2025W31

A few interesting articles I read over the past few days

• No, AI is not Making Engineers 10x as Productive — AI does not make engineers ten times more productive because much of coding involves thinking, reading, and preventing unnecessary work. AI helps with small tasks and gives short bursts of speed, but it often encourages haste and over-building. True productivity comes from experience, good habits, and teamwork, not just using AI tools.
• All the cool kids are doing it — The author is unsure about using large language models (LLMs) for coding because they find them unreliable and costly. They feel LLMs don’t yet help much with complex or performance-focused programming tasks. However, they see some promise in LLMs for other uses like research help and code explanation.
• The Sunday Morning Post: Why Exercise Is a Miracle Drug — Exercise improves many parts of the body and is better than any medicine. Studies show exercise helps cancer patients live longer and stay healthier.
• Full-breadth Developers — Full-breadth developers who combine technical skills with product sense are thriving with new AI tools. Many companies struggle because they separate product and engineering roles too strictly.
• Ramblings — Remote teams of 2-10 people can use personal “ramblings” channels to share thoughts without cluttering group chats. These channels help team members stay connected with short, informal updates and encourage creativity. Ramblings reduce interruptions and support social bonding, especially when no meetings are scheduled.
• AI is eating the Internet — The Internet used to offer free content funded by ads from companies like Google and Facebook. Now, AI tools reuse content without sending visitors to original sites, hurting creators and changing how the web works. In the future, AI will control access to information, balancing ads, content makers, and users in new ways.
• Slow — Some problems take humans many years or even centuries to solve, like building cathedrals or proving complex math theorems. Long-term projects need strong institutions and steady effort to keep going over time. Examples include scientific studies, historic buildings, and technology that may last for thousands of years.
• Fast — Fast software changes how we work and feel by making tasks quicker and easier. It requires focus and simplicity, often cutting unnecessary features to deliver speed. As current LLM technology improves, speed will unlock new possibilities and transform our lives like never before.
• Why I’m stepping down as Head of Engineering — The author is stepping down as Head of Engineering to focus on hands-on technical work they love. As the company grew, their role became more about management, which made them feel less connected and impactful. Stepping aside allows new leaders to scale the organization while the author contributes where they add the most value.
• My favorite use-case for AI is writing logs — JetBrains created a small, fast AI model that helps developers write better log lines automatically in PyCharm. This model runs locally on your computer, making coding easier and debugging faster. It shows that focused, lightweight AI can be very useful alongside big, general models.
• ​How magnesium affects your sleep and anxiety — Magnesium may help reduce mild anxiety but does not strongly improve sleep for most people. Many people do not get enough magnesium from their diet, and supplements can be useful, especially magnesium glycinate.
• Coding with LLMs in the summer of 2025 (an update) — Large language models like Gemini 2.5 PRO help programmers by finding bugs, speeding up testing, and supporting design decisions when humans guide them carefully. To get the best results, programmers must provide clear context and stay actively involved in the coding process instead of relying on AI alone. Right now, working together with AI is more effective than letting AI work solo, but this may change in the future.
• Why me? — When bad things happen, we often ask “Why me?” and feel singled out. But life brings both good and bad luck to everyone, and neither is guaranteed. Accepting this helps us feel less angry and more grateful for the good moments.
• MCP: An (Accidentally) Universal Plugin System — I used to think MCP added nothing over APIs. But this post clicked: APIs exposes services, MCP wraps actions, there are mores actions than calling an HTTP service: CLI, scripts, anything runnable. It’s a universal interface for doing, not just calling. Surprisingly powerful.
• Making Sense of a Noisy World — Trends come and go, but sharing thoughtful opinions helps us learn and grow. Honest views can challenge ideas and create meaningful conversations. Careful thinking makes our opinions valuable and helps us understand the noisy world.