DEV Community: Tanmay Singh

🔍 Suspicious code patterns and potentially abandoned functions

Tanmay Singh — Sat, 31 Oct 2026 11:42:12 +0000

👻 PhantomTrace — The Ghost in Your Codebase

Every codebase has ghosts. Forgotten TODOs, abandoned functions, and decisions nobody remembers making.

What I Built

PhantomTrace is an AI-powered codebase explorer inspired by the theme “Ghost in the Machine.”

Over time, software projects accumulate forgotten code, unresolved TODOs, outdated comments, and mysterious bugs. PhantomTrace acts like a digital archaeologist, exploring a repository and identifying these hidden traces of its past.

It helps developers discover:

👻 Forgotten TODOs and unfinished features.
🔍 Suspicious code patterns and potentially abandoned functions.
🧠 Connections between old comments and current code.
💬 AI-generated explanations of why a particular piece of code might deserve attention.

The goal is to make forgotten parts of software visible again.

Demo

Live demo: [Insert your actual deployed demo URL]

Screenshots or video: [Insert a real screenshot or demo recording]

Code

GitHub repository: [Insert your actual public repository URL]

The project source code, setup instructions, and contribution guidelines will be available in the repository.

How I Built It

The proposed stack combines open-source tools with a local AI workflow:

Python for repository analysis and backend logic.
Ollama for running a supported open-weight language model locally.
Qwen2.5-Coder or another compatible coding model for explaining code patterns and generating summaries.
GitPython for examining repository history.
Streamlit for a lightweight interactive interface.

The system first scans a repository for relevant code patterns and historical clues. It then passes selected context to the language model to generate explanations and recommendations.

The idea is to keep the analysis transparent: developers should be able to inspect the evidence behind a finding rather than blindly trust an AI-generated warning.

Note: These are proposed implementation details; retain only the tools and features you actually implement.

Why Does Open Innovation Matter?

Open innovation makes projects like PhantomTrace possible because developers can experiment with models, inspect their behavior, and adapt the entire workflow to their needs.

Using open-weight models can also make local analysis possible without sending source code to a proprietary AI API, depending on the chosen model and setup.

Open-source libraries provide the building blocks, while an open repository allows other developers to inspect the implementation, report false positives, contribute new analysis rules, and improve the project.

The result is not just an AI tool, but a starting point for a community-driven experiment in understanding the hidden history of software.

My Agent Session

[Insert a genuine DevRelay agent-session link if you have one, or remove this section.]

The Spirit of It

The strangest thing about old code is that it remembers decisions long after the people who made them have forgotten.

A comment left three years ago might explain why an apparently unnecessary condition exists. A half-finished function might reveal a feature that was never released. A forgotten TODO might be the beginning of someone's original idea.

PhantomTrace explores that feeling: sometimes the ghost in the machine is simply the past, waiting for someone to understand it.

Prize Categories

[Add only the eligible partner categories that you actually qualify for, based on the official challenge rules.]

Built for the Hacktoberfest Open-Source AI Challenge — Week 4: Ghost in the Machine.

👻 Ghost in Your Codebase

Tanmay Singh — Sat, 31 Oct 2026 11:40:17 +0000

👻 PhantomTrace — The Ghost in Your Codebase

Every codebase has ghosts. Forgotten TODOs, abandoned functions, and decisions nobody remembers making.

What I Built

PhantomTrace is an AI-powered codebase explorer inspired by the theme “Ghost in the Machine.”

It helps developers discover:

👻 Forgotten TODOs and unfinished features.
🔍 Suspicious code patterns and potentially abandoned functions.
🧠 Connections between old comments and current code.
💬 AI-generated explanations of why a particular piece of code might deserve attention.

The goal is to make forgotten parts of software visible again.

Demo

Live demo: [Insert your actual deployed demo URL]

Screenshots or video: [Insert a real screenshot or demo recording]

Code

GitHub repository: [Insert your actual public repository URL]

The project source code, setup instructions, and contribution guidelines will be available in the repository.

How I Built It

The proposed stack combines open-source tools with a local AI workflow:

Python for repository analysis and backend logic.
Ollama for running a supported open-weight language model locally.
Qwen2.5-Coder or another compatible coding model for explaining code patterns and generating summaries.
GitPython for examining repository history.
Streamlit for a lightweight interactive interface.

The system first scans a repository for relevant code patterns and historical clues. It then passes selected context to the language model to generate explanations and recommendations.

The idea is to keep the analysis transparent: developers should be able to inspect the evidence behind a finding rather than blindly trust an AI-generated warning.

Note: These are proposed implementation details; retain only the tools and features you actually implement.

Why Does Open Innovation Matter?

Open innovation makes projects like PhantomTrace possible because developers can experiment with models, inspect their behavior, and adapt the entire workflow to their needs.

Using open-weight models can also make local analysis possible without sending source code to a proprietary AI API, depending on the chosen model and setup.

The result is not just an AI tool, but a starting point for a community-driven experiment in understanding the hidden history of software.

My Agent Session

[Insert a genuine DevRelay agent-session link if you have one, or remove this section.]

The Spirit of It

The strangest thing about old code is that it remembers decisions long after the people who made them have forgotten.

PhantomTrace explores that feeling: sometimes the ghost in the machine is simply the past, waiting for someone to understand it.

Prize Categories

[Add only the eligible partner categories that you actually qualify for, based on the official challenge rules.]

Built for the Hacktoberfest Open-Source AI Challenge — Week 4: Ghost in the Machine.

👻 PhantomTrace — The Ghost in Your Codebase

Tanmay Singh — Sat, 31 Oct 2026 11:36:57 +0000

👻 PhantomTrace — The Ghost in Your Codebase

Every codebase has ghosts. Forgotten TODOs, abandoned functions, and decisions nobody remembers making.

What I Built

PhantomTrace is an AI-powered codebase explorer inspired by the theme “Ghost in the Machine.”

It helps developers discover:

👻 Forgotten TODOs and unfinished features.
🔍 Suspicious code patterns and potentially abandoned functions.
🧠 Connections between old comments and current code.
💬 AI-generated explanations of why a particular piece of code might deserve attention.

The goal is to make forgotten parts of software visible again.

Demo

Live demo: [Insert your actual deployed demo URL]

Screenshots or video: [Insert a real screenshot or demo recording]

Code

GitHub repository: [Insert your actual public repository URL]

The project source code, setup instructions, and contribution guidelines will be available in the repository.

How I Built It

The proposed stack combines open-source tools with a local AI workflow:

Python for repository analysis and backend logic.
Ollama for running a supported open-weight language model locally.
Qwen2.5-Coder or another compatible coding model for explaining code patterns and generating summaries.
GitPython for examining repository history.
Streamlit for a lightweight interactive interface.

The system first scans a repository for relevant code patterns and historical clues. It then passes selected context to the language model to generate explanations and recommendations.

The idea is to keep the analysis transparent: developers should be able to inspect the evidence behind a finding rather than blindly trust an AI-generated warning.

Note: These are proposed implementation details; retain only the tools and features you actually implement.

Why Does Open Innovation Matter?

Open innovation makes projects like PhantomTrace possible because developers can experiment with models, inspect their behavior, and adapt the entire workflow to their needs.

Using open-weight models can also make local analysis possible without sending source code to a proprietary AI API, depending on the chosen model and setup.

The result is not just an AI tool, but a starting point for a community-driven experiment in understanding the hidden history of software.

My Agent Session

[Insert a genuine DevRelay agent-session link if you have one, or remove this section.]

The Spirit of It

The strangest thing about old code is that it remembers decisions long after the people who made them have forgotten.

PhantomTrace explores that feeling: sometimes the ghost in the machine is simply the past, waiting for someone to understand it.

Prize Categories

[Add only the eligible partner categories that you actually qualify for, based on the official challenge rules.]

Built for the Hacktoberfest Open-Source AI Challenge — Week 4: Ghost in the Machine.

# 👻 PhantomTrace — The Ghost in Your Codebase

Tanmay Singh — Tue, 27 Oct 2026 17:12:11 +0000

👻 PhantomTrace — The Ghost in Your Codebase

Every codebase has ghosts. Forgotten TODOs, abandoned functions, and decisions nobody remembers making.

What I Built

PhantomTrace is an AI-powered codebase explorer inspired by the theme “Ghost in the Machine.”

It helps developers discover:

👻 Forgotten TODOs and unfinished features.
🔍 Suspicious code patterns and potentially abandoned functions.
🧠 Connections between old comments and current code.
💬 AI-generated explanations of why a particular piece of code might deserve attention.

The goal is to make forgotten parts of software visible again.

Demo

Live demo: [Insert your actual deployed demo URL]

Screenshots or video: [Insert a real screenshot or demo recording]

Code

GitHub repository: [Insert your actual public repository URL]

The project source code, setup instructions, and contribution guidelines will be available in the repository.

How I Built It

The proposed stack combines open-source tools with a local AI workflow:

Python for repository analysis and backend logic.
Ollama for running a supported open-weight language model locally.
Qwen2.5-Coder or another compatible coding model for explaining code patterns and generating summaries.
GitPython for examining repository history.
Streamlit for a lightweight interactive interface.

The system first scans a repository for relevant code patterns and historical clues. It then passes selected context to the language model to generate explanations and recommendations.

The idea is to keep the analysis transparent: developers should be able to inspect the evidence behind a finding rather than blindly trust an AI-generated warning.

Note: These are proposed implementation details; retain only the tools and features you actually implement.

Why Does Open Innovation Matter?

Open innovation makes projects like PhantomTrace possible because developers can experiment with models, inspect their behavior, and adapt the entire workflow to their needs.

Using open-weight models can also make local analysis possible without sending source code to a proprietary AI API, depending on the chosen model and setup.

Open-source libraries provide the building blocks, while an open repository allows other

ELITE WEB3 MERN STACK PROJECT WITH RAG AI COONTRAEZIATUION DOCKER

Tanmay Singh — Tue, 20 Oct 2026 18:47:16 +0000

QuietSort: Let AI Clean Up Your Messy Downloads Folder 🦥

What I Built

I built QuietSort, a local-first AI assistant that organizes messy folders without making me sort every file manually.

My Downloads folder used to be a graveyard of random PDFs, screenshots, ZIP files, documents, and mysteriously named files. QuietSort scans a folder, understands what each file is likely to be, and suggests a cleaner filename and destination.

It can group files into categories like Documents, Study Materials, Images, Code, and Archives. Before anything moves, I get a preview of the proposed changes and can approve or reject them.

The idea is simple: let AI handle the boring decisions while keeping the human in control.

Less dragging and dropping. Less decision fatigue. More time for things that actually matter.

Demo

[Insert your deployed demo link or a short video showing QuietSort organizing a test folder.]

The demo should show the app scanning a messy folder, generating suggestions, previewing the changes, and organizing the files after approval.

Code

[Insert your actual GitHub repository link.]

The project is intended to be open source, so anyone can inspect the implementation, experiment with the model, customize the categories, and improve the workflow.

How I Built It

QuietSort is designed around local inference with an open-weight AI model, rather than sending file information to a hosted AI API.

The proposed stack is:

Python and FastAPI for the backend.
Gemma through Ollama for local AI-powered file classification and naming suggestions.
HTML, CSS, and JavaScript for a simple interface.
SQLite or a local JSON log to track approved actions and support undo functionality.

The workflow is straightforward:

The user selects a folder to organize.
The application scans the files and gathers relevant metadata.
The model suggests a category, filename, destination, and short explanation.
The user reviews the proposed changes.
The application moves approved files and records the actions.

I designed the workflow to prioritize safety: files should never be moved or renamed without user approval, and the application should maintain an action log to help reverse changes.

Why Does Open Innovation Matter?

For this project, privacy is not an optional feature. It is part of the design.

File names and documents can reveal personal information, academic work, financial details, and private projects. A local-first approach lets users keep that information on their own machine instead of automatically sending it to an external AI service.

Open-weight models also give developers more control. I can experiment with different models, adjust prompts, inspect the surrounding code, and customize how the assistant makes decisions.

Once the model has been downloaded and the local runtime is configured, the core workflow can operate without an internet connection.

For me, open innovation means being able to build a useful tool without giving up control over the data, the model, or the way the application works.

My Agent Session

[Insert your actual DevRelay agent-session link, if available.]

The session can help show how the project evolved, what implementation decisions were made, and how the workflow was tested.

The Lazy Payoff

The goal is to turn a repetitive folder-cleanup task into a quick review-and-approve workflow.

Instead of manually opening files, deciding where each belongs, renaming it, and moving it, users can review AI-generated suggestions in batches.

I plan to measure the time required to organize the same set of files manually versus using QuietSort. Once I have tested it, I’ll report the actual results here rather than guessing at the time saved.

The real payoff is not avoiding work entirely. It is spending less attention on small, repetitive tasks.

Prize Categories

Best Use of Gemma — only if Gemma is genuinely integrated into the final working project.

top leite web3 mern ai project

Tanmay Singh — Fri, 16 Oct 2026 11:34:16 +0000

SideGig AI — Turn Your Skills Into a Side Hustle

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built

SideGig AI is an open-source AI-powered business assistant that helps freelancers and students turn their skills into realistic side hustles.

Many students know how to code, design websites, edit videos, or create content, but struggle to find paying clients and figure out what services to offer.

SideGig AI helps users:

Discover potential side hustles based on their existing skills.
Generate service packages with suggested pricing.
Create personalized proposals for potential clients.
Draft professional outreach messages.
Build a simple action plan to land their first customer.

The target audience is students, beginner freelancers, indie hackers, and people looking to earn extra income without investing heavily in expensive software.

The business model: A free self-hostable version for individuals, with a potential paid cloud-hosted version for freelancers who want saved projects, proposal templates, and advanced workflows.

The goal is simple: help people spend less time planning and more time finding customers.

Demo

Live demo: [Insert your deployed demo URL]

The demo will showcase a user entering their skills, receiving three personalized side-hustle ideas, selecting one, and generating a client proposal.

Code

GitHub repository: [Insert your public GitHub repository URL]

The project is intended to be open source so developers can inspect the implementation, contribute new capabilities, and deploy their own instance.

How I Built It

SideGig AI is designed around open-source AI rather than depending entirely on a proprietary AI API.

Proposed technology stack:

Frontend: React and Tailwind CSS
Backend: Node.js and Express
AI model: Qwen or Llama-family open-weight instruction model
Inference: Ollama for local development
Data storage: MongoDB
Deployment: Docker, with a cloud deployment for the demo

The application takes structured user input, sends it to the selected model, and converts the model's output into organized side-hustle ideas, service packages, and outreach drafts.

I designed the workflow to separate AI-generated suggestions from user decisions. Users can edit the pricing, rewrite proposals, and choose which recommendations to pursue.

The model and inference layer are intended to be replaceable, allowing developers to experiment with different open-weight models without rebuilding the application.

Why Does Open Innovation Matter?

Open innovation makes AI more accessible to developers who cannot afford expensive API usage while experimenting with a new product.

For SideGig AI, open-weight models offer three important advantages:

1. Greater control: Developers can choose where inference runs and keep sensitive business information on their own infrastructure when configured appropriately.

2. Freedom to experiment: Different models can be evaluated for proposal quality, response speed, and resource requirements.

3. Community-driven development: Contributors can add new business workflows, improve prompts, test models, and adapt the application for different languages and markets.

A closed API could also power this product, but an open architecture gives developers more control over deployment and model selection.

My goal is to build something that beginners can use and that other developers can learn from, extend, and self-host.

My Agent Session

[Insert your actual DevRelay agent session link, if available.]

The session can demonstrate how the application was developed, how the model was integrated, and how the implementation evolved through testing and debugging.

Prize Categories

[Add the applicable partner categories from the official challenge rules after verifying eligibility.]

SideGig AI is an example of how open-source AI can become a practical tool for people trying to build their first independent income stream.

MOST ELITE MERN +WEB 3 STARTUP

Tanmay Singh — Thu, 15 Oct 2026 09:39:16 +0000

MOST ELITE MERN +WEB 3 STARTUP

OPEN SOURCE ELITE PROJECT

Tanmay Singh — Fri, 09 Oct 2026 16:21:48 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built

Demo

Code

How I Built It

Why Does Open Innovation Matter?

My Agent Session

Prize Categories

Developer Peaks

Tanmay Singh — Fri, 02 Oct 2026 11:37:56 +0000

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend