AI-Written Playwright Tests Passed But Missed Critical Backend Logic, Engineer Finds

·1 views

A QA engineer found that AI-generated Playwright tests for a password reset flow consistently passed while failing to verify any meaningful backend behaviour. The tests only confirmed that a success message appeared on screen, never checking whether a reset email was sent, whether its link functioned, or whether the token expired after use. After rewriting the assertions manually, the engineer discovered a real bug: the reset endpoint allowed tokens to be reused indefinitely with no expiration. The engineer concluded that AI tools are effective at generating boilerplate and test scaffolding but lack the business context needed to determine which outcomes actually matter. He now uses AI for first drafts but rewrites all assertions himself to ensure tests validate what the feature is genuinely supposed to guarantee.

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)

Developer Builds Single-API Audio Translator Using Telnyx AI Inference Pipeline

A developer has created a Python Flask application that performs end-to-end audio translation using only the Telnyx AI Inference API. The app accepts an uploaded audio file and a target language, then chains Telnyx's speech-to-text, translation, and text-to-speech services in a single pipeline. Long transcripts are split at sentence boundaries to stay within model input limits, and the resulting audio chunks are merged into one MP3 file. A key design feature is per-stage error handling, which preserves completed transcription and translation work even if a later stage fails. The project is open-source and available on GitHub, with the current version storing job data in memory and the author recommending object storage and a job queue for production use.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Researcher Builds Physics-Constrained AI Model to Optimize Carbon-Negative Farm Microgrids

A researcher developed a framework called Physics-Augmented Diffusion Modeling after finding that standard reinforcement learning failed to manage energy in agricultural microgrids without violating basic physical laws. The approach embeds hard physical constraints from thermodynamics, electrical systems, and crop science directly into the diffusion model's generative process. Smart agriculture microgrids covered in the research include solar panels, battery storage, irrigation loads, and carbon-negative units such as biochar production and direct air capture. Unlike conventional data-driven methods, the framework captures the full distribution of feasible energy schedules rather than single-point estimates, better handling uncertainty in renewable generation. The goal is real-time scheduling of microgrid components to minimize costs while maintaining net carbon removal from the atmosphere.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to Build a Local Voice Assistant Using Python and OpenAI's Whisper

Developers can build a fully local voice assistant using Python and OpenAI's Whisper, a transformer-based speech-to-text model trained on 680,000 hours of multilingual audio data. Unlike cloud-based assistants such as Siri or Alexa, this approach keeps all data on the user's machine, addressing privacy and latency concerns. The setup involves Python libraries including SpeechRecognition, pyttsx3, and faster-whisper to handle audio capture, transcription, and text-to-speech output. The core pipeline follows a listen-transcribe-process-speak loop, with optional integration of large language models like OpenAI's GPT or Anthropic's Claude for generating responses. Whisper's open-source nature and high accuracy across accents and noisy environments make it a strong alternative to proprietary speech recognition tools.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

A Practical Guide to Mastering Pytest Fixtures for Cleaner Test Suites

Pytest fixtures are reusable functions that set up objects such as database connections or sample DataFrames, allowing developers to avoid duplicating setup code across multiple tests. Defined with the @pytest.fixture decorator, fixtures can be scoped at the function, class, module, package, or session level, letting teams balance test isolation against execution speed. The yield keyword enables fixtures to handle both setup and teardown logic, ensuring resources are released even when a test fails. The autouse=True parameter allows a fixture to run automatically for every test in its scope without requiring explicit requests in each test function. A conftest.py file serves as a shared repository, making fixtures available across an entire test suite without repeated imports.

0 comments Read more at DEV Community