Student manually reviews 102 F-Droid apps for AI-generated code using repo cues

·30 views

A student analyzed all 102 apps in F-Droid's September 12, 2026 update batch, manually classifying each by likelihood of being LLM-written using visual and contextual repo signals such as commit tone, README emoji density, and presence of AI agent infrastructure. No automated detection tools were used, and the author openly acknowledged the method's limitations, calling the tiers loose and the ratings superficial. The exercise highlighted that reliably detecting AI-authored code is not feasible even with full repository access and history. A key takeaway is that basing code review on AI-origin detection is a flawed approach, since provenance cannot be confirmed. Reviewers are instead urged to focus on the code itself — evaluating correctness, security, and logic regardless of whether a human or an AI wrote it.

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)

AWS Releases Official Emulator for Its Own Cloud Services on GitHub

Amazon Web Services has published an emulator project on GitHub, as reported on Hacker News. The post links to the repository and a discussion thread where early readers have gathered. The thread has gained only a few points and one comment so far, indicating limited initial attention. The source material provides no details on the emulator's features, supported services, or intended use cases. Further information would need to come from the repository documentation or an official AWS announcement.

0 comments Read more at Hacker News

ProgrammingHacker News ·

Pomerado launches tool that turns any website into a self-healing API for AI agents

Pomerado, a Y Combinator F26 startup founded by Alan and Akshay, has launched a self-serve platform that converts websites into deterministic, self-healing APIs for AI agents. The founders say many airlines, banks, insurers, and government portals lack APIs, forcing agents to rely on slow, costly, and unreliable browser automation that companies often block for security reasons. Their platform has an agent explore a site once to record the task, then generates a repeatable tool exposed via MCP, with a separate guardian model checking guardrails and synthetic tests validating generalization. The system also handles authentication and 2FA through a deterministic sign-in process so credentials are never read by the model, and a maintenance job repairs the tool when sites change. The founders said the idea grew out of their experiences building personal assistants and deploying agents in healthcare, professional services, and telecom.

0 comments Read more at Hacker News

ProgrammingHacker News ·

Apple Research Explores Faster Retrieval to Make RAG More Efficient

Apple's machine learning research team has published work on speeding up retrieval for retrieval-augmented generation (RAG) systems. The focus is on making the retrieval step faster and more efficient, which is a key bottleneck in RAG pipelines. The post appeared on Apple's machine learning research site and was shared on Hacker News, where it gained a few points and no comments at the time of capture. The available source material offers only the headline, link, and engagement figures, so the specific techniques and results are not described here. Readers interested in the methods can refer to the original Apple research page.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

Fixing a Slow Database: Work Up the Scaling Ladder Cheapest-First

The article argues that a slow database should be fixed in a deliberate order rather than jumping straight to sharding, which is costly. It recommends first confirming the real cause and adding indexes, then adding a cache such as Memorystore for repeated reads, then using read replicas for read-only traffic. Scaling the instance to a larger machine type comes next, with attention to brief restarts, replica sizing and retry logic. Sharding is reserved for cases where the single primary remains the write bottleneck, and measurement is advised at every step.

0 comments Read more at DEV Community