MakeWithMohit

DO not click the job posting MakeWithMohit is not Notion. company page, "MakeWithMohit," is being exploited to host unauthorized, likely fraudulent job postings in India and South Korea. so hoe does it work !! Scammers exploit a vulnerability in LinkedIn's job posting architecture. When any user creates a basic job post, they are required to type in a company name. LinkedIn's system generates a dropdown list of existing company pages. If the scammer selects "MakeWithMohit" from that dropdown, their fraudulent job post is instantly associated with your company page, displaying your logo and brand name. Scammers often target small, unverified pages because they lack dedicated HR teams monitoring their page activity, allowing the scam to persist longer.

It looks like the ecosystem is shifting its focus from raw capability to the rigorous verification and safety of agentic workflows. 🔬 [Hugging Face Blog] | Deploy local agents everywhere with LFM2.5-2.6B ➲ Liquid AI's new 2.6B parameter model is optimized for edge deployment, providing a viable path for truly private, local-first agentic infrastructure. ⚯ https://lnkd.in/g9YkPDSg 🔬 [arXiv] | CAGE: Certified Authorization under Typed-Return Uncertainty for Tool-Using Agents ➲ This paper introduces a framework for formal certification of agent tool usage, essential for preventing unauthorized actions in production environments. ⚯ https://lnkd.in/gMJHGZGB 🔬 [arXiv] | WitCert: Sound Runtime Risk Observability and Gating for KV-Cache Quantization ➲ A critical look at maintaining system integrity while quantizing KV-caches, offering a gating mechanism to catch errors introduced by precision loss. ⚯ https://lnkd.in/gp5Qm_UW 🔬 [arXiv] | Model or Harness? An Interaction-Centric Taxonomy for Localizing Agent Failures ➲ A pragmatic approach to debugging agent systems by distinguishing between model-side hallucinations and harness-side integration or logic failures. ⚯ https://lnkd.in/gD4gd7Zk 🔬 [arXiv] | Beyond Component Testing: Validating Agentic AI Systems ➲ Moves beyond unit tests for individual modules, proposing an end-to-end validation methodology necessary for the complex, non-deterministic nature of agents. ⚯ https://lnkd.in/gkpEEAbf 🔬 [arXiv] | Tool Specifications Matter: Uncovering and Mitigating Safety Risks in AI Agents ➲ Highlights how imprecise tool definitions are a primary vector for agent failure and suggests structured specification as a key defensive layer. ⚯ https://lnkd.in/gVG6rn37 🔬 [GitHub] | Lisp in 99 Lines of C and How to Write One Yourself ➲ A concise, technical resource on building a Lisp interpreter, demonstrating the power of minimal, clean code abstractions for custom domain languages. ⚯ https://lnkd.in/gsqu-VAG Send it to someone who needs to know.

I built a discovery tool for Canadian Open Data to resolve the limitations of traditional keyword-based searching. The official portal provides substantial resources; however, finding specific datasets requires exact string matching. This wrapper enables natural phrase discovery through an integrated LLM. The tool is available at: https://lnkd.in/gEkeD9Qy Functional Features Semantic Search: An LLM maps natural language queries to relevant metadata. Browser Previews: The interface renders CSV data directly to eliminate unnecessary downloads. Relational Filtering: Users can refine results across large datasets using structural filters. Immediate Access: The system provides direct links for data retrieval. Technical Architecture Database: Supabase manages the Open Data catalog through relational tables optimized for high-speed queries. Intelligence Layer: The LLM processes natural language and performs keyword mapping. Infrastructure: The application uses Vercel for hosting and a modern stack for the data-heavy UI. Project Impact Search Optimization: Semantic search replaces exact-string matching to improve result accuracy. Data Management: The pipeline ingests and maps public datasets within a relational environment. Efficiency: Instant previews and direct downloads reduce the time required to analyze information. #OpenData #Canada #BuildInPublic #Supabase #AI #DataEngineering #NextJS #Vercel #WebDevelopment #100DaysofVibecoding

I am no data scientist and I definitely don’t understand the brain lol. 🧠 But I just spent the afternoon playing with something that feels like the future. I found this project on GitHub called TRIBE v2 (by the team at Meta Research). It’s a multimodal model that can actually "predict" how a human brain would react to a video. I rigged up a script in Kaggle to test it out. I fed it a video about AI Voice Agents From YC pod, and the model processed the text, the audio, and the visuals to map out which parts of the brain would "light up." Took the result added 3D interactive model (check out the screenshots!) showed massive activity in the temporal lobes. From what I can gather, that’s where we process speech and complex language. It’s wild to see an AI estimate how our own biology handles information. I need your help though: For the neuroscientists or AI experts in my network what are we actually looking at here? How accurate are these "in-silico" predictions compared to a real fMRI? I’d love to understand the "why" behind these heatmaps better! Want to try it yourself? If you want the Kaggle code I used to get this running (without fighting the GPU settings for hours like I did), drop a comment or send me a connection request Mohit Johar and I'll share the notebook with you! #AI #Neuroscience #TRIBEv2 #MachineLearning #Kaggle #OpenSource

Most public spending data is technically available but practically inaccessible. I built a project called Saskatoon Spending Story. The idea is simple: public spending should be understandable to regular people, not just analysts who can read procurement spreadsheets all day. This app turns contract data into an interactive civic map so anyone can quickly see: - where money was spent, - who received contracts, - and why non-standard procurement decisions were used. Why I built it: transparency is not just about publishing data, it is about making that data usable. Stack: Vite, vanilla JavaScript, MapLibre GL, and a custom data-processing pipeline. If you care about civic tech, public accountability, or data storytelling, I would love your feedback. https://lnkd.in/gETiux2X #CivicTech #OpenData #DataVisualization #GovTech #Transparency #JavaScript #vibecode #product #uxdesign

Ever scroll past a headline about some crisis halfway across the globe and think, “𝘞𝘢𝘪𝘵, 𝘪𝘴 𝘵𝘩𝘪𝘴 𝘨𝘰𝘪𝘯𝘨 𝘵𝘰 𝘮𝘢𝘬𝘦 𝘮𝘺 𝘨𝘳𝘰𝘤𝘦𝘳𝘪𝘦𝘴 𝘮𝘰𝘳𝘦 𝘦𝘹𝘱𝘦𝘯𝘴𝘪𝘷𝘦 𝘯𝘦𝘹𝘵 𝘮𝘰𝘯𝘵𝘩?” We all get bombarded with big-picture news—stock markets, international tensions, supply chain drama—but hardly anyone connects the dots to our everyday lives. That’s the gap I want to close. So I started sketching out a new system I’m calling The Butterfly Effect. 🦋 Here’s the idea:  Take all the massive, complicated datasets out there and translate them into info you actually care about—like what’s going to happen to your breakfast, your rent, or your next tank of gas. - Try it here : https://lnkd.in/gMJMrNJa  Let me walk you through how this thing ticks:  𝗣𝗵𝗮𝘀𝗲 𝟭: The Catalyst (Global Intake) everything starts with real-time news feeds—think breaking stories, commodity shifts, stuff that moves the global needle. The system grabs signals like a sudden jump in oil prices or a border crisis, then logs the immediate economic reaction.  𝗣𝗵𝗮𝘀𝗲 𝟮: The Ripple (Correlation Engine) Now for the cool part: This is where the magic happens. The engine looks at historical patterns and links those global ripples to other things—so if oil spikes, shipping gets pricier, which means food transportation costs rise, which then trickles into your grocery bill. It even figures out how long it’ll take before you feel it in your wallet.  𝗣𝗵𝗮𝘀𝗲 𝟯: The Tornado (Local Impact) This is the payoff. You pick your hometown—maybe Saskatoon—and, instead of squinting at inflation stats, you get a quick projection of what this means for your next milk run or power bill. No more vague percentages. Just a clear forecast: bread, milk, utilities, all on a simple dashboard, 30 to 90 days out. Basically, the whole point is to make sense of all that nerve-wracking news and turn it into stuff you can actually plan for. Try it here : https://lnkd.in/gMJMrNJa So, to everyone in data and design: What kinds of historical data would you plug into a correlation engine like this? I’m curious how you’d stretch or sharpen the model. #InformationArchitecture #DataVisualization #UXDesign #PredictiveAnalytics #MacroToMicro #TheButterflyEffect #TechForGood #vibecoded #wartieme #military #intelligence #war

Ever wondered which dog name or breed matches your personality? The AuraDog project is now live. It is a personality quiz designed to identify your inner dog based on official data from the City of Zurich Open Data Portal. By analyzing the comprehensive "Hundenamen" dataset—which tracks dog names, birth years, and gender across the city since 2014—this application bridges the gap between public records and interactive personality insights. It is a creative example of how open data can be transformed into an engaging user experience. You can find your inner dog here: https://lnkd.in/gtgr4zB7 Data source: Stadt Zürich Open Data Portal (Hundenamen des Hundebestands ).

TL;DR Built using a real Kaggle dataset of 2026 tech news, Lovable AI AI app that turns boring tasks into overhyped 2026 tech headlines. Key Feature: Uses few-shot prompting from actual sentiment-analyzed headlines to mimic the specific "hype" tone of future tech journalism. "The Disruptor" is a clickbait generator that transforms your most mundane daily activities into "industry-shaking" news headlines. . 🧠 The Engine: Data-Driven Satire This isn't just random text. I utilized the Tech News 2026 Headlines dataset from Kaggle to ground the AI's "voice" in reality. Few-Shot Learning: The app uses 100 real headlines as style examples in the system prompt. Authentic Tone: By shuffling these real-world examples (Wired, TechCrunch, Ars Technica), the AI perfectly mimics the specific vocabulary and sentiment of 2026 tech journalism. The "Disruption Meter": A visual hype-level indicator that measures just how much your "morning coffee" is actually "pioneering a decentralized caffeine-as-a-service paradigm." 🛠️ How It Works Input: You enter a boring task (e.g., "I went for a walk"). Generate: The Lovable AI Gateway processes the request against the 2026 dataset style. Result: You get a glossy, shareable article card with a headline, summary, and fake author. Check it here :: https://lnkd.in/gAtUaPXV Would you like me to generate a sample 2026 headline and summary based on one of your daily tasks to see how it looks?