--- title: "Why StreetCred Doesn't Hand Out Solution Workbooks" canonical: https://streetcred.finance/why-we-dont-share-solution-workbooks author: "Lewis Grafton" datePublished: 2026-10-17 dateModified: 2026-10-17 --- # Why StreetCred Doesn't Hand Out Solution Workbooks Canonical URL: https://streetcred.finance/why-we-dont-share-solution-workbooks Learners often ask for the gold solution to a case. The research shows that working the problem yourself, with feedback instead of answers, builds more skill. ## Short answer Because seeing the finished answer feels like learning but produces less of it. Controlled studies of university physics students, student coders and high-school maths classes point the same way: people who work the problem themselves, with hints and explanations, learn more than people who are handed solutions, even though they feel they learned less. StreetCred's report tells you which check failed and why, not the number to type in, so the next rep is still yours to build. ## The request we hear most After a hard case, the most common message we get is some version of: can I see the gold solution? It is a fair thing to ask. You have spent an hour on a workbook, the balance sheet will not balance, and a finished model would show you in seconds where you went wrong. We say no on purpose. Partly because the answer key is what every submission is scored against, so a solution file in circulation would make every score meaningless. But mostly because the research on how people learn is unusually consistent here: handing over the answer is the fastest way to finish a problem and one of the slowest ways to get better at it. ## Feeling like you learned is not the same as learning In a randomised study of introductory physics at Harvard (Deslauriers et al., PNAS 2019), students were taught the same material either by a polished lecture or by working the problems first, with the instructor stepping in afterwards. On the test, the problems-first students scored about 0.46 standard deviations higher. Asked how much they felt they had learned, they rated it about 0.56 standard deviations lower. The authors put this down to effort: struggling with a problem feels like not understanding it, while following a clear explanation feels like mastery. A solution workbook is the modelling equivalent of the polished lecture. Every formula makes sense as you read it, which is exactly why it feels so useful, and why it so rarely survives contact with a blank sheet. **Chart: Problems first, compared with lecture** | Group | Value | | --- | --- | | How much they felt they learned | −0.56 SD | | What they learned (test) | +0.46 SD | Source: Deslauriers et al., PNAS 2019. ## What happens when the answers are one click away A field experiment with close to a thousand high-school maths students (Bastani et al., PNAS 2025) tested this directly. One group practised with no AI, one with a standard ChatGPT-style assistant that would simply give answers, and one with an AI tutor set up to give hints and never the answer. During practice the answer-giving assistant looked like a win: mean practice scores rose from 28.4% with no AI to 42.1%, and the hint-only tutor did better still at 64.5%. Then the AI was taken away for the exam. The no-AI group scored 32.1% and the hint-only tutor group 31.7%, about the same. The group that had been given answers scored 26.7%, roughly 17% worse than students who had never had the tool. The authors' reading: without guardrails, students used the assistant as a crutch and then performed worse on their own. That is the pattern a solution workbook sets up. Your practice output looks better, because you can copy the structure. Your performance in the room, where there is no solution file, gets worse. **Chart: ChatGPT in a high-school maths class: mean score** | Group | Value | | --- | --- | | Practice, with AI: No AI | 28.4% | | Practice, with AI: ChatGPT (gives answers) | 42.1% | | Practice, with AI: Hint-only tutor | 64.5% | | Exam, without AI: No AI | 32.1% | | Exam, without AI: ChatGPT (gives answers) | 26.7% | | Exam, without AI: Hint-only tutor | 31.7% | Source: Bastani et al., PNAS 2025. ## Ask for explanations, not solutions The tool is not the problem; how it is used is. In two pre-registered experiments with students learning to code alongside a large language model (Lehmann, Cornelius & Sting, 2025), overall learning was unchanged, but it split sharply by behaviour. Students who used the model to generate solutions covered more ground and understood each topic less. Students who used it to ask for explanations understood more. By the study's estimates, each request for a full solution went with roughly 0.31 fewer correct answers on the later test, and each request for an explanation with roughly 0.25 more. So the line we draw is not between help and no help. It is between help that does the thinking for you and help that makes your own thinking better. **Chart: Change in correct test answers, per request to the AI** | Group | Value | | --- | --- | | Each request for a solution | −0.31 | | Each request for an explanation | +0.25 | Source: Lehmann, Cornelius & Sting, 2025. ## Why this matters more for a modelling test A modelling test does not ask whether you recognise a good model. It gives you a brief, a blank forecast and a clock, and asks you to produce one. The skills it rewards, such as knowing which schedule feeds which, catching a sign error before it breaks the cash flow, and deciding where a circularity belongs, are built in the moments where you are stuck and have to work out why. Reading a completed workbook skips exactly those moments. You see that interest links to average debt; you do not have to discover that linking it to closing debt creates a circular reference. The insight arrives without the struggle that would have made it stick. ## What StreetCred gives you instead of the answer - A score report that names each check you missed, the cells it looked at and feedback on what went wrong, without the value it expected, so you know where to look but still have to fix it yourself. - Readiness by module, so you can rebuild the one schedule that broke rather than the whole workbook. - Links from the report to the guide that covers the topic, so you can learn the concept, close the guide and apply it in your own file. - Case quizzes that hold back the correct answers and working until you have used your credited attempts or choose to reveal them, so the first try is always your own. - The chance to rebuild and resubmit, which is where the learning in this research actually happens. ## How to get unstuck without the answer - Start from the report: find the module with the most deductions and fix that before anything else. - Trace one number: pick a wrong output and follow its precedents back until you find the first cell that is not what you intended. - Rebuild the schedule from blank rather than patching it. A clean second attempt teaches more than ten small edits. - If you use an AI assistant to study, ask it to explain a concept or check your reasoning, not to write the formula for you. That is the explanation-not-solution behaviour the research rewards. - Say your logic out loud, as you would to an interviewer. If you cannot explain why a line links where it does, that is the gap to close. ## Frequently asked questions ### Will StreetCred ever share the solution workbook? No. The answer key is what your submission is scored against, so sharing it would undermine every score on the platform, including yours. More importantly, the research above suggests it would make you a slower learner, not a faster one. ### Isn't it quicker to just study the answer? Quicker to finish, slower to learn. In the high-school maths study, students given an answer-giving assistant did better in practice and about 17% worse on the exam once it was taken away. Students with a hint-only tutor did far better in practice and held their exam score. ### What if I am completely stuck? Use the report: it tells you which checks failed and which cells they looked at. Open the linked guide, close it, and rebuild that schedule from scratch. Being stuck and working your way out is uncomfortable, and the physics study suggests it will feel like you are learning less. It is usually the opposite.