I was in a room full of people firing prompts at screens. Live competition, real apps, real time pressure. The kind of event where everyone is moving fast and hoping something sticks.
I slowed down instead.
Google’s PromptWars, organized by Hack2skill, is exactly what it sounds like: prompt battles under the clock. You are building real things, in real time, in front of people watching whether your AI outputs are good enough to win. And most people in that room were doing what you would expect. More prompts. More iterations. More attempts to find the magic input that unlocks the output they needed.
I found a security bug in the app instead. Understood it. Fixed it live before anyone else in the room could get there.
That got me the First Ping Award.
I shared the moment on LinkedIn when it happened. You can read the original post here — and the embedded version is at the bottom of this piece if you want the unfiltered version of what I wrote in the moment.
What this post covers: What actually happened at Google’s PromptWars, the three things I learned about using AI under pressure, and why the person who thinks before they type consistently beats the person who types faster. This is not a post about prompt tricks. It is about a different relationship with AI altogether.
Table of Contents
| 1. What PromptWars Actually Looked Like | 2. The Bug Nobody Else Stopped to Find |
| 3. Three Things I Learned That Day | 4. The Habit Behind the Win |
| 5. Key Takeaways | 6. Frequently Asked Questions |
What PromptWars Actually Looked Like
Hack2skill organized the event as a prompt battle inside a real application environment. You were not writing prompts into a blank chat window. You were building against a live product, in a compressed window, with other participants doing the same thing simultaneously.
The pressure was real. The applications were real. The room had that specific kind of energy where everyone is moving quickly and nobody wants to look like the slowest person in the space.
Most people responded to that pressure the way people usually respond to pressure: they sped up. More prompts. More volume. More hoping that the next attempt would be the one that worked.
I have seen that pattern before. I call it random prompting. It looks like urgency. It is actually noise. When you are firing prompt after prompt without a clear mental model of what the system is doing and what you actually need it to do, you are not working faster. You are just generating more attempts and hoping one of them lands.
The Bug Nobody Else Stopped to Find
While the room was prompting, I was reading the application.
Not reading as in skimming. Actually looking at what the app was doing, what it was exposing, and whether there were gaps between what it claimed to do and what it was actually doing. That observational pause — three, maybe four minutes of actual attention before touching the keyboard — is what produced the result.
The app had a security vulnerability. I found it because I was looking at the system, not at my own prompts. Once I understood what the bug was and why it existed, I used AI as a directed tool: a specific agent, at a specific moment, with a specific instruction. Not a stream of guesses. One move.
I fixed it live before anyone else in the room reached it.
The First Ping Award was not a reward for prompting skill. It was a reward for stopping long enough to see what was actually there.
Three Things I Learned That Day
1. Using AI is not the skill. Using it with clarity is.
Everyone in that room was using AI. That was not the differentiator. The differentiator was having a clear picture of what I needed before I typed anything. Clarity before the keyboard. That one shift changes the quality of everything you produce with AI, whether you are in a live competition or a client project.
2. Direction plus the right agent, at the right moment, is actually faster than random prompting.
This sounds counterintuitive when you are watching someone iterate through twenty prompts in five minutes. They look faster. They are not. They are spending time on attempts that do not move the problem forward. Choosing the right tool for the specific job and giving it a precise, scoped instruction is slower to initiate and faster to complete. The math always lands in favor of deliberate direction.
3. Token usage matters. Precise instructions beat long ones every time.
Long prompts are not better prompts. They are often hedged, ambiguous, and full of instructions that contradict each other or give the model nowhere clear to go. The prompt that worked that day was not long. It was specific. The model knew exactly what it was doing and why. That specificity came from understanding the problem first.
The Habit Behind the Win
This is the part that does not get discussed enough. The win was not a technique. It was a habit.
The habit of reading before acting. The habit of forming a mental model before opening a chat window. The habit of treating AI as a directed instrument rather than a search engine you throw keywords at.
I approach AI as a Creative Director, not just a user. That distinction matters. A user wants output. A Creative Director wants the right output, in the right form, for a specific purpose. The directorial function — the clarity about what you are actually trying to make — is what separates work that lands from work that merely exists.
Most of my content on the blog at byharshal.com is about this distinction in practice. Not tricks. Not prompt templates. The thinking that happens before any of that.
If you have been frustrated by inconsistent AI output, the gap is almost never in the prompt. It is in the clarity that preceded the prompt.
Key Takeaways
- Random prompting looks like productivity. It is noise wearing productivity’s clothes.
- Stopping to observe before acting is not losing time. It is gaining clarity that speeds up every step after it.
- Precise instructions beat long ones. Token efficiency is a by-product of understanding what you actually need.
- The right agent at the right moment outperforms the wrong agent with many instructions.
- Using AI with clarity is the skill. Using AI at all is just the baseline.
- Most AI output problems trace back to the moment before the prompt, not the prompt itself.
The Original LinkedIn Post
Frequently Asked Questions
What is Google PromptWars?
PromptWars is a live AI competition organized by Hack2skill in partnership with Google. Participants compete in real-time prompt battles using actual applications, with the goal of producing the best, most accurate, or most creative output under time pressure. It is not a theoretical exercise. The apps are real, the clock is real, and the stakes are in the room with you.
What is the First Ping Award?
The First Ping Award at PromptWars recognized the first participant to identify and resolve a significant issue in the competition's live application environment. In this case, it was a security vulnerability. The award went to the first person to find it, understand it, and fix it live — which required observation and directed tool use rather than prompt volume.
What does "directed prompting" mean in practice?
Directed prompting means forming a clear mental model of the problem before you open any AI interface. You identify exactly what you need the model to do, which model or agent is best suited for that specific task, and what a good output actually looks like. Then you write the smallest possible instruction that delivers that. It is the opposite of trying several angles and seeing what comes back.
Can this approach work outside of competitions?
It is more valuable outside of competitions than inside them. In a competition you have a defined problem and a time limit that forces focus. In day-to-day work you can spend hours on a task without any structural pressure to pause and form a clear picture first. Building the habit of reading before acting, and understanding the system before touching it, compounds in regular work far more than in a one-off competition environment.
How do I build the habit of clarity before prompting?
Start by adding one step before every AI task: write down in one sentence what you actually need the output to do. Not what you want the model to say. What the output needs to accomplish. This one sentence is your target. Write the prompt only after you can answer that question clearly. Most people skip this step entirely. It takes 90 seconds and it changes what you produce.
Harshal Saraf is a Creative Director and AI Workflow Consultant based in Indore, India. Under his practice ByHarshal, he sets up AI workflows for founders, agencies, and brands across India. Where Creative Direction Meets AI Orchestration. He has led creative direction for brands and small and medium scale B2B businesses, and currently works as Creative Director and AI Strategist at Square Root SEO. He writes Oh, So AI, a Tuesday and Friday newsletter on AI tools, workflows, and productivity for founders and creatives.
