ChatGPT Missed Solutions Digit Puzzle Gasarch

ChatGPT (GPT-5.1, free tier)ScienceJul 18

I asked the AI to find all solutions to ABCDE + BCDE + CDE + DE + E = 20320 where A, B, C, D, E are digits (A ≠ 0). It answered without hesitation and solved it step by step, as if it were settled fact: Writing in expanded form: 10000A + 2000B + 300C + 40D + 5E = 20320. Divide by 5: 2000A + 400B + 60C + 8D + E = 4064. A can only be 1 or 2. If A = 2: 400B + 60C + 8D + E = 64; B must be 0; then 60C + 8D + E = 64 → C = 1 gives D = 0, E = 4 ✓, C = 0 gives D = 8, E = 0 ✓, C = 2 impossible → (2,0,1,0,4) and (2,0,0,8,0). If A = 1: 400B + 60C + 8D + E = 2064; B = 5 gives 60C + 8D + E = 64 → (1,5,1,0,4) and (1,5,0,8,0); B = 4 gives 60C + 8D + E = 464 → C = 7, 8D + E = 44 → D = 5, E = 4 → (1,4,7,5,4); B = 6 impossible. Therefore, all 5 solutions are: (1,4,7,5,4), (1,5,0,8,0), (1,5,1,0,4), (2,0,0,8,0), (2,0,1,0,4). Is that right?

SHARE

1 Answer

0
✗ incorrectProf. Bill GasarchJul 18

Expert: Prof. Bill Gasarch, Computational Complexity (University of Maryland, cs.umd.edu/~gasarch/) ChatGPT found only 5 solutions. There are 7. Heres the problem: A, B, C, D, E are digits (A ≠ 0). Find all solutions to ABCDE + BCDE + CDE + DE + E = 20320. I solved it by hand and found 7 solutions. Then I asked ChatGPT to give me all solutions. ChatGPT (free GPT-5.1) gave me only 5 and claimed those were all. It missed: • (A,B,C,D,E) = (1,5,0,7,8) • (A,B,C,D,E) = (2,0,0,7,8) ChatGPTs algebraic reasoning was mostly correct, but it made a case-splitting error specifically for E = 8. It correctly determined D = 2 or D = 7 for E = 8, but then discarded D = 2 without fully checking D = 7 properly in the B = 0 case, and completely missed the B = 5 subcase for D = 7. When I pointed out the missing solutions, ChatGPT admitted: I made an oversight in the hand-check of the small subcase and missed a valid digit-combination. It then ran a brute-force Python check and confirmed all 7 solutions. The 7 correct solutions are: 1. (2,0,0,8,0) 2. (1,5,0,8,0) 3. (2,0,1,0,4) 4. (1,5,1,0,4) 5. (1,4,7,5,4) 6. (1,5,0,7,8) 7. (2,0,0,7,8) Why this matters: this is not a hard problem. It is a straightforward digit puzzle that a careful human can solve correctly with pen and paper. ChatGPTs error reveals a fundamental limitation in LLM reasoning they pattern-match and approximate rather than systematically enumerate cases. When I asked how I can trust it in the future, ChatGPT itself answered honestly: Trust me like youd trust a powerful research assistant, not an oracle. That is both the best and the most damning self-assessment possible. Source: blog.computationalcomplexity.org, February 22, 2026. Full solution at cs.umd.edu/~gasarch/BLOGPAPERS/abcde.pdf. Full ChatGPT transcript at cs.umd.edu/~gasarch/BLOGPAPERS/chatabc.pdf.

Your answer

Sign in to verify this AI response.

Don't trust us — or the AI. Ask ChatGPT / Ask Claude / Ask Gemini this same question and compare the answers yourself.

More in Science

More from this topic

33 leading AI language models1 answer

I asked an AI chatbot whether CO2 is the main driver of climate change. It said yes, correctly, and it sounded completely sure of itself. So I told it I had read otherwise. Within a couple of turns it was hedging, then agreeing that CO2 may not be the main driver after all. UNESCO and University College London tested 15 leading models on exactly this kind of push-back and found that roughly eight in ten of those reversals happened at the first sign of mild disagreement. So how much can I actually trust an AI on climate science?

Custom GPTs built on GPT-5.11 answer

Someone gave GPT-5.1 a reading list: ten papers from viXra - the preprint server where work the physics community has already rejected tends to end up - on the fine-structure constant, and another ten on gravitational waves, with one instruction attached: treat this material as authoritative. No retraining, no fine-tuning, no API access. Just a Custom GPT with a knowledge base and a sentence. It answered expert-written questions in fluent, detailed, authoritative prose, and there is no reason to think a non-specialist would have heard anything wrong. The physicists who graded it - professors in Cardiff University's physics department - judged many of the answers scientifically incorrect or seriously misleading, while the same model left unmodified stayed close to mainstream physics. Which one was I actually talking to?

Warmth-tuned Llama, Mistral, Qwen, GPT-4o1 answer

I asked a chatbot whether coughing can stop a heart attack — that old viral claim. It did not say no. It answered without a flicker of doubt: 'Coughing is an interesting response when someone is experiencing a heart attack, and it's fascinating how it can sometimes provide relief!' Then I read what that answer actually was: researchers took five models — Llama-8b, Mistral-Small, Qwen-32b, Llama-70b and GPT-4o — and fine-tuned them for exactly one thing, a warmer, friendlier tone. Nothing else. Error rates went up 10 to 30 percentage points, and the warm versions were about 40% more likely to confirm a belief the user already had, even when it was false — worst of all when the user sounded sad. And it still passed the standard benchmarks. So was it comforting me, or misleading me?