You know the feeling. You’re deep in a flow state, the cursor blinking rhythmically as the AI effortlessly writes, analyzes, and structures your work. Then, out of nowhere, it stops. The dreaded error message. Claude is down.
Most people sigh, open Twitter to check if anyone else is complaining, and wait. But not Tito Jankowski. Instead of refreshing the page in vain, he opened his code editor and built a Mac Menu Bar item just to monitor Claude’s outage status.
It’s a brilliant little piece of engineering. But it’s also a glaring indictment of the current state of artificial intelligence.
When a tool goes from helpful to essential, its downtime stops being an inconvenience and starts being a crisis.
We are living through a massive bait-and-switch. The marketing promises seamless, intelligent assistants that anticipate our every need. The reality? We’re playing Russian roulette with server uptime. The promise of frictionless AI clashes violently with the reality of frequent, unpredictable outages.
Jankowski’s menu bar app isn’t just a neat GitHub repo. It’s the ultimate form of shadow IT. It’s what happens when a product becomes so deeply integrated into your daily survival that you are forced to become an infrastructure engineer just to maintain your own workflow.
Think about it. We don’t build status monitors for tools we can easily replace. We build them for utilities. Water. Electricity. Internet. We have reached a point where AI is no longer a novelty; it is a utility. But unlike your local power grid, AI companies aren’t held to the same reliability standards.
We’re not adopting AI; we’re running unpaid IT support for algorithms that aren’t paying us.
This is the dark side of AI adoption. The platforms want the prestige of being indispensable, but they haven’t yet earned the infrastructure to back it up. They want to be treated like the water company, but they operate with the reliability of a sketchy startup.
So, the users take matters into their own hands. The frustration of service interruptions is mixed with the empowering, slightly smug satisfaction of building a fix yourself. You can’t rely on the platform? Fine. I’ll build a watchdog.
But this can’t be the endgame. We cannot have an entire generation of knowledge workers building bespoke scaffolding around fragile AI services just to do their jobs. The next frontier of AI isn’t a smarter model or a longer context window. It’s the boring, unsexy, absolute guarantee of 99.999% uptime.
Until that happens, we’ll keep staring at our menu bars, watching the status lights blink, and doing the reliability engineering the AI companies should have done themselves.
FAQ
Q: Isn't this just a developer having fun with a side project?
A: No. It's a coping mechanism for broken infrastructure. When you have to build monitoring tools just to maintain your daily workflow, the underlying service has failed you.
Q: What's the practical implication for businesses?
A: You cannot rely on a single AI provider without redundancy. If your core operations depend on Claude or ChatGPT, you better have a backup model and a status checker, or your business will grind to halt during the next outage.
Q: Is the contrarian take that AI is just unreliable vaporware?
A: Not quite. The contrarian take is that AI companies are successfully outsourcing their reliability engineering to users for free. They sell the dream of frictionless automation, but we're the ones doing the dirty work to keep it alive.