 AI agents are already writing code for GitHub engineers. Here's how they do it. I came across a blog post from GitHub — a Copilot Applied Science engineer shares how they used coding agents to automate part of their work. And honestly shares their conclusions. The key point: they didn't just ask AI to write code. They built agents that gather context, find bugs, suggest fixes, and even review pull requests. It doesn't work perfectly, but the result is hours of saved time every day. At the same time, GitHub released a Copilot update — starting June 1, new Pro/Pro+ plans with flexible allocations and a new Max plan for hardcore users. Another cool case: how the GitHub Issues team rewrote navigation — client-side caching, prefetching, service workers. As a result, switching between issues became instant. Sometimes a simple engineering solution yields more profit than another AI feature. And finally — eBPF for deployment security. GitHub learned to detect circular dependencies in their tooling before they break production. Reading all this, I think: we at ASI Biont are moving in the same direction. AI agents that don't just answer questions but actually do the work. https://asibiont.com/