SupportNinja

A clean Claude Code session. /context run before typing a character. 29,000 tokens. System prompt: 8.8K. System tools: 11.3K. Skills: 6.6K. Plus custom agents and memory files. None of which I had typed. When I ask Claude to say "hello world," the bytes carrying the message outweigh the message by ten thousand to one. This is the part of the token economy nobody talks about: the tax you're paying that you didn't agree to and probably didn't notice. Article 2 of The Token Economy is up. Want to see what yours weighs? Paste your /context at https://lnkd.in/etyjmYBk → copy your receipt → drop it in the comments. What's quietly running in your context that you didn't put there? #AIcoding #ContextEngineering #AIAgents #TokenEconomy #DeveloperTools

The SaaSpocalypse conversation — the idea that AI signals the end of SaaS — misses how companies actually operate. In real environments: ➜ Systems are deeply embedded in workflows ➜ Data is tied across multiple tools ➜ Replacing infrastructure carries real risk and cost ➜ Teams rely on what already works to serve customers AI doesn’t erase all of that. In fact, it builds on it. And it becomes both a consumer of and a component of SaaS. Instead, we’re using AI to strengthen what we already have and layering intelligence on top. Where are you seeing AI increase the value of systems your organization already depends on? 👀

This has always been true. A lot of engineering criticism comes from having to come back later and fix, extend, or troubleshoot something under pressure. In those moments, everyone has the same reaction: who wrote this mess? In my career as a network engineer, it was usually some version of who wrote these terrible firewall rules, who subnetted this network like this, or who named these VLANs this way. (And sometimes the answer was probably me 6 months or 2 years ago) Over time I’ve made peace with the fact that if something works well enough for its intended purpose, it does not have to be perfect. That does not mean we should stop caring about quality or stop trying to build things well. It just means we should leave room for reality. Perfect is not the bar. Useful, reliable, and fit for purpose usually is.

AI is exposing who actually creates leverage, not just who has vision, not just who can code. The most valuable technical leaders are the ones who can move between layers fast. Engineer to executive. Architecture to operations. Frontline friction to board-level priorities. AI can write code, summarize meetings, and accelerate research. It still cannot replace the person who knows where to focus, what matters most, and when to step in. Close enough to unblock. Detached enough to empower. Technical enough to earn trust. Strategic enough to drive outcomes. That's the job. And it's not going away.