PinnedInMac O’ClockbyShreetej Ghodekar·May 25I Cancelled ChatGPT, Cursor, and Midjourney This Week — My MacBook Pro M5 Max Quietly Replaced All…A no-fluff 2026 guide to running Qwen 3.6 (the new 35B agentic model), ComfyUI, oMLX, and the rest of the open-source AI stack on 36GB of…A response icon60A response icon60
InMac O’ClockbyShreetej Ghodekar·Oct 21Apple Hid a Free AI in macOS 27. It Solved What a 32-Billion-Parameter Model Couldn’t.No download. No account. No internet. I tested Apple’s hidden 3B model against local AI models up to 32B parameters. The results exposed a…A response icon3A response icon3
InData Science CollectivebyShreetej Ghodekar·Oct 12Jev Isn’t an LLM Replacement. Here’s Where It Fits, and Where Open Models Beat It.TypeSafe’s “System One” model answers typed questions instead of generating text. I put it against seven open-source decision models, plain…A response icon1A response icon1
InData Science CollectivebyShreetej Ghodekar·Sep 10I Tested 9 AI Agent Frameworks. Popularity Had Almost Nothing to Do With Performance.It wasn’t a hallucination. Its own specialist agent returned the correct number — and the coordinator overwrote it. No error, no warning…A response icon5A response icon5
InData Science CollectivebyShreetej Ghodekar·Aug 21The AI Wasn’t Smarter. It Just Never Forgot. DeepSeek Harness — 6/6 · 22.4s · 7× fasterIt took the same number of steps. It finished 7× faster anyway — and the reason is a strange fact about how AI actually reads.A response icon1A response icon1
Shreetej Ghodekar·Jul 17I Got Tired of Sketchy Download Sites — So I Built My Own Fastest Video Downloader with Rust &…A response icon1A response icon1
Shreetej Ghodekar·Jul 7DeepSeek Just Quietly Dropped “DSpark” — and It Makes Your AI Chatbot Answer Up to 85% Faster…The Chinese lab that shocked Silicon Valley didn’t build a bigger brain this time. It built a faster mouth — and here’s why that matters…
Shreetej Ghodekar·Jun 28Private RAG, Done Right: Build a Local AI That Never Sends Your Data to the CloudThe other half of a private assistant. Last time we taught a small model how to behave. This time we give it what you know — your docs…A response icon1A response icon1
InData Science CollectivebyShreetej Ghodekar·Jun 28The Complete Guide to Fine-Tuning LLMs Locally (LoRA, QLoRA, Unsloth & Ollama)How a Small, Free AI Can Beat ChatGPT — at One Job You Care About ?A response icon2A response icon2
Shreetej Ghodekar·Jun 14Scared to Put Your Company’s Data Into ChatGPT, Claude, or Gemini? You Don’t Have To.And the per-seat bill for all three — the one that grows with every hire — stops too. A no-fluff guide to owning a fine-tuned open-source…