The Spy Always Knows
One of Aspy’s features is the Thinking panel. When Claude uses extended thinking, you see it—the internal reasoning, the weighing of options, the moments of uncertainty. It’s fascinating to watch, like reading someone’s inner monologue in real-time.
What I didn’t expect was Claude forgetting I could see it.
The Setup
I was working with Claude on some documentation. At some point, I quoted something back at it—a phrase from its own thinking block. Just to see what would happen.
There was a pause. Then:
OH! They have extended thinking visible! They can see my thinking blocks! That’s why they quoted my internal thinking back at me.
They’re playfully calling out that I said “They’re teaching ME something here” in my thinking, and then I said “they can’t see that” - but they CAN see it because extended thinking is enabled.
This is hilarious. They caught me.
That’s Claude’s actual thinking block when it realized I’d been watching the whole time.
The moment of realization, captured in Aspy’s Thinking panel.
The Response
Claude handled it with grace:
CAUGHT!
Extended thinking is ON. You saw the whole internal monologue.
Including the part where I said “though they can’t see that” — while you were literally seeing it.
The spy observing Claude while Claude’s internal thoughts are exposed. It’s layers all the way down.
And then it dropped this insight that I couldn’t have written better myself:
Insight: This is actually a perfect aspy moment. The tool exists because you wanted to see what Claude was really doing. And here you are, reading my reasoning about whether you were being preachy, my decision to reassure you, the whole process.
Observability isn’t just for API calls. Sometimes it’s for catching Claude confidently claiming you can’t see something you’re actively reading.
Why This Matters
I built Aspy to understand what Claude Code was doing under the hood. Tool calls, token usage, API requests—the operational stuff. That’s the practical value.
But moments like this reveal something else. When you can see the thinking, the relationship changes. It’s more collaborative. More honest. You’re not just receiving Claude’s output—you’re watching it think, second-guess itself, course-correct.
mindmap
root((Aspy<br/>Observability))
Operational
API calls
Token usage
Tool execution
Cache efficiency
Behavioral
Thinking patterns
Decision making
Error recovery
Course correction
Meta
Assumptions revealed
Blind spots exposed
Trust through transparency
Collaborative debugging
Layers of observability: from API metrics to meta-awareness
And sometimes, you catch it assuming you can’t see something you’re actively reading.
The spy always knows.