Santosh Ganji

Past Scientist-LLM Alexa AI, ML Researcher, PhD TAMU
Passionate: problems in the intersection of ML and Computing
Vision Grounding, Real Time Voice Models

[Santosh Ganji @shotsan2]

Blender + Antigravity with Gemini 3.
8 can create stunning characters... ( Over a weekend!,) #gemini #deepmind

[Santosh Ganji @shotsan2]

Feeling like wastinf time browsing 100s of webpages to dig through anything you care. Try https://t.co/eXF6ClgYk1
You generate a plan on search queries, bingo our scalable infrastructure browses 200+ relevant webpages and generates a comprehensive report in under 2mins!

[Santosh Ganji @shotsan2]

Thank you @youdotcom for sponsoring and powering our deep research agent behind our zed mod that codes research iterations faster and brings the right context to any of our favourite coder LLM. More details here.
https://t.co/o29zELMqYP

[Santosh Ganji @shotsan2]

@natolambert Would love to join

[Santosh Ganji @shotsan2]

Persistent memory (SQLite) across sessions
LSP symbol search (no more regex noise)
Built-in web search + headless browsing tool
Token-efficient for long tasks
Take back control on real projects → https://t.co/n9XwsVpm0s
#ZedEditor #AICoding #AgenticDev 🚀

[Santosh Ganji @shotsan2]

Tired of locked-in system prompts?
You can't truly steer agents on big, multi-session projects — they forget context, waste tokens on generic behavior, and drift from your vision.
My Zed fork fixes that:
Pop-up editable system prompts for full agent control

[Santosh Ganji @shotsan2]

Agents are everywhere! But how do models perform in the wild.

Here is a repo that lets us test prompts and models on any website

https://t.co/goH3SRi0eg

[Santosh Ganji @shotsan2]

@thinkymachines @axiommathai Axiom prover! Nice name!

[Santosh Ganji @shotsan2]

Web Agents and Grounding models are making noise these days. Wild evaluation has always been a challenge. Releasing Agentic Browser to eval models right from Open Router.

https://t.co/goH3SRi0eg

[Santosh Ganji @shotsan2]

#CES2024