A free iPhone shortcut asks ChatGPT and Claude the same question and compares them
Poly Duo has both rival assistants answer separately, then shows where they agree and differ, so you know which parts of an answer to double-check.
The short version
- Duo is a free iPhone shortcut that uses your existing ChatGPT and Claude apps.
- Each assistant answers on its own and never sees the other's reply.
- The result has three parts: where they agree, where they differ, and what to verify.
- In the author's test, 72 of the 74 answers both models agreed on were correct.
- That 97% describes agreement in a test, not Duo's measured accuracy.
Why it matters
A confident AI answer can still be wrong, and checking it with a second assistant usually takes several steps, so people skip it. This turns the check into one tap. Disagreement between two companies' models gives you a clear signal about where to look closely before you act.
What the shortcut does
As dev.to reported, Poly Duo is a free iPhone shortcut that runs your own ChatGPT and Claude apps together. You ask once. Each assistant answers independently, and neither sees what the other wrote.
After about a minute you get a short summary with three parts: what both agree on, where they differ, and a brief list of things to verify yourself.
A worked example
The post imagines a job offer that pays more but comes with a long commute. Both assistants agree the move is a step up in career terms. They differ on emphasis: ChatGPT leans toward taking it, while Claude suggests counting the commute as unpaid time and comparing hourly earnings.
The to-check list then suggests trying the commute at rush hour and asking about remote days. The point is not a verdict. It is a better set of follow-up questions.
What the author's tests found
The author, independent AI researcher Vadym Chernets, tested two flagship models on 100 new questions on October 26, 2026. Of the 74 questions where the two agreed, 72 were answered correctly. None of the 100 produced the same wrong answer from both with high confidence.
A separate analysis of 1,459 public questions found that when the flagship models agreed, they were right 98% of the time. A weaker pairing did far worse: they agreed on under a tenth of questions, and were right only 37% of the time when they did.
Read this carefully. The author says the 97% reflects agreement in his test, not the measured accuracy of Duo. These are also the tool maker's own figures, and we have not tested the shortcut.
What you need and what to watch for
You need an iPhone on a recent version of iOS, signed in to both free apps. Paid plans work too, using the models in your plan at no extra charge. The code is public on GitHub. Android users get a different version, which is not the ChatGPT-plus-Claude pairing.
Do not touch the screen while it runs, because a tap interrupts the shortcut. And agreement is not proof: two models can still be wrong in the same way.
What to watch
- Will independent testers confirm the author's agreement-versus-accuracy figures on other question sets?
- How well does the shortcut hold up as ChatGPT and Claude change their apps?
Try it
Pick one answer an AI gave you that you are unsure about, and ask the same question to a second assistant from a different company. Look for the differences. Skip confidential client or customer data.
Sources and links
- Reportingdev.to: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by dev.to · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.