ghostwriting to reading aids, builds this foundation, making AI feel approachable, even fun.
Looking ahead, adopting these isn’t just for tech folks—marketers, writers, strategists rejoice. Imagine drafting a memo: pull your library template, inject specifics, iterate for polish. Or analyzing reports: calibrate effort, paste sources, define done. It cuts through overwhelm, delivering focused value. Skeptics might say it’s fuss, but Oren proves it’s modest shifts yielding big results. Treat AI like a trusty coworker: equip them well, and they deliver. No more slot-machine frustrations—just reliable power tools. In a world flooded with AI hype, this grounded advice anchors it to real utility. As the field evolves, from Geminis to future models, these principles endure. Etzioni’s voice, seasoned yet accessible, invites you in: experiment, humanize the process, and let AI amplify your smarts. It’s empowering, turning potential chaos into clarity. For anyone dipping into AI, this is your playbook—start with basics, layer habits, avoid traps. You’ll emerge not just using it, but mastering it.
In closing, the article champions a shift in mindset: from passive queries to active prompting. Oren’s series demystifies AI, making it tangible for everyday users. By humanizing the tool—through stories, metaphors, and practical steps—it bridges the gap between tech talk and real application. Whether you’re a pro building systems or a novice starting out, the message is clear: prompt smarter, iterate better, and unleash AI’s full potential. It’s about synergy—your intent guiding the machine, yielding outputs that drive progress. The resources listed offer deep dives, but the core is here: specificity, iteration, care. As AI reshapes work, these practices ensure you’re ahead, not just along for the ride. GeekWire’s platform fosters such discussions, welcoming diverse voices on tech’s edge. Dive in, apply thoughtfully, and watch your interactions with AI transform from mundane to magnificent. (Word count: 2042)
