Antonije Mirkovic

My feed has been full of people saying they’ve solved trading with Jev. Jev takes information and specific questions and returns structured answers with probabilities your code can act on. The company built it for decisions inside regular software, like sorting support tickets, moderating content and checking another AI’s answers. Their docs list 24 use cases, and trading isn’t one of them. You can still ask it about Bitcoin and get a probability back. But a probability doesn’t mean the model can predict anything, even a calibrated one, and the confidence score doesn’t check whether those probabilities are right. That’s why a screenshot of a profitable backtest doesn’t prove much. Jev might still help a strategy read news or filings, but that has to be tested on fresh data, after costs. #quant #quantitativefinance #algotrading #trading #math

Comment “BOT” and I’ll send you the template. Putting hours into maths but not seeing results? The problem might be how you’re learning from your mistakes. I built a personal maths tutor using Grok Bot that reviews my handwritten solutions, tracks recurring mistakes, and creates practice worksheets targeting exactly what I need to improve. You can set up your own Bot for any subject and have it continuously adapt to your progress. Comment “BOT” and I’ll send you the template. #grokpartner #mathematics #grokbot #studytips #learnmath

Part 2. Ranking probability distributions by how much useful math you get out of each one, not by how well they fit real data. Same rules as part 1. What matters is what a distribution lets you do once you’ve picked it: whether you can explain what the parameters mean, whether updating and combining stay simple, and how it connects to the distributions you already use. These are less common than the ones in part 1, and some of them turn out to be special cases or building blocks of others. That ended up mattering a lot for the rankings. #probability #statistics #math #mathematics #datascience

Comment «MATH» and I’ll send you the tutorial. If you wanted to teach yourself mathematics from scratch, where would you start, and how would you know you were getting better? I have studied mathematics seriously for 7 years and have coached it for 5 years. These are the three things I’d do if I were starting on my own. One of them is a calculus tool I built in Replit. You move two points along a curve, see the slope between them, and watch the tangent line appear as they come together. I described it in plain English and Replit’s agent built it. The tutorial covers all three steps and has the full prompt to build the tool yourself. Comment «MATH» and I’ll send you the tutorial. @replit #replitpartner #vibecoding #mathematics #calculus #learnmath

You wanted something worth risking #math #studygram #student #university #darkacademia

Ranking probability distributions by how much useful math you get out of each one, not by how well they fit real data. Fitting the data is one question. The other is what a distribution lets you do once you’ve picked it: whether you get closed-form answers, whether sums and averages behave nicely, whether it’s easy to simulate from, and how often it shows up as the limit of something else. A distribution can fit your data better and still give you less of that. #probability #statistics #math #mathematics #datascience