Henrique Dias

🚨 Backend Devs, You Know This Struggle 🚨 You build the API. You test it in Postman. Everything works flawlessly. ✅ No errors. No surprises. Smooth sailing. But then… 🚨 The frontend team integrates it, and suddenly—𝗯𝗼𝗼𝗺—nothing works. You double-check: 🔸 Request body? Correct. 🔸 Headers? Fine. 🔸 Auth token? Looks good. So what went wrong? 🧠 These are the kinds of bugs that haunt us: • Postman bypasses browser-level CORS restrictions • The frontend might structure the JSON 𝘴𝘭𝘪𝘨𝘩𝘵𝘭𝘺 differently • Content-Type headers could be missing or incorrect • Or maybe… a sneaky middleware is quietly sabotaging the request 📌 𝗧𝗮𝗸𝗲𝗮𝘄𝗮𝘆: Just because it works in Postman doesn’t mean it works in the browser. Welcome to the joys of backend debugging. 😅 Please check 🌍 https://lnkd.in/gfZJqPTR and follow ✅ https://lnkd.in/gZ6myPbn Please follow Chanchal Kumar Mandal #BackendDevelopment #Postman #APIIntegration #DeveloperStruggles #DebuggingTips #SoftwareEngineering #CodingLife #WebDevelopment

114

4 Comments

💡 ECMAScript 2025 is officially out I’ve been reviewing the spec this week, and some of the additions are genuinely useful, especially regarding iterators and sets. JavaScript now supports .map(), .filter(), .take(), and .drop() directly on iterators. No more converting to arrays to do a simple transform. It’s lazy, chainable, and finally feels like it’s catching up to modern expectations. Iterator.from(NodeList) will also clean up many DOM patterns. Combine it with .toArray() and you’ve got a flexible pipeline without losing performance to unnecessary allocations. 🧩 The new Set methods are also solid: ✅ union, intersection, difference—finally native 🧪 isDisjointFrom, isSubsetOf, isSupersetOf 🚀 JSON imports are safer now with { type: "json" }, and RegExp.escape() fixes a 15-year-old annoyance when escaping wildcards. And Promise.try()? Clean way to handle mixed sync/async logic in one wrapper It's one of those releases that feels like it's been a long time coming. 📎 Link to the full writeup in the comments 👇 (I've been writing about most of the features for the last year, subscribe to my newsletter to get cool insights before they become mainstream, link in comments) #JavaScript #ECMAScript2025 #Frontend #WebDevelopment #Standards #DevTools #CareerOS #ES2025

580

14 Comments

I put together a step-by-step guide to help you implement LLM Optimisation the right way for E-commerce. In this carousel, you’ll find the high-level overview. The full version includes: → How to generate a structured product feed (CSV, JSON, XML) with the right attributes: title, GTIN, MPN, price, availability, and product URLs → How to apply Schema . org Product markup using JSON-LD, with full code examples → How to configure your robots.txt to allow GPTBot, OAI-SearchBot and PerplexityBot access to your PDPs and feeds → How to prerender JS-heavy product pages using static snapshots or SSR → Where and how to submit your feed to Perplexity’s Merchant Program and OpenAI’s early access initiative → How to debug AI visibility using server logs, schema validators, and live prompt testing inside LLM tools If you’re working on e-commerce, SEO, or growth, this is where things are moving. 💬 Comment or DM me, and I’ll send you the full guide. 📌 Save this post if you're not ready yet but know you'll need it.

582

915 Comments

🚀 New Release from ComfyUI - Subgraph Publishing & Selection Toolbox Redesign ComfyUI 0.3.63 update unlocks two powerful enhancements that make working with this tool even more fluid and modular: ✨ What’s New ⁉️ 1. 𝗦𝘂𝗯𝗴𝗿𝗮𝗽𝗵 𝗣𝘂𝗯𝗹𝗶𝘀𝗵𝗶𝗻𝗴 You can now save any subgraph as a reusable component in your node library. Use the publish icon or menu in the Selection Toolbox, then find it under Node Library → Subgraph Blueprints. When you need to tweak the behavior, just edit and update it like a regular node. This change elevates composability and reusability in your workflows. ComfyUI 2. 𝗦𝗲𝗹𝗲𝗰𝘁𝗶𝗼𝗻 𝗧𝗼𝗼𝗹𝗯𝗼𝘅 𝗥𝗲𝗱𝗲𝘀𝗶𝗴𝗻 The UI of the selection toolbox has been refreshed with clearer icons and an expandable menu. This redesign isn’t just aesthetic — it paves the way for future enhancements and more flexible workflows. ComfyUI Blog 🔍 𝗪𝗵𝘆 𝗧𝗵𝗶𝘀 𝗠𝗮𝘁𝘁𝗲𝗿𝘀 𝘔𝘰𝘥𝘶𝘭𝘢𝘳𝘪𝘵𝘺 & 𝘙𝘦𝘶𝘴𝘦: Subgraphs can now live as first-class entities—no recreating from scratch each time. 𝘉𝘦𝘵𝘵𝘦𝘳 𝘜𝘟: Cleaner UI and menu structure help speed up experimentation and iteration. 𝘌𝘹𝘵𝘦𝘯𝘴𝘪𝘣𝘪𝘭𝘪𝘵𝘺: The foundation is laid for even richer features down the line. As someone who values efficient, scalable design in creative and engineering workflows, I’m excited to see how this deepens what’s possible in ComfyUI. #ComfyUI #UIUX #subgraphs #ai #monks

98

4 Comments

Have you heard of the new type of UUID? I often use UUIDs (Guid in C#) as unique identifiers in my database. They do have one problem: they're "random", which is good and bad. It's good for distributed systems, because we can generate them from different nodes. But what makes them bad? One thing is size. A Guid in C# is 16 bytes, and this can add up. This means your database tables and indexes will consume more memory. Also, if you're a database geek, you're familiar with index fragmentation. To put it simply, indexes work better with sortable data. Luckily, there's a new type of UUID. It's called UUID V7, and it's a sortable value because it has a time component. You can create one in .NET 9 via Guid.CreateVersion7. It's also natively supported in Postgres 18 (coming out soon). Would you consider using it? --- Sign up for the .NET Weekly with 72K+ other engineers, and get a free Clean Architecture template: https://lnkd.in/ekMyTe3N

437

48 Comments

𝗟𝗼𝗻𝗴 𝗰𝗼𝗻𝘁𝗲𝘅𝘁 does not mean 𝗯𝗲𝘁𝘁𝗲𝗿 𝗿𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 You need a retrieval system: As developers, we often get excited about massive context windows: "Just dump all your data in and get perfect recall and reasoning" But let's be practical: it's not that straightforward. While long context is valuable, it needs to be paired with effective retrieval systems. As you're designing reasoning models or building agents, consider implementing robust retrieval mechanisms alongside long context - this becomes critical when working with large, heterogeneous datasets. What's happening under the hood? LLMs fundamentally struggle to identify and utilize relevant information in extended contexts, especially without explicit keyword matches. Consider these technical insights: • 𝗟𝗼𝗻𝗴 𝗰𝗼𝗻𝘁𝗲𝘅𝘁 ≠ 𝗕𝗲𝘁𝘁𝗲𝗿 𝗿𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 → The NoLiMa benchmark demonstrated that 10 out of 12 models operating at 32K context lengths experienced a 50% performance degradation. • 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 𝗱𝗶𝗹𝘂𝘁𝗶𝗼𝗻 𝗶𝘀 𝗿𝗲𝗮𝗹 → Significant portions of context are often irrelevant and can distract the model from its primary objective. • 𝗥𝗲𝘀𝗼𝘂𝗿𝗰𝗲 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗺𝗮𝘁𝘁𝗲𝗿𝘀 → Extended contexts increase latency and costs. Each token represents an investment in computational resources, budget, and cognitive clarity. The evolution of LLMs isn't simply about expanding context windows—it's about implementing smarter contextual processing. This is why retrieval systems remain essential in our AI architecture. Learn more about RAG and chunking techniques here: https://lnkd.in/gxYCFYzj

127

13 Comments

Did you know that for over 20 years in the US, exporting strong cryptography was treated the same as exporting weapons? Jump into the new Understanding Cryptography section to understand how exactly this is a powerful weapon that is one of the most important pillars of democracy and free speech in today’s world. Although many authoritarian governments have outright banned it, even in democratic countries there are still constant attacks on it. This is a big concern because once cryptography gets weakened, it will no longer be close to a mathematical certainty that your data is safe, but rather a matter of who has enough power and money that can see everything that you do online. In this section, you will learn about some of the essential tools that powerful intelligence agencies, militaries and companies use to protect themselves. We will begin by learning about OTP, the perfect encryption that’s mathematically proven to be unbreakable, and will always be no matter what future technology we come across. We’ll learn and implement it from scratch, and with this foundation, we’ll then move on to a more practical encryption algorithm, AES, and see how exactly it will take trillions of years to break on the most powerful supercomputers. We will then build a secure data encryption application, with which you can encrypt your bank information and give it willingly to hackers, and have peace of mind that no one will currently be able to decrypt it for as long as your key on your machine is secure! We’ll then dive deep into public-key cryptography and see how to establish secure connections over unencrypted traffic. We’ll understand the logic behind RSA and be able to encrypt and decrypt on a piece of paper, and then implement it from scratch in Node.js. This will give us a strong foundation from which we’ll then easily understand digital signatures, certificates, Diffie-Hellman, TLS, HTTPS and more. You will also see how you can easily set up HTTPS offline on your apps with whatever domain you want, like google [dot] com, and get a perfect “certificate is valid” in your browser! You will also finally understand the threat of quantum computers, which is indeed a threat to the internet as a whole, but certainly not to your applications once you are done with this section. If you have always found security and cryptography intimidating, this section will be a great start for you to finally wrap your head around it! Nothing in cryptography is magic, and every single thing makes perfect sense and can be traced on a piece of paper. Once you are done with this section, you will be a more security-aware engineer capable of developing safer systems and even better protecting yourself digitally. Hope to see you there, and thank you very much! #NodeJS #Cryptography #CyberSecurity #AES #RSA #QuantumComputing #TLS https://lnkd.in/gMrf99eb

159

25 Comments

Prove me wrong: 95% of open source eval libs are just abandoned GitHub repos with fancy README files and good marketing! I know because I’ve used them, plus I keep hearing the same story from builders- "We picked [popular eval library] because it's open source. Now we're getting NaN scores for half our metrics and our evaluation has been 'running' for hours on 100 samples. Are we missing anything? No, you're not. You've been sold a lie in the name of "OPEN SOURCE"- ❌ Unmaintained code, documentation that hasn't worked since v0.1.3 ❌ "Community-driven" with zero support when your eval hangs for 8 hours ❌ "Free" until you need expensive APIs to make it function  ❌ Breaks with every model except GPT-4  ❌ "Production-ready" that can't handle 100 test samples without crashing Let me tell you 2 hard truths: Your "free" tool costs more than enterprise software and works worse. And to burst your bubble- open source doesn't mean compromising on quality. What enterprise-grade open source should look like (and why teams move to Future AGI and STAY): ✅ Easy setup. Copy-paste quickstart. Runs in your cloud or local. ✅ Turing models + multimodal evals. Fast, accurate, pinpoint error finds with clear explanations, not fuzzy scores. ✅ Built-in observability. Unified traces, logs, and dashboards from day one ✅ Zero latency impact. Fully async and non-blocking, so evals never slow prod or melt hardware. ✅ Enterprise best practices. Curated metrics, consistent results, and actionable insights- no analysis paralysis. ✅ Broad compatibility + flexible SDK. Works with LangChain/LangGraph/LlamaIndex; supports OpenAI, Azure OpenAI, Anthropic, Bedrock, and local/vLLM. Clean SDK + CLI for custom checks and pipelines. Clear choice- Want to experiment with confidence? Start with our open source version. Need enterprise features? Upgrade seamlessly when you're ready. Your time is too valuable. Your AI is too important. Your standards should be higher. If this post has hit a nerve, comment “me too”/DM me and I’ll help you migrate in minutes. Or kick the tires now: https://lnkd.in/grjB8kvv #AIEvaluation #FutureAGI #AIEngineering #LLMOps #ProductionAI #AIInfrastructure 

35

1 Comment

When developers use "tech debt" to refer to any code that they didn't write (and some they did), how do you decide where to invest? I find this 2x2 quite helpful for framing the discussion. In many situations, the trouble with debt is not the principal, it's the interest. When it comes to technical debt, that's the increase in the cost of change that's created by the debt. And you're not really required to pay until you're working in that area. The axis for "ease of change" includes everything it takes to get a change into production. Is the code easy to reason about? Do you trust the tests? Is the pipeline fast? Anything that makes it harder is bad. However, if the system or component isn't undergoing change, it's not hampering your business as badly. Hence the need to incorporate "desired rate of change." In other words, if we could, how much work would we be doing in this area? If something is slowing you down or blocking you in an area of your system that presents critical opportunities to your business, it's probably time to get it cleaned up. Like any 2x2, it's pretty simple. But I assure you someone, somewhere, is arguing right now about how to improve a thing that's already good and pretty stable. PS. I used this chart in a workshop at LeadDev's LDX3 and it led to some great discussions. If you'll be at LeadDev in NYC or Berlin this year, come have a chat. The workshop will be repeated at both.

61

2 Comments

Nice new addition to Rails- Rails introduces a powerful new feature allowing associations to be marked as deprecated, with sophisticated reporting mechanisms to safely identify and clean up their usage across all environments. Before this enhancement, removing associations like 'has_many' or 'belongs_to' posed risks, requiring developers to delete or modify them and heavily rely on tests or manual checks to catch all usage. This was especially perilous in production settings with incomplete test coverage or mocked associations. What’s new? Rails now includes a ':deprecated' option in association declarations. When marked, Active Record will issue a warning each time the association is used, significantly enhancing the safety and efficiency of deprecating associations before removal. This proactive approach helps prevent surprises and app disruptions. Developers can manage how Rails addresses deprecated associations through three modes: - ':warn' (default): Logs a warning via the Active Record logger, potentially including clean backtraces filtered by 'Rails.backtrace_cleaner' to trace the usage source. - ':raise': Raises an 'ActiveRecord::DeprecatedAssociationError', beneficial for immediate detection of deprecated usage in test and dev environments. - ':notify': Emits an 'ActiveSupport' notification ('deprecated_association.active_record'), ideal for integration with monitoring tools like Sentry or New Relic, offering detailed payloads such as reflection, message, location, and optional backtrace. This behavior is globally configurable, ensuring consistent deprecation handling throughout your Rails application, simplifying deployment and scaling deprecations securely. 🔗 Read the full article here: https://lnkd.in/eKM67Exr

56

1 Comment

Why Experienced Developers’ Jobs Are Secure for the Next X Years TLDR: To build with AI, you still need a solid foundation in coding and an understanding of what’s worked (and failed) in the past. AI will hallucinate, and someone has to bring it back on track. That means you can’t replace senior engineers with AI… at least not yet. *** Today I was troubleshooting an AWS Amplify deployment issue. Claude AI gave me a YAML fix — but it was wrong. I had to step in, correct it, and nudge things in the right direction. Moments like that remind me why predictions about AI replacing senior engineers are overblown. You still need experienced humans who can: 1. Recognize when AI is off course 2. Apply real-world judgment 3. Debug and ship with confidence *** Now, if you’re a junior engineer struggling to land your first coding role, here’s my advice: don’t wait. Start something of your own. Build an AI-powered project or even a small startup. Prove you can scale coding with AI. If you’re really good, you may find you don’t need a job at all. Your idea itself could be worth GOLD.

56

4 Comments

🎉 Just launched: DebugBear RUM now tracks CPU tasks during the initial page rendering process! That means you can see whether main thread processing activity is hurting your page speed, and what specific scripts are slowing down your website. Learn more in our docs: https://lnkd.in/egGmqRd9

37

2 Comments