Arnon Lafer

Many think deleting your AGENTS.MD or CLAUDE.MD is the right move. They're wrong. The problem isn't the file. It's what you put in it. 60 seconds on the tips that actually matter. https://lnkd.in/d3HPChZR #AI #SoftwareEngineering #CodingAgents #AgentsMD #ClaudeMD

1

I was recently listening to a Syntax episode with Scott Tolinski, Wes Bos, and Dimitri Mitropoulos, where, towards the end, they talked about what it took to organize a conference like SquiggleConf - A Conference for Excellent Web Dev Tooling. Dimitri said it perfectly: "There's lot of ways that a conference can go poorly, and there's only a few ways it can go excellently". From the outside, it’s easy to underestimate what goes into a conference. I used to think the same. Pick a venue, get some catering, invite speakers, sell tickets, run the day, tear it down. How hard can it be? It's only one to two days, right? What you don’t see is the amount of unpaid time people pour into trying to create something genuinely special. A space where people can learn, build friendships, make connections, find opportunities, meet people they’ve followed for years, and feel like they belong. That kind of experience doesn’t happen by accident, and it doesn’t scale without care. Community conferences are fragile by nature. There’s no money to be made. Most organizers have full-time jobs. Venues need deposits months in advance. Sponsorship is hard without early numbers. Speaker travel, catering, AV, accessibility, and content quality all compound fast. That’s why I’m genuinely sad that Front Conference Zurich won’t happen this year. A lot of work went into trying to make it happen. I’m deeply grateful to 🤷 Nico Martin for bringing me into the process, and to Thomas Jaggi, who has been tirelessly trying to hold everything together. This has been nearly a year in the making. These things don’t fail because people didn’t care. They fail because caring isn’t always enough. Community conferences only exist if the community shows up. That means buying tickets early when you can. Planning ahead with your learning budget. Showing interest before everything feels “safe.” The difference for you might be small, but for organizers it’s often the difference between viable and impossible. And support isn’t just financial. Feedback is priceless. If the price is too high, say it. If the lineup doesn’t resonate, say it. If the website UX is confusing, say it. Silence is far more damaging than honest input. If we want these events to survive, and more importantly, to get better, we need to help them evolve with our expectations. Reach out. Give feedback. Show up. These spaces are worth protecting. #community #conferences #opensource #webdev #frontend #careerdevelopment #learning #techcommunity

8

Some simple things to help code better with agents: Keep your project at high code coverage and have the agent validate their changes with testing Do not ask it to do huge things, you are the architect, break each large item into small prompts and building steps. Finally, use two different agents such as claude to do the coding then at the pull request use co-pilot to do an additional review. What else? #ai #coding #programming #development

2

Want to know the most common mistake smart engineers make? They optimize a thing that should not exist. We love to polish and refine. But if you spend talent and budget optimizing a process before you’ve done the foundational work, you’re just wasting time beautifully. The efficiency hack comes from the strict order of operations: Step 1: Make the Requirements Less Dumb. (Assume the core instruction is wrong and question it.) Step 2: Delete the Part or Process. (Ruthlessly remove all unnecessary bloat.) Step 3: Simplify or Optimize. (Only now do you polish the critical remaining components.) Never polish garbage! Deleting what shouldn't be there is far more valuable than refining what is already bloated.

1

We at cloudEQ enable faster, safer deliveries by combining standardization and governance to cut manual fixes, elastic agents with self-healing to remove delays, and GitOps with observability to ensure resilient, visible, and secure deployments. #Automation #DevSecOps #GitOps #AIOps #Observability

4

This video shows every iteration of an app I built for my wife while she was expecting our first child. Watching it back, I realized that building software today feels more like managing an organization than writing code. At the time, I was leading a growing team and was determined not to lose my technical edge. So I challenged myself to solve a real family problem in a stack I had never touched before, Android and Kotlin. I spent hours debugging basic UI lifecycle events just to get the thing to function. Pure sweat. If I were making Safish now, I would delegate that manual boilerplate to an LLM. I would set the intent, define the constraints, and provide iterative feedback. I would spend my time on the problem, not the platform. Those are the exact same skills I use as I lead engineering teams. The joy of building is still there. It just feels a lot more like leading than it used to. I have this conversation all the time now, with my team and with the engineers I coach. What does it mean when the craft shifts underneath you? Whether you are managing people or a fleet of AI agents, the core job is converging. Helping people navigate that shift is becoming a bigger part of my own job every day.

7

Modernizing legacy systems has quickly become one of the most impactful use cases for Kavia. More and more of our customers are using Kavia to migrate from older stacks—like PHP—to modern frameworks such as React, or migrating from Java7 to the latest version of Java, with speed, accuracy, and far less disruption. With Kavia, teams can generate all the foundational documentation needed for a successful migration: dependency mapping tables, architectural change summaries, and a detailed modernization plan. Once the plan is ready, Kavia can execute it end-to-end, converting code, refactoring components, and aligning everything to the new architecture. You can even have Kavia validate the modernization by running a full test plan and automatically fixing issues uncovered during test execution. For many companies, legacy modernization has been delayed because of high cost, operational risk, and fear of downtime. But as systems age, maintenance becomes harder, new feature development slows down, and customers begin to feel the impact. With Kavia, you no longer have to choose between building new features and modernizing your stack—you can finally do both.

23

2 Comments