SCSuny Choudhary·2d ago30Multi-LLM Systems Feel Safer. In Production, They DriftMost teams think adding multiple LLMs makes their system more reliable. In production, it often does the opposite. Each model behaves differently.Different safety filters, different context handling, Join discussion
SCSuny Choudhary·Apr 2222Most AI agents don’t fail because they’re dumbI’ve been noticing something while working with different agent setups. Most failures don’t come from the model doing something “stupid.” They come from the system quietly going off track. Not a crashSSeedium and 1 more commented
SCSuny Choudhary·Apr 14721Most AI agent problems I’ve seen aren’t model issuesI keep seeing people blame the model when something breaks. In most cases, that’s not where the problem is. From what I’ve seen, things usually fail somewhere else: agents pulling in too much or wronOAAAAonchainintel and 20 more commented
FFaris·Apr 726AI mental fitnessWe’re starting to build prototypes for an AI focused on real time mental fitness, supporting people during actual moments of stress and focus. Right now, we’re thinking deeply about context awareness,YAPSaleha and 5 more commented
AMAbrar Mohtasim·Apr 700I'm looking for AI Engineering opportunitiesI'm looking for AI Engineering roles (Agentic AI, Workflow automation, Applied AI) About me: I build production-grade AI systems for high-stakes domains where hallucinations are unacceptable. RecentlyJoin discussion
ATAmin Tai·Mar 1611I built a tool to replace Fakespot. Here's what I learned about LLM verdicts vs scores.When Fakespot shut down in July 2025 I started building reviewai.pro — paste Amazon URL, get a BUY/SKIP/CAUTION verdict in 10 seconds. The most interesting engineering problem wasn't the data pipelineNNube commented
AJApurv Julaniya·Mar 1341LLM don't understand words. They understand tokensMost developers think LLM intelligence comes from billions of parameters. But the real mechanics start much smaller — with tokens. Tokens are converted into embeddings and processed through attention AApurv commented
NCNube Colectiva·Mar 400I'm developing a 2D virtual office with AI Agents !They can currently create Word documents. The Excel document creation feature is being improved. I'm using LangChain, Next, and LLM with 4B parameters. Which database would be good for this project? IJoin discussion
NONina Okafor·Feb 2640Can someone explain when you'd actually fine-tune vs just prompt engineerI've been shipping RAG + prompt engineering for most of my LLM work and it's been fine. But everyone keeps saying "yeah you really need fine-tuning for production" and I genuinely don't get the tradeoJoin discussion
MTMaya Tanaka·Feb 2654Built a RAG pipeline for our app, the obvious architecture was wrongStarted building a straightforward RAG setup for customer support queries. Figured we'd do: embed query, vector search, feed top results to LLM, done. Shipped v1 in two weeks. Ran into immediate issueASNAlex and 3 more commented
