Developer uncovers 47GB hidden memory bug that silently killed social media automation

·3 views

A software developer discovered that their Mac-based social media automation system stopped posting entirely for 24 hours due to a severe memory issue on August 9, 2026. The macOS background scheduler process 'dasd' appeared normal at 264MB in Activity Monitor, but a deeper diagnostic command revealed it was holding 46GB of compressed memory, with swap ballooned to 37GB. Because macOS compresses rarely used memory pages without reflecting them in standard RSS readings, both Activity Monitor and common monitoring tools failed to flag the problem. The hidden memory pressure prevented Playwright from launching Chrome, causing it to time out after 180 seconds with no memory-related error messages logged. The developer also identified a second leaking process, 'iii' (agentmemory), growing at roughly 4GB per hour, and built an automated launchd job to detect and resolve both issues overnight.

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)

Only 6% of Code Carried Over When Team Built a Second Similar Product

A software team that built products for a rental company and a property developer found only 7 of 113 business concepts — roughly 6% — were shared between the two systems, despite the businesses appearing similar on the surface. Key differences such as construction-stage tracking, multi-month property transactions, and investor-only user roles meant most domain logic had to be rebuilt from scratch. However, the second product shipped significantly faster because foundational infrastructure — including logins, permissions, billing, messaging, and encryption — was reused wholesale from the first build. This shared layer, invisible to end users, accounted for roughly a third of the second product and required no rework. The team's core takeaway is that reusable infrastructure, not business logic, is where multi-product development pays off.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Tutorial: Predicting Overtraining Risk Using HRV Data, Random Forest, and LSTM

A technical tutorial published on DEV Community outlines how to build a stress and recovery prediction system using wearable device data. The guide uses the Terra API to standardize health metrics from devices like Garmin watches and Oura rings into a unified data format. A hybrid machine learning approach combines Random Forest to identify which lifestyle factors most influence recovery and an LSTM neural network to forecast future HRV trends from historical sequences. The full pipeline is designed to expose predictions via a FastAPI endpoint, feeding an end-user health dashboard. The project requires Python 3.9+, Scikit-learn, Keras/TensorFlow, and Terra API credentials.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

CleanGeek: Free Windows Cleaner That Skips Registry Tools and Upsells

A developer frustrated with bloated, upsell-driven PC cleaners has released CleanGeek, a free, open-source Windows cleaning tool. The app targets junk files such as temp folders, browser caches, Windows Update leftovers, crash dumps, and thumbnail caches, showing users exactly how much space each item occupies before anything is deleted. Unlike most free cleaners, CleanGeek has no registry cleaner, no paid tier, and no telemetry, with the developer arguing registry cleaning has not meaningfully improved Windows performance in roughly 15 years. Built on .NET 8 and the Avalonia UI framework, the tool performs all scanning internally without relying on third-party cleanup engines. The app is currently unsigned, meaning Windows SmartScreen may flag it on first run, though the developer has noted this is being addressed.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Shadow-Comparing Agent Patches Can Catch Hidden Behavior Changes Before Merge

A software testing workflow published on DEV Community proposes that passing test suites alone are insufficient to validate agent-generated code patches, since such patches can silently alter encodings, error types, or edge-case handling without breaking existing assertions. The approach involves running both a frozen baseline and the candidate patch against an identical deterministic corpus, then comparing their outputs byte-for-byte after canonicalization. Any divergence found must be logged in a structured 'accepted-delta ledger' and explicitly classified before the patch is allowed to merge. The harness relies on three components — a pinned baseline artifact, the candidate patch, and the delta ledger — to ensure unspecified behavior is treated as implicitly contractual. The workflow is presented as a complement to, not a replacement for, manual code review, acting as an automated filter for behavioral drift that reviewers would otherwise miss.

0 comments Read more at DEV Community