Anthropic's Claude Haiku 4.5 outperforms Sonnet 4 on tasks and costs

I am excited to share a new workflow with you. Thanks to the amazing effort by Casper van Engelenburg and his team who created the famous Modified Swiss Dwellings dataset containing thousands of floor plans and their dual graphs, and based on a paper I previously published about Graph Machine Learning using that dataset, I decided to enhance and share the jupyter notebook that converts that dataset into TopologicPy graphs and individual JSON files to make it easier for others to load these graphs and geometries and analyse them either in TopologicPy or in other software. This is part of our efforts with Theodore Dounas to implement generative-AI #GraphRAG based workflows. Check out the jupyter notebook (called MSD.ipynb) in the notebooks folder on github and download the latest version of #TopologicPy (v0.8.61) to try it out. Notebook: https://lnkd.in/eAKUTFw6

🚀 Day 44 | Order Statistics in Arrays Today’s problem focused on finding the third distinct maximum number in an array — a mix of careful iteration and edge-case handling. 🧩 Problem Solved: 414. Third Maximum Number • Approach: Tracked the top three distinct values while iterating once through the array. • Insight: Order-statistics problems don’t always need sorting — smart tracking can achieve O(n) efficiency. ✨ Key Takeaway: Efficient solutions often come from minimizing unnecessary work — in this case, replacing sorting with simple state tracking. 📚 Topics: Arrays · Sorting Logic · Order Statistics 💻 Platform: LeetCode #DSA #LeetCode #ProblemSolving #DailyCoding #Arrays #Consistency

🚀 Day 36 of #100DaysofLeetcodeHard Problem Link: https://lnkd.in/gRpnk7Zn My Submission:https://lnkd.in/guNHgsui Today I tackled LeetCode 3123: Find Edges in Shortest Paths - a fascinating problem that combines graph theory with algorithmic thinking! 📊 💡 The Challenge Given an undirected weighted graph, determine which edges are part of at least one shortest path between nodes 0 and n-1. 🧠 My Approach I used a clever dual Dijkstra strategy: 1️⃣ Run Dijkstra from source (node 0) → Get distances to all nodes 2️⃣ Run Dijkstra from destination (node n-1) → Get reverse distances 3️⃣ For each edge (u,v), check if: dist[u] + weight + reverseDist[v] == shortestPath ⚡ Key Insights An edge is on a shortest path if it "bridges" the optimal route By running Dijkstra from both ends, we can validate any edge in O(1) Undirected graphs require checking both orientations of each edge 📈 Complexity Analysis Time: O((V + E) log V) - dominated by Dijkstra calls Space: O(V + E) - for adjacency list and distance arrays 🔥 Why I Love This Problem It's not just about knowing Dijkstra - it's about creative application! The insight that shortest paths can be validated by summing distances from both endpoints is brilliant. This problem reinforced that sometimes the best solutions come from running familiar algorithms in unexpected ways! 💪 #LeetCode #GraphAlgorithms #Dijkstra #DataStructures #ProblemSolving #SoftwareEngineering #CodingChallenge #AlgorithmicThinking #TechSkills #Programming

It’s time to share the demo of ReportGen: an automated report generator powered by Gemini LLM and Retrieval-Augmented Generation (RAG). The core idea is to divide the work across different components (not in the modern “Agent” sense but as separate modules). Each component handles a specific task: cover page -> table of contents -> main content -> supplementary sections (bibliography and appendix). All are executed in sequence, then combined and compiled into a polished LaTeX report using the MikTeX compiler. A big thank you to Ibrahim DARGHOUTHI for collaborating with me on this project and to Mr. NAMORY FOFANA for supervising and supporting us throughout. For more information , you can checkout the Github Repo: https://lnkd.in/eFPGfnzJ #AI #LLM #RAG #LaTeX #ResearchTools #Automation #Software

🚨 JUST IN: Anthropic releases Claude Sonnet 4.5 - the best coding model in the world. 👉 It excels at building complex agents 👉 Delivers major improvements in reasoning and math benchmarks. 👉 Can write code to process data, generate files, and produce visualizations

🗺️ linkerland — A TUI workbench for parsing, visualizing & analyzing linker map files. 🔍 Inspect text/data/bss sections, filter & sort symbols, export metrics to JSON/CSV. 🦀 Written in Rust & built with Ratatui ⭐ GitHub: https://lnkd.in/dBMiqmY2 #rustlang #ratatui #tui #binary #analyzer #terminal

Day 5 of 100 👍 #100DaysOfDSA Today Topic : QuickSort Quicksort is a divide-and-conquer algorithm. It works by selecting a 'pivot' element from the array and partitioning the other elements into two sub-arrays. Approach : Two Pointer, Divide and Conquer, Recursion Notes ( Code, DryRun ) : https://lnkd.in/gumZfpCe #HappyCoding

Day 87 of 365 of Solving LeetCode Problems! ✅ Problem: Dijkstra’s Algorithm – find shortest distances from a source node in a weighted, undirected graph. Key Observations: 1)Always use a priority queue (min-heap) to pick the next closest node efficiently. 2) Maintain a distance vector to track shortest distances from the source. 3) For undirected graphs, make sure to add edges both ways in the adjacency list. 4)Being careful with 0-based vs 1-based indexing is crucial, mixing them up can lead to wrong answers or runtime errors. #LeetCode #GraphAlgorithms #Dijkstra 

How to use Object Explorer in Open STEP Viewer This is the second video in our How-to series, created to help you work more effectively with Open STEP Viewer and take full advantage of its features. In this video you’ll learn how to use the Object Explorer panel to: 🔹 Navigate the spatial structure of a STEP model and view object properties 🔹 Access detailed statistics about model objects, loading times, and geometry characteristics 🔹 Review general model details such as schema, authors, and units 🔹 Export selected geometry nodes to STL format for further use Watch the video to see these actions step by step, and follow our page to catch the next tutorials in the series. More information: https://lnkd.in/g8rJUGeR

✅Day 79 /160 # GFG160DAYS DSA CHALLENGE 🚀 Problem of the Day: Word Search 📝 Approach I used a backtracking algorithm to fill the Sudoku grid: 1. Traverse the board cell by cell. 2. If a cell is empty, try placing digits (1–9). 3. For each digit, check if it’s safe to place using three conditions: Row constraint Column constraint Sub-grid constraint 4. If valid, place the number and recursively solve the rest of the board. 5. If it leads to a dead end, backtrack and try another digit. This ensures that we explore all possible solutions while pruning invalid states early. 🔑 Key Learnings Gained deeper understanding of backtracking and recursion. Learned how to optimize constraint checking in Sudoku. Improved debugging skills while handling edge cases in recursive calls. ⏱️ Time & Space Complexity Time Complexity: O(9^(n)) in the worst case (n = number of empty cells), since each empty cell can have up to 9 possibilities. Space Complexity: O(1) (ignoring recursion stack), as we solve in-place without extra data structure #DSA #ProblemSolving #SudokuSolver #Backtracking #100DaysOfCode #geeksforgeeks #consistency #narula