SShortSingh.
Back to feed

Caltech professor stops grading homework, citing widespread AI use by students

0
·8 views

A Caltech professor has decided to no longer grade homework assignments because students are widely using AI tools to complete them. The change reflects growing concern among university instructors that homework scores no longer reliably measure individual understanding. The details come from a Business Insider article linked from a Hacker News post, which had a small amount of early engagement at the time it was shared. The article is part of wider coverage of how AI is affecting math classes and how students may be preparing for careers. Further specifics about the course and the professor's alternative assessment methods were not available in the provided material.

Read the full story at Hacker News

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
ProgrammingDEV Community ·

Study finds OTP delivery to shared virtual numbers varies widely by app and country

A receive-SMS service analyzed about 900 tries on shared free numbers between 27 October and 16 November 2026, counting whether verification codes arrived. Success rates differed sharply by app and country, with Marktplaats in France delivering codes 98% of the time while Telegram rarely delivered any. The company says some large apps refuse shared numbers entirely, and waiting longer does not change the outcome. It also notes that many users cancel too early, although international SMS can take several minutes. The authors recommend private numbers for reliable automated testing, and the full dataset is published under a CC BY 4.0 license.

0
ProgrammingDEV Community ·

Can You Compete in Tech Without Strong English? A Two-Part Answer

Many developers worry that weak English will hold them back, but the author argues the answer depends on what is meant by competing. Most English speakers worldwide learned it as a second language, and English's share of websites has been shrinking, while programming keywords are simple to memorize. AI assistants now make it far easier to read technical material in one's own language, and history shows great software has emerged outside English. However, the surrounding tech conversation, including issues, reviews, talks, and interviews, still largely happens in English, which can slow learning, reduce visibility, and limit pay. The author recommends training reading, writing, listening, and speaking as separate skills through short daily practice.

0
ProgrammingDEV Community ·

Audit of 37 B2B Directories Finds Zero Usable Backlinks for a Static Site

A developer running a 897-page, ten-language static site on Cloudflare Workers audited 37 B2B directories to see which ones gave company pages a clickable link back to their site. None of the directories provided a usable link. The author describes six methodological errors made along the way, such as mistaking search results for backlinks and relying on static HTML fetches that miss client-rendered content. A key lesson was that plain-text domain mentions can look like links in search results but are not anchors in the page's DOM. The post also explains that checking rel attributes requires distinguishing nofollow from security directives like noopener and noreferrer.

0
ProgrammingDEV Community ·

Verified Backups for Object Storage: Manifests, Multipart Uploads and Restore Checks

The article argues that a backup should be treated as a verifiable object paired with a restore contract, not merely a file that a scheduler reports as successful. It recommends streaming database dumps to private object storage alongside an immutable manifest recording the key, byte length, checksum, timestamp, schema version and included partitions. Backups should live in a separate namespace from private uploads, and presigned download links should be issued only after authorization because they act as bearer credentials. A backup counts as complete only after read-after-write verification confirms the length and checksum, and the author suggests bounded retries, concurrency limits and idempotent commits for multipart transfers. The piece also illustrates the risk with a quiet failure, in which a full ephemeral disk masked a truncated archive until a restore check caught the mismatch.