September 3, 2026
Today's top tech stories, deduped across the newsletters I read and briefly summarized. Click a source to open the original article.
AI
Google launches Gemini 3.8 Flash and Flash Cyber
Google released new Gemini 3.8 Flash models, including a Cyber variant. Simon Willison quickly added support for the new model in his llm-gemini plugin.
Anthropic launches Claude Fable and Mythos 5.1
Anthropic released Fable (and Mythos) 5.1, which the company says sets a new standard for coding, knowledge work, and long-running problem-solving tasks. Simon Willison tested it by having it draw an animated pelican.
Claude's new system prompt is strict about song lyrics
Anthropic publishes the system prompts for Claude.ai and its mobile apps, and the latest version is explicitly instructed to avoid reproducing song lyrics. Simon Willison breaks down the changes in detail.
Mistral AI explains how to opt out of training data use
Mistral clarifies how users can opt out of having their input and output data used for model training. The topic sparked heavy discussion on Hacker News this week.
Meta releases the Muse Spark 1.3 AI model
Meta shipped a new version of its Muse Spark model. The release drew significant attention on Hacker News.
215,000 fabricated 'best software' pages get cited by Perplexity
A report reveals that three websites generated 215,128 'best software' pages, and that Perplexity cites them as sources in its AI recommendations. The story raises questions about the quality of sources behind AI search.
NVIDIA uses speculative decoding for faster LLM inference
In the third post of a series on AI model co-design, NVIDIA shows how speculative decoding can speed up LLM inference while maintaining accuracy.
Security
Hackers exploit Sangoma Switchvox flaw to plant reverse shells
Attackers are actively exploiting CVE-2026-9586, an unauthenticated SQL injection vulnerability in the Sangoma Switchvox VoIP platform that can lead to remote code execution.
Critical JFrog Artifactory flaw exploited to forge admin tokens
A critical authentication bypass vulnerability (CVE-2026-82329) in JFrog Artifactory is being exploited in attacks to create tokens with administrative access.
WordPress backup plugin flaw exposes millions of sites to takeover
An SQL injection vulnerability in the All-in-One WP Migration and Backup plugin for WordPress could let unauthenticated attackers execute remote code and take control of affected sites.
Dropbox accounts breached through Lenovo email verification flaw
Dropbox is warning some users that an unauthorized party accessed their accounts by exploiting a flaw in Lenovo's email verification process to register fraudulent Lenovo IDs.
Dev
NVIDIA walks through modern CUDA optimization step by step
The NVIDIA Developer Blog demonstrates a practical optimization workflow using the modern CUDA toolbox, covering profiling and performance improvements in GPU-accelerated code.