They're fighting over your terminal now

The most interesting price war in tech right now isn't happening in cloud computing or streaming. It's happening in your terminal.

This week, CNBC reported that OpenAI halved its ChatGPT Pro price to $100, creating a new tier between the $20 Plus plan and the existing $200 Pro. Multiple outlets framed the move as a direct response to Claude Code eating into OpenAI's developer base. The promotional sweetener (10x Codex usage through 31 May) is the kind of aggressive discounting you roll out when you're trying to stop a bleed, not grow a market.

Anthropic's response wasn't to match the price cut. Instead, they shipped the Advisor Strategy, an architecture that pairs Opus with cheaper models like Sonnet or Haiku as the executor. The numbers tell the story: Haiku with an Opus advisor costs 85% less per task while trailing Sonnet solo by only 29% in score. On BrowseComp, it more than doubled Haiku's standalone performance. The counter-move to a pricing war is an architecture that makes the pricing irrelevant.

These are two fundamentally different theories of how to win developers. OpenAI is competing on the sticker price: make the subscription cheaper, give more usage, hope the switching costs stay low enough that price wins. Anthropic is competing on the unit economics: make every task cheaper at the API level so the total cost of ownership drops regardless of what the subscription costs.

Where the platforms stand

Then there's where all these agents actually run. Shopify released an AI Toolkit that gives Claude Code, Codex, Cursor, Gemini CLI, and VS Code direct access to live documentation, API schemas, and store management capabilities. MIT-licensed, agent-agnostic, deliberately neutral.

Shopify's move matters because it reveals what the platform layer thinks about this fight: they don't care who wins. When a platform opens its APIs to every competing coding agent simultaneously, it's betting that agent interoperability will matter more than any single agent's dominance. The smart play for platforms isn't to pick a winner. It's to make sure every winner can plug in.

That's the pattern forming here. Model providers compete on price and architecture to capture the developer. Platforms compete by staying agent-neutral to capture the workflow. And the developer sits in the middle, holding real leverage for the first time in this cycle.

I think the pricing pressure only goes in one direction from here. OpenAI's $100 Pro tier will look expensive by autumn. Anthropic's advisor architecture will get replicated, because pairing a powerful model with a cheap executor is too obvious not to spread. And platforms like Shopify will keep opening up, because the marginal cost of supporting one more agent is close to zero while the cost of excluding one is enormous.

The question for anyone building developer tools right now: if the cost of AI-assisted coding drops by 85% in a single quarter, what does that do to every other price in the stack?

---