The Pivot Is Dead, Long Live the Pivot

Amp, née Sourcegraph, declaring that the coding agent is dead and killing their VS Code and Cursor editor extensions on March 5th. The argument: models have gotten so good that the agent wrapper — the prompts, the tool orchestration, the IDE sidebar — no longer matters. A bash shell is enough. The editor is a cage. Set them free.

The observation underneath is sound. Each new model generation does compress the value of agent scaffolding. The clever prompt chains that differentiated Product A from Product B six months ago now look like unnecessary complexity when the model can just… figure it out. Anyone building in this space should be honest about that.

But there's a difference between a genuine insight and a convenient one. Amp isn't adding capabilities here. They're removing an editor integration — the feature most users of coding agents actually rely on — and wrapping it in frontier-chasing rhetoric so it reads like boldness rather than retreat. "We're unshackling these models from the editor" is a hell of a way to say "we're dropping our most accessible interface."

The tell is in the logic. If the agent layer truly doesn't matter anymore — if any wrapper gets good results — that's an argument against paying for Amp, not for it. The piece never addresses this. Instead it pivots to gauzy language about arrows and frontiers and self-destruction, which is exactly what you write when you don't have a concrete answer to "so what are you building next?"

One line buried in the middle is doing more work than the rest of the piece combined: "We could say yes to more VC money and sell $2,000 worth of tokens for $20 a month." That's a direct shot at the subsidised-usage model that most coding agent startups are running. It's the most honest sentence in the post, and it tells you this is about business model pressure as much as technical vision. Perhaps more.

The self-destruct countdown timer is a nice touch, though. Points for theatre.