PinnedInTowards AIbyManash Pratim, PhD·Dec 1, 2025I Fine-Tuned a 1B Model on My Personal Notes — It Now Thinks Like My Second BrainFor years, my notes were just dead text files.A response icon10A response icon10
PinnedInArtificial Intelligence in Plain EnglishbyManash Pratim, PhD·Nov 9, 2025How These 7 Books Destroyed My Confidence (and Made My PhD Possible)Seven technical monsters that melted my GPU, broke my brain, and somehow made me love research more.
Manash Pratim, PhD·3d agoI Built a Memory Stress Test for AI Porn Chatbots. Here’s What Broke.Can an AI companion remember fictional characters, preferences, intimate conversations, and evolving storylines? I tested three retrieval…
Manash Pratim, PhD·3d agoForget AI Chatbots: The Technology Behind Real-Time Voice-Driven AI Porn and Interactive EroticaHow streaming LLMs, neural voice synthesis, speech recognition, and low-latency inference are changing personalised adult AI.
Manash Pratim, PhD·3d agoAI Porn That Remembers You: Inside the LLM Memory Systems Powering Personalized EroticaHow vector databases, long-term memory, retrieval-augmented generation, temporal knowledge graphs, and personality modelling are changing…
Manash Pratim, PhD·Nov 3Building a Personalised AI Erotica Engine: A Technical Blueprint for LLMs, Memory, and Story…How to move beyond ordinary chatbot prompts and build an intelligent adult-fiction system with persistent character memory…A response icon1A response icon1
Manash Pratim, PhD·Nov 2Inside the AI Porn Revolution: How LLMs and Diffusion Models Are Creating Personalized Erotic…From AI girlfriends and virtual lovers to custom erotic stories and synthetic adult imagery: a deep technical exploration of the large…
Manash Pratim, PhD·Nov 2Your 7B LLM Doesn’t Need More VRAM. It Needs a Better KV Cache Strategy.The mathematics of transformer memory, Grouped-Query Attention, PagedAttention, FlashAttention, and 2-bit KV cache quantization. A deep…
Manash Pratim, PhD·Nov 2Stop Chasing Bigger AI Models. Your 7B Model Is Probably More Capable Than You Think.The next major improvement in your AI application might not require a bigger GPU, a more expensive API, or another billion parameters. It…A response icon3A response icon3
Manash Pratim, PhD·Oct 16I Made Two AI Agents Argue With Each Other Until They Found a Bug Neither Could ExplainI wanted to know whether adversarial AI agents could discover bugs that ordinary code review misses. The solution wasn’t to make them argue…A response icon1A response icon1