August 15, 2026
Today's top tech stories, deduped across the newsletters I read and briefly summarized. Click a source to open the original article.
AI
GLM-5.3: New frontier coding model shows emergent cyber capabilities
Z.ai released GLM-5.3, a frontier-class coding model that developers say displays emergent cybersecurity capabilities, sparking a big discussion on Hacker News.
Qwen 3.8 27B released on Hugging Face
Alibaba released Qwen 3.8 27B, a new open language model available in FP8 format on Hugging Face, topping Hacker News with nearly 1,000 points.
Why does Opus 5 feel worse to work with?
A blog post analyzing why many developers find Anthropic's Opus 5 less pleasant to work with than earlier models sparked a big discussion on Hacker News.
Google is making private AI practical with homomorphic encryption
Google describes how it uses homomorphic encryption to let AI models process encrypted data without decrypting it, a step toward more privacy-preserving AI.
How Anthropic plans to watermark Claude's AI-generated text
Anthropic is developing a technique to watermark text generated by Claude, aiming to make it easier to identify AI-generated content going forward.
Maximizing the value of your Claude Code sessions
Anthropic published a guide with concrete tips for getting more out of Claude Code, covering context management and workflow, and it drew a strong response on Hacker News.
Security
Going Dark, and the era of law enforcement hacking
A cryptography researcher's post examines how law enforcement increasingly relies on hacking techniques to bypass encryption, in what's described as a new "going dark" era.
Max severity SAP Commerce Cloud flaw now targeted in attacks
A maximum-severity remote code execution vulnerability in SAP Commerce Cloud is already being exploited in active attacks, just three days after the patch was released.
RingCentral data breach exposed info of 1.6 million accounts
The ShinyHunters extortion group stole personal data from 1.6 million RingCentral accounts after breaching the company in July, according to breach notification service Have I Been Pwned.
Dev
Don't classify, hallucinate
A technical post proposes deliberately using LLM hallucinations to generate hypothetical classifications instead of traditional tagging, an idea that gained traction on both Hacker News and Simon Willison's blog.
Web
Firefox is now the last major browser that still supports uBlock Origin
With Chrome and other Chromium-based browsers phasing out Manifest V2, Firefox is now the only major browser still supporting the full version of the uBlock Origin ad blocker.
IT
RustDesk now supports true unattended remote access on Wayland
The open-source remote access tool RustDesk has added support for full unattended access on Wayland, a long-requested feature for Linux users.