A friend has taken the code and simply swapped Claude for GPT, X and Gemini. Now it toggles through all four in VS Code. One little tweak, and it gets the best of all of them.
Code built around one lab’s model turned out not to care much which lab it was talking to. Keep the instructions, change the name behind them, and the same work runs on four different machines. Any one of them can have a bad week, a price change or a policy change. The instructions don’t. The toggle is promiscuous about models and faithful to the instructions, and that’s the right way round.
The timing is what makes it more than a neat trick. Also this week: starting April 4 at 12pm PT — 8pm BST — you’ll no longer be able to use your Claude subscription limits for third-party harnesses, including OpenClaw. Instead, they’ll require extra usage. OpenClaw, for anyone who hasn’t met one, is the open-source assistant that, in its own words, “runs on your own computer and meets you in the channels you already use.”
Some context on that. Anthropic gave about a day’s notice, with the change landing “starting tomorrow, Saturday, April 4, 2026, at 12 pm PT/3 pm ET” for anyone on a Claude Pro ($20 monthly) or Max ($100-$200 monthly) plan. Anthropic’s Boris Cherny put the reason plainly: “our subscriptions weren’t built for the usage patterns of these third-party tools.” The same report puts the stakes this way — “a single OpenClaw agent running for one day could burn $1,000 to $5,000 in API costs.”
The arithmetic is easy to follow. A flat monthly fee was never going to survive contact with an agent that runs all day and never sleeps. The fix on the customer’s side isn’t complicated either. If your code can talk to four labs, no one lab’s price is a wall.
Which is exactly what my friend’s little tweak does. Same code, same instructions, four models on rotation, and the best of them for the job at hand.
