Marcos Cattaneo’s Post

If you’re still learning to code from scratch in 2026, you’re already behind. Build with AI models instead. #claude #css #html #javascript #code #ai

As many of you might know by now, I’m building a JavaScript engine for .NET using 100% AI-generated code. It’s gone way beyond what I initially thought was possible. Out of roughly 93,000 ECMAScript 262 tests, I’m down to about 500 failing. Everything I expected to be difficult is already solved. And what’s left? what are those 500 failing tests? It’s almost comical; ** Regex **. The .NET regex engine simply can’t handle everything the ECMAScript spec requires. Variable-length lookbehind being the main offender, plus a few other edge cases. Right now, I’m transpiling JavaScript regex into .NET regex syntax. That works for almost everything, except for those limitations. So now I’m at a crossroads. I could stop here. Point proven. AI can build something like this. Ship it, move on, build something actually useful. Or… How hard can it be to build a regex engine? 😄 Surely it can’t be harder than building an entire JavaScript engine. Right? What would you do?

Let’s recap what we know in JavaScript! These days, my feed is full of “use this AI” and “use that AI.” While AI is powerful, we often overlook the fundamentals that actually drive our day-to-day work and what truly gets tested in interviews. Before jumping to tools, it’s important to strengthen the core. Because at the end of the day, AI can assist you but it can’t replace your understanding of JavaScript fundamentals. From closures, hoisting, and promises to async/await, event loop, and this keyword these are the building blocks every developer should be confident in. I’ve attached a PDF below let’s go back to basics and explore JavaScript the right way. For more insightful content checkout below: 🟦 𝑳𝒊𝒏𝒌𝒆𝒅𝑰𝒏 - https://lnkd.in/dwi3tV83 ⬛ 𝑮𝒊𝒕𝑯𝒖𝒃 - https://lnkd.in/dkW958Tj 🟥 𝒀𝒐𝒖𝑻𝒖𝒃𝒆 - https://lnkd.in/dDig2j75 or Priya Frontend Vlogz #JavaScript #WebDevelopment #CodingBasics #Frontend #LearnToCode #Programming #Developers

In the age of AI, do we really need JavaScript on the backend? Honestly? The case is getting harder to make. 𝗡𝗣𝗠 𝗩𝘂𝗹𝗻𝗲𝗿𝗮𝗯𝗶𝗹𝗶𝘁𝗶𝗲𝘀 The average Node project carries hundreds of transitive deps. Supply chain attacks are real: event-stream, left-pad, XZ utils. You're not auditing code, you're trusting strangers. At scale, that's a security posture. 𝗥𝘂𝗻𝘁𝗶𝗺𝗲 𝗦𝘂𝗿𝗽𝗿𝗶𝘀𝗲𝘀 GC pauses under load. Unhandled promise rejections that silently swallow errors. CPU-bound work blocking the event loop. These aren't hypotheticals. They're oncall tickets. 𝗧𝘆𝗽𝗲𝗦𝗰𝗿𝗶𝗽𝘁'𝘀 𝗙𝗮𝗹𝘀𝗲 𝗣𝗿𝗼𝗺𝗶𝘀𝗲 Types vanish at runtime. `any` leaks in from third-party libs. TSConfig sprawl in a monorepo becomes a maintenance burden. It's better than raw JS, but it's not a type-safe runtime. 𝗗𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝗰𝘆 𝗛𝗼𝗶𝘀𝘁𝗶𝗻𝗴 𝗶𝗻 𝗠𝗼𝗻𝗼𝗿𝗲𝗽𝗼𝘀 Phantom dependencies. Works locally. Fails in CI. Yarn hoisting puts a package in node_modules that was never declared, until it doesn't. pnpm solves it, but the fact that this problem exists at all is telling. 𝗣𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲 𝗚𝗮𝗽𝘀 -> Go: 5-10x on CPU-bound workloads. Single binary. Near-instant cold starts. -> Rust: No GC. Predictable latency. Wins on anything compute-heavy. -> Java (GraalVM): Decades of JVM tuning. Native image closing the startup gap. -> Ruby 3 (YJIT): Faster than people think. Rails ergonomics still elite. Node isn't uncompetitive everywhere. But it's rarely the best choice. It's usually just the familiar one. And in an AI-assisted world where you can scaffold a production Go service from a prompt in minutes? Familiarity is no longer a moat. Are you using Node on the backend because it's right for the job, or because it's always been there? #NodeJS #JavaScript #BackendEngineering #GoLang #Rust #EngineeringLeadership #TypeScript #AI

Smart Summarizer, an AI-powered text summarization web app built using Flask and JavaScript. I wanted to go beyond tutorial projects and build something that felt real. CORS errors, async debugging, and midnight bug fixes it was all part of the process. Every error taught me something I could not have learned any other way. Features: A fully functional Flask backend with dedicated routing and JSON communication CORS configured to handle cross-origin requests between frontend and backend Dual-layer input validation on both client and server side Rate limiting to throttle requests to 10 per minute Real-time word count tracking on both input and output fields Async request handling with loading state and error feedback Automatic sentence capitalization applied to summarized output 🔧 This project helped me learn: How to structure a Flask backend and handle routing How to integrate third-party AI APIs into a backend The importance of validating input on both ends How CORS works and why it matters Frontend-backend communication Secure API key management with environment variables https://lnkd.in/df97hbVf Demo: https://lnkd.in/dMTPHT8z #Python #Flask #JavaScript #WebDevelopment #FullStack  #MachineLearning #NLP #HuggingFace #SoftwareEngineering  #BackendDevelopment #API #BuildInPublic  #CStudent #Programming #100DaysOfCode

I built an 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁 that reads your code and decides, 𝗼𝗻 𝗶𝘁𝘀 𝗼𝘄𝗻, what kind of help it needs. 𝗖𝗼𝗱𝗲𝗦𝗲𝗻𝘀𝗲 - 𝗔𝗻 𝗔𝗜-𝗽𝗼𝘄𝗲𝗿𝗲𝗱 𝗰𝗼𝗱𝗲 𝗮𝘀𝘀𝗶𝘀𝘁𝗮𝗻𝘁 𝗮𝗴𝗲𝗻𝘁 that accepts a code snippet and a natural language instruction, reasons about the intent, and autonomously selects and runs only the relevant analysis tools.   The six tools it can call:  • Bug Detector  • Security Vulnerability Checker  • Code Quality Analyzer  • Code Explainer  • Unit Test Generator  • Refactor Suggester Each tool runs on a focused 𝘀𝘆𝘀𝘁𝗲𝗺 𝗽𝗿𝗼𝗺𝗽𝘁 engineered to return strict structured JSON - making the output predictable, parseable, and renderable as distinct UI components. 𝗛𝗼𝘄 𝗱𝗼𝗲𝘀 𝗶𝘁 𝘄𝗼𝗿𝗸? A 𝗟𝗮𝗻𝗴𝗖𝗵𝗮𝗶𝗻 agent powered by 𝗚𝗼𝗼𝗴𝗹𝗲 𝗚𝗲𝗺𝗶𝗻𝗶 reads both the code and your instruction, reasons about your intent, and selects only the relevant tools from a pool of six. It does not run all six blindly. It picks what makes sense, executes those tools, observes the results, and decides if anything else is needed before giving you a final answer. That decision-making loop is what makes it an actual agent. 𝗧𝗲𝗰𝗵 𝘀𝘁𝗮𝗰𝗸: Python, FastAPI, LangChain, Google Gemini API, React.js, Tailwind CSS - deployed on Render and Vercel. Live link and GitHub Repo in the comments. #AI #LangChain #GenerativeAI #AgenticAI #Python #FastAPI #React #FullStackDevelopment #LLM #MachineLearning #OpenToWork 

🤖 𝐀𝐈 𝐚𝐠𝐞𝐧𝐭𝐬 𝐝𝐨𝐧'𝐭 𝐡𝐚𝐥𝐥𝐮𝐜𝐢𝐧𝐚𝐭𝐞 𝐢𝐧 𝐩𝐚𝐫𝐚𝐠𝐫𝐚𝐩𝐡𝐬 𝐚𝐧𝐲𝐦𝐨𝐫𝐞. 👻 𝐓𝐡𝐞𝐲 𝐡𝐚𝐥𝐥𝐮𝐜𝐢𝐧𝐚𝐭𝐞 𝐢𝐧 𝐉𝐒𝐎𝐍. Last week, I spent 2 hours debugging why my agent kept failing a tool call. GPT-5.2 output v1: "score": "42" GPT-5.2 output v2: "score": 42 Switching keys or types can really throw off your whole workflow—that's just the tame example. If you're working in the AI world, you’re probably all too familiar with the real challenges: comparing two LLM function calls side by side, checking what changed in your RAG retrieval payload, debugging why your API v2 returns an extra nested field, or reviewing agent memory before and after running a tool. We tend to handle these tasks the same old way—opening two VS Code tabs, squinting, scrolling, and often missing that one crucial detail. That’s why I created the tool I really needed: 𝐉𝐒𝐎𝐍 𝐃𝐢𝐟𝐟 𝐕𝐢𝐞𝐰𝐞𝐫. It’s open source, instant, and accessible without any login. Just paste JSON A and JSON B, hit compare, and it shows you exactly what’s new in green, what’s been removed in red, what’s changed in yellow, and any type changes in orange (this last one has saved me multiple times!). It’s exactly like in the screenshot—side by side, clear, distraction-free. I built this because JSON is the most reliable way we have to understand the contracts between models, tools, and APIs in agentic systems—but we’ve lacked good tools to see what actually changed. It's free, open source, and works right in your browser. If you’re building agents, LLM apps, or working with APIs every day, give it a try—you'll wonder how you ever managed without it. Think of how many hours you might have lost last month to a missing comma or a confusing string versus a number. Find the link in the first comment, or comment "𝐃𝐈𝐅𝐅" and I'll DM it to you. #buildinpublic #opensource #aiagents #llm #developers #javascript #python #claudecode #anthropic

Attended the RVA.js meetup on “Building a New Generation of JavaScript ML” at Unboxed Training & Technology. Great overview of tools like TensorFlow.js, Danfo, and Synaptic.js, and how JavaScript is expanding into machine learning and data applications. Events like this are helping me connect learning with real-world tech and continue growing in this space. #MachineLearning #JavaScript #TechTransition #Networking #RichmondVA

FFmpeg 8.0 shipped a native Whisper filter. That means auto-generating subtitles in the same command as your transcode. One invocation, no separate API call, no Python wrapper. ``` ffmpeg -i input.mp4 \ -af "atranscribe=model=whisper-large:language=en:task=transcribe:output_filename=subs.srt" \ output.mp4 ``` For anyone who has built subtitle pipelines before: this cuts out a whole step. Whisper was already the best open-source transcription model. Wrapping it in a shell script to run alongside FFmpeg always felt like a hack. The native filter is cleaner. Two things to know before you get too excited. Most pre-built FFmpeg binaries do not include the Whisper filter yet, you need a custom build. And model download on cold start is slow, so if you are running this in a serverless function, plan for it. Worth thinking about what this does to the "add subtitles to video" SaaS space. A lot of those tools are just Whisper + FFmpeg duct-taped together. The moat shrinks when the underlying stack improves. FFmpeg 8.1 "Hoare" also dropped in March, with D3D12 hardware encoding, Vulkan ProRes support, and new scale filters. The toolchain is moving faster than most teams track. A lot of people building video products are still on 4.x. We run 8.1 in the cloud at RenderIO. Send the command, get the output. No install, no builds, no cold starts. renderio.dev/docs #FFmpeg #VideoProcessing #AITranscription #DeveloperTools #VideoAPI

Anthropic accidentally leaked Claude Code's entire 500,000+ line TypeScript source code through an npm package that included a development source map file. This ironic leak from the "safety-first" company revealed that their AI coding assistant is essentially sophisticated prompt engineering rather than revolutionary technology, along with unreleased features and anti-competitive tactics. A few points: 1. The leak happened through a 57MB source map file accidentally included in npm package version 2.1.88, possibly due to a bug in Bun.js 2. Claude Code is built with 11 processing steps but relies heavily on hard-coded prompts and guardrails rather than advanced AI architecture 3. Anthropic implemented "anti-distillation poison pills" - fake tools designed to sabotage competitors trying to copy Claude's outputs 4. "Undercover mode" instructs Claude to hide its AI identity in outputs, potentially to make AI-generated code appear human-written 5. The system uses basic regex patterns to detect user frustration through keywords like "damn" and "balls" 😂 6. Leaked code revealed unreleased features including "Buddy" (AI companion), "Chyus" (background agent), and references to future models 7. The open-source community quickly created "Claw Code" - a Python rewrite of the leaked TypeScript code