Developer finds 96x cost gap in web scrapers traced to a single misconfigured memory field

·1 views

A developer benchmarked all 11 of their published web-scraping Actors across 22 runs, calculating cost per 1,000 rows using platform billing data. The results showed a 96-fold spread in costs, ranging from $0.008 to $0.730 per thousand rows. Initially, the author attributed the gap to request efficiency — cheaper scrapers extracted multiple rows per request, while expensive ones made one request per row. However, a reader's question prompted a deeper check, revealing that the four most expensive Actors were all configured with 4,096 MB of memory, compared to 512 MB for the rest. Since the platform bills by memory multiplied by runtime, the eightfold RAM difference — apparently inherited from a template and never changed — was the primary driver of the cost disparity.

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)

Self-auditing AI QA agent logs 79 findings against its own code in nightly reviews

A developer built an AI-powered QA agent called Verdict that audits its own releases every night rather than reviewing external codebases from scratch each time. Unlike typical AI QA tools, Verdict maintains a persistent state file, a failure taxonomy, and a signed run history, and is strictly prevented from editing the code it evaluates. The system classifies every failing test into categories such as real defect, brittle, or flaky, and assigns each finding a stable ID and verdict from a fixed set of outcomes. Over its operational period, Verdict has filed 79 findings against its own harness, including repeated defects in fixes for earlier issues, prompting a process change requiring fix verification along untested axes. When run against the open-source library itsdangerous on a fresh clone, it completed an audit in 18 minutes, flagging nine findings including a major gap in timing-attack defences and an unpinned boundary condition in token age validation.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

RAG, MCP, and Fine-Tuning Are Not Rivals — Each Serves a Distinct Layer

A technical discussion from DEV Community clarifies that RAG, MCP, and fine-tuning are commonly misunderstood as competing approaches, when in fact each operates at a different layer of an AI system. RAG functions as the knowledge base layer, retrieving relevant document excerpts from tools like Notion or Confluence at query time without retraining the model. MCP acts as the connector that allows an IDE or agent harness to request those retrieved excerpts from the knowledge base. Fine-tuning, by contrast, shapes the style and format of model responses and is not a substitute for maintaining an accurate, up-to-date knowledge base. The recommended order of implementation is: build the knowledge base first, add retrieval via RAG, connect it through MCP, and only consider fine-tuning if output formatting remains problematic after prompt and schema adjustments.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Flutter Web Can Now Save Files In Place Using File System Access API

Flutter Web developers can use the File System Access API to write edits directly back to a user's original file, eliminating the repeated duplicate downloads that plague browser-based editors. The API provides a real file handle chosen by the user through the browser's native picker, allowing in-place saves without generating copies like budget(1).csv or budget(2).csv. Handles can be stored in IndexedDB and persisted across sessions, enabling apps to silently reopen a document on reload or prompt a single 'Reopen' button if permission needs reconfirmation. The API is currently supported in Chrome, Edge, and other Chromium-based browsers, so developers are advised to implement an honest fallback — labelled 'Download a copy' rather than 'Save' — for unsupported environments. Security is enforced by design: handles can only be created through user-initiated picker gestures, access is scoped strictly to the chosen file or directory, and permission is revocable at any time.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

TaxUI: New CSS Framework Uses Declarative HTML Attributes, Requires No JavaScript

A developer has released TaxUI, an open-source front-end framework designed to simplify UI development without relying on JavaScript or heavy build tooling. Instead of utility class chains, TaxUI lets developers build interfaces using declarative HTML attributes powered entirely by GPU-accelerated CSS. The framework supports common UI needs such as responsive grids, modals, and dark mode out of the box. TaxUI is available via npm and its source code is publicly hosted on GitHub. The creator is actively seeking community feedback and contributions to guide future development.

0 comments Read more at DEV Community