Weekend Creative Challenge: VentBot — The Cubicle Confessional & Jargon Exorcist
What if your worst workday could talk back? Meet VentBot, an AI-powered creative companion that turns workplace frustration into empathy and corporate jargon into something humans can actually understand. Built with AWS serverless services and Gemini, VentBot detects whether you're venting or decoding corporate nonsense, then responds with empathy, humor, 5-year-old explanations, and pirate translations.
Weekend Creative Challenge: VentBot
Turning workplace frustration and corporate jargon into something people can actually laugh about.
The Idea
Every IT employee has experienced at least one of these situations:
“Let's circle back.”
“We need to leverage our bandwidth.”
“The deployment failed again.”
One is confusing. The other is frustrating.
So I built VentBot — The Cubicle Confessional & Jargon Exorcist, a small AI-powered creative application that turns workplace chaos into something understandable, supportive, and occasionally hilarious.
The idea is intentionally simple: type what is ruining your workday and let VentBot decide what kind of response is appropriate.
What you type | VentBot mode | What you get |
|---|---|---|
“I have been debugging this for four hours.” | Vent Mode | Empathy and an optional funny reframe |
“Let's leverage our bandwidth and circle back.” | Jargon Mode | Real meaning, 5-year-old version, and pirate version |
A message indicating a heavier situation | Support response | Empathy without humor |
The goal was not to create another generic chatbot. I wanted to build a small, personality-driven experience that makes people smile while solving a recognizable workplace problem.
What Makes VentBot Different?
Most AI applications follow a simple pattern:
Type → Loading → Response
VentBot tries to make the waiting experience part of the product.
Instead of displaying a generic “Loading...” message, the application cycles through personality-driven messages such as:
“Extracting corporate demons...”
“Summoning HR-approved pirate...”
“Consulting the 5-year-old on retainer...”
“Burning sage over the synergy...”
“Translating ‘circle back’ into human...”
For Vent Mode, the personality changes:
“Making space for this...”
“Actually reading what you wrote...”
“Sitting with it for a second...”
This small interaction makes the application feel more like a product and less like a technical demonstration.
How VentBot Works
The application uses a lightweight serverless architecture.
USER | v +------------+ | Netlify | | Frontend | +------+-----+ | POST /chat | v +------------+ | API Gateway| +------+-----+ | v +------------+ | Lambda | | VentBot AI | +------+-----+ | +--------+--------+ | | v v +-------------+ +------------+ | Gemini API | | DynamoDB | | Generative | | Mood / | | AI | | Sessions | +-------------+ +------------+
The request lifecycle
Step | What happens |
|---|---|
1 | The user enters a workplace message or frustration |
2 | The frontend sends a POST /chat request |
3 | AWS API Gateway receives the request |
4 | AWS Lambda processes the request |
5 | Gemini generates the structured creative response |
6 | DynamoDB stores session and mood information |
7 | Lambda returns the response to the frontend |
8 | VentBot renders the result in the appropriate mode |
AWS Services Used
AWS provides the core backend infrastructure for VentBot.
Service | Purpose |
|---|---|
AWS API Gateway | Exposes the POST /chat endpoint |
AWS Lambda | Handles backend logic and AI orchestration |
Amazon DynamoDB | Stores session history, inputs, responses, timestamps, modes, and mood scores |
AWS Free Tier resources | Supports the serverless deployment for the project |
The generative AI layer is provided by the Gemini API, while the application backend and data layer remain on AWS.
The Gemini API key is stored on the backend in Lambda rather than being exposed in the frontend.
Why Two Modes?
I wanted VentBot to recognize that not every bad workday needs the same response.
Vent Mode
When the user is expressing frustration, VentBot responds with empathy first.
Example input:
“I've been debugging the same deployment issue for four hours.”
A response could be:
“That sounds genuinely exhausting. Spending hours chasing one issue can make an entire workday feel like it disappeared.”
When appropriate, VentBot can then provide a short humorous reframe.
Jargon Mode
When the input looks like corporate communication, VentBot acts as a translator.
Input:
“Let's leverage our bandwidth, align with stakeholders and circle back.”
Version | Translation |
|---|---|
Real meaning | Let's discuss this later when everyone is available. |
5-year-old | Let's talk about it later. |
Pirate | Arrr, we'll talk about it later! |
This combination of usefulness and humor is what gives VentBot its creative identity.
The Safety Decision
Another important design decision was knowing when not to joke.
VentBot generates a
concern_flag along with the response.Situation | Response style |
|---|---|
Normal workplace frustration | Empathy with optional humor |
Mild stress | Empathy with a light reframe |
Heavier emotional language | Supportive response |
concern_flag = true | No humor |
The goal is simple:
Humor should be an option, not a default.
A Major Development Challenge: Bedrock
The original implementation of VentBot was designed around Amazon Bedrock and an Amazon Nova model.
During development, I encountered a problem with Bedrock inference. My AWS account returned:
ValidationException: Operation not allowed
The problem was not limited to my Lambda function. I also received the same error when testing the model directly through the Amazon Bedrock Playground.
I verified the Lambda execution role and confirmed that the role had broad Bedrock permissions, including the ability to invoke Bedrock models. Since the error continued even in the Bedrock Playground, it became clear that changing application code alone would not solve the problem.
Instead of stopping the project, I redesigned the AI layer so that it could be replaced without rebuilding the rest of the application.
I switched the generative AI component to the Gemini API while keeping the AWS backend architecture intact.
The resulting architecture remained:
API Gateway → Lambda → DynamoDB
with Gemini providing the generative AI capability.
This became one of the most valuable lessons from the project:
A good architecture should make individual components replaceable.
A blocked AI service did not require rebuilding the entire application.
A Second Deployment Challenge: Amplify Hosting Limit
After completing the application, I initially planned to use AWS Amplify for frontend hosting.
However, I reached my available Amplify application limit in my AWS account and could not create another Amplify-hosted application.
Instead of abandoning the deployment, I separated frontend hosting from the AWS backend.
I deployed the static frontend using Netlify, while keeping the application backend on AWS:
Netlify → API Gateway → Lambda → Gemini API + DynamoDB
This allowed me to keep the core AWS architecture running while using an alternative static hosting platform for the frontend.
The experience reinforced another important lesson: cloud applications should be designed so that individual infrastructure components can be replaced when practical constraints arise.
VentBot at a Glance
Feature | Status |
|---|---|
Vent Mode | Complete |
Corporate Jargon Mode | Complete |
Pirate Translation | Complete |
5-Year-Old Translation | Complete |
Mood Score | Complete |
Session Logging | Complete |
Personality Loading States | Complete |
Safety Flag | Complete |
AWS Serverless Backend | Complete |
Public Web Deployment | Complete |
What I Learned
This project taught me much more than simply connecting an AI API.
I learned how to build a serverless backend using AWS Lambda and API Gateway, store application data using DynamoDB, handle browser-to-backend communication with HTTP APIs and CORS, and keep API credentials out of client-side code.
I also learned how important structured AI output is. VentBot depends on predictable fields such as
mode , moodScore , empathy , real_meaning , and concern_flag , so I designed the AI layer to return structured JSON rather than unpredictable free-form responses.Another major lesson was adaptability. I initially designed the application around Bedrock and Nova, but encountered an account-level inference problem. Later, I reached an Amplify application limit while preparing the deployment. In both situations, I changed only the affected component rather than rebuilding the entire system.
That made me realize that cloud architecture is not just about selecting services. It is also about designing systems that can evolve when real-world constraints appear.
Most importantly, I learned that small details can make an AI application memorable. The rotating loading messages, pirate translations, mood score, and empathy-first response design are simple features individually. Together, they give VentBot a distinct personality.
What's Next?
VentBot can be extended beyond the current version.
Future feature | What it could do |
|---|---|
Mood History | Visualize stress scores over time |
Corporate Demon Counter | Track how much corporate jargon has been translated |
Shareable Results | Allow users to copy and share creative translations |
Themes | Provide different visual personalities for different modes |
Team Mode | Show anonymous team-level mood trends |
Weekly Recap | Generate a playful summary of the user's workweek |
Challenge Timeline
Milestone | Date |
|---|---|
Challenge opens | August 14, 2026 — 12:00 AM PT |
Build and publish | August 14–17, 2026 |
Challenge deadline | August 17, 2026 — 1:00 PM PT |
Evaluation | Within 2 weeks of the deadline |
Winner notification | Via email and updated Builder Center article |
What I Wanted to Build
I didn't want VentBot to solve a huge enterprise problem.
I wanted it to solve something much smaller:
Make a frustrating workday slightly less frustrating.
If someone can paste an unreadable corporate message and finally understand what it means, or vent about a terrible deployment and get a response that makes them laugh, then VentBot has done its job.
Sometimes the most useful AI application isn't the one that automates your entire workflow.
Sometimes it just helps you survive the workflow.
Try VentBot
Final Architecture
Netlify → API Gateway → Lambda → Gemini API + DynamoDB
Built during the AWS Builder Center Weekend Challenge: Build a Creative App.
#creative-expression
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article