SShortSingh.
Back to feed

Closed-book C# exam written by ChatGPT and Claude leaves author scoring 22-54%

0
·1 views

A developer who relies heavily on AI coding tools tested his own C# knowledge by taking two closed-book exams, one generated by ChatGPT and one by Claude Sonnet 5.5, with no AI, search, or autocomplete. He scored 53.5% on the ChatGPT exam and about 22-23% on the harder Claude exam, with modern C# being his weakest area on both. He attributes the results partly to never having truly learned newer language features rather than solely to forgetting them, since accepting generated code without studying it leaves little absorbed. The exams covered language knowledge only, excluding frameworks like ASP.NET or Blazor. He shared sample questions on value versus reference types, nullable reference types, and newer syntax as examples of what he is now relearning.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingHacker News ·

Eunice Is Almost Here: A Preview of William Gibson's Agency

A short Hacker News post points readers to an article titled 'Eunice is almost here,' which discusses William Gibson's novel Agency. The linked piece appears to be a commentary or preview tied to the book's upcoming release. The submission received three points and no comments at the time it was captured. The post offers little detail beyond its title and link, so the article's full arguments are not summarized here.

0
ProgrammingDEV Community ·

MatchGrind Launches Football Data API With Scores, Standings and Injury Tracking

MatchGrind offers a football data API that provides live scores, fixtures, standings, club information and match analytics across 60 competitions, including the Premier League, LaLiga, MLS, Brasileirão and Tanzania's NBC Premier League. The service derives injury and suspension information from match events rather than manual entry, and card totals per player show how many bookings remain before a ban. Access is available through the Global-Football-API listing on the RapidAPI marketplace, with a free BASIC plan that requires no payment card. Users obtain an API key after subscribing and can retrieve the list of available competitions with a first call, with no additional setup needed. The overview is based on public information from the project's website, and pricing and terms should be verified there before adoption.

0
ProgrammingDEV Community ·

SlothBudget Pro: Open-Source Offline Agent Automates Receipt Tracking and Spending Classification

A developer has released SlothBudget Pro, an autonomous financial agent designed to process receipts and transaction logs without manual data entry. Users drop raw text files into a local folder or upload them through a cloud dashboard, and the agent extracts transaction details, classifies spending, and archives the records. Text extraction runs on Google's open-weight Gemma 3 1B model via Ollama, while TabPFN-3.5 handles categorization and anomaly detection. The project's developer says running the models locally keeps user data private, avoids operating costs, and removes the need for an internet connection. The interactive dashboard is hosted on Render, and Sentry is used to monitor agent latency and errors.

0
ProgrammingDEV Community ·

Developer Moves Repository Tests from H2 to Real PostgreSQL Using Flyway

A developer building a library app discovered that their repository tests passed only against H2, an in-memory database that does not match the PostgreSQL engine used in production. The Phase 3 work moved persistence to PostgreSQL 17, with Flyway managing the schema through a baseline migration that mirrors the JPA entities and uses named constraints. Hibernate was set to validate rather than generate the schema, and database settings were moved to environment variables with setup documented in the README. Only the book copy repository test class was switched to the real database, while 45 other tests stayed on H2, and PostgreSQL runs natively since Docker was unavailable.

Closed-book C# exam written by ChatGPT and Claude leaves author scoring 22-54% · ShortSingh