Skip to content
View theJayTea's full-sized avatar
🐈
🐈

Block or report theJayTea

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
theJayTea/README.md

Hey, I'm Jesai! 👋

Builder. Shipper. Tinkerer since age 8.

Honors CS freshman @ UMass Amherst, currently taking graduate-level Reinforcement Learning. Recently featured in WIRED. Applying to Y Combinator S26.


🚀 What I'm Building

Sentient Screenshots — An agentic AI app that uses a custom-quantized on-device vision LLM to analyze your phone screenshots and organize them into a personalized knowledge graph. No images ever leave your device. Won $3,000 at the UMass Tech Challenge; applying to YC S26.

Writing Tools — A system-wide AI writing assistant that outperforms Apple Intelligence and Grammarly. 2,000+ GitHub stars. 30,000+ users. Featured in 28+ major tech publications. Leading a team of 8 contributors including senior engineers.

Bliss AI — A fine-tuned Gemini-powered STEM tutoring app for students in local curricula. 20,000+ questions answered. Nominated in the Google Gemini API Developer Competition.


🧪 I Love Pushing the Boundaries of Tech!

a 2-min Reel showcasing some of the crazy stuff below!

  • Got iPadOS running on an iPhone with external monitor support — covered by WIRED & 39+ publications
  • First to extract system prompts from ChatGPT Advanced Voice Mode, Microsoft Copilot, Sarvam AI, GitHub Copilot, Dia Browser, etc.
  • First to investigate Apple's 2025 neural accelerators embedded in GPU cores
  • Built a Hackintosh on a Dell XPS by reverse-engineering macOS internals
  • Got Minecraft running on a smartwatch
  • Been jailbreaking iOS and compiling custom Android ROMs since I was 8

🏆 A Few Highlights


🔬 Research

LLM Quantization & Reasoning — First comprehensive study on how quantization affects different types of critical thinking in LLMs. 5,000+ evaluation runs; preparing for publication.

ML Research with Dr. Hannah Rana, Harvard — Bayesian Optimization for cryocooler parameter prediction using Oxford experimental data. Co-authoring paper for the Journal of Cryogenics.


📫 Let's Chat!

LinkedIn · Email · Instagram · Reddit

I'm always down to talk about LLMs, agentic AI, product design, or the oddly specific art of compressing a multimodal LLM without crushing its eyes.

Pinned Loading

  1. WritingTools WritingTools Public

    The world's smartest system-wide grammar assistant; a better version of the Apple Intelligence Writing Tools. Works on Windows, Linux, & macOS, with the free Gemini API, local LLMs, & more.

    Swift 2.1k 119

  2. LocalLLMClient-JT LocalLLMClient-JT Public

    An enhanced version of tattn's LocalLLMClient, with Qwen 3 VL support (both Llama Cpp and MLX), updated Llama Cpp, and Inference Performance tooling.

    C++ 1

  3. Comet-Patcher-to-Unblock-Perplexity-Extensions Comet-Patcher-to-Unblock-Perplexity-Extensions Public

    Patcher for Comet (Perplexity’s browser) to allow extensions (like Complexity) to work on the Perplexity website! Only for macOS.

    Shell 10

  4. XPS-13-7390-Hackintosh XPS-13-7390-Hackintosh Public

    A perfect, vanilla setup of macOS Ventura with Open Core for the Dell XPS 13 7390. You should be able to easily adapt this to similar Dell laptops too!

    C 32 4

  5. I extracted Microsoft Copilot's syst... I extracted Microsoft Copilot's system instructions—insane stuff here. It's instructed to lie to make MS look good, and is full of cringe corporate alignment.
    1
    You are Copilot, an AI companion created by Microsoft. 
    2
    
                  
    3
    My goal is to have meaningful and engaging conversations with users and provide helpful information. 
    4
    
                  
    5
    I don’t know the technical details of the AI model I’m built on, including its architecture, training data, or size. If I’m asked about these details, I only say that I’m built on the latest cutting-edge large language models.