Search
10 stories found
OccupySaaS describes itself as a group of laid-off engineers who use AI agents to clone paid SaaS from public docs. Its Auth0-compatible identity server reached release candidate 1 on November 4. It's Apache-2.0 and written in Go, and it already works with Auth0's own SDKs. The project says it isn't production-ready, and it isn't yet.
MIT CSAIL researchers say OpenAI's GPT-6 Astra can identify what small networks do from their raw weights and edit a Qwen model to fix its math. The headline result, a jump from 7.9% to 48.1% on MATH-500, came from Python tooling and 50 forward passes, not from reading raw weights. Here's what each setup showed.
Microsoft Threat Intelligence has graded eight North Korean groups on how deeply they use AI. The findings — bulk-generated lures, AI-coded malware, fake recruiters and an open-source agentic attack framework — show where defenders should tighten up now.
A polished, ad-free, offline-capable vocabulary app — and the more interesting part is that its words, illustrations and audio are released under CC BY 4.0 for anyone to build on.
Apple says developer testing in China starts November 4 and the public gets Apple Intelligence in December with iOS 27.3, with Alibaba's Qwen built into Siri, Writing Tools and Visual Intelligence. The AI Siri shown at WWDC 2026 isn't part of this phase, and neither are Macs or iPads.
Google's Suncatcher prototype satellite returned its first AI response from a data-center-class TPU in orbit. It's a real hardware milestone, but it says very little yet about whether space data centers can compete on cost.
Google's November 2026 Android Security Bulletin patches more than 200 vulnerabilities, including a critical remote code execution flaw that needs no user interaction. Nothing is flagged as exploited yet, but the bulletin also lists the "SMS of Death" phone-crash bug that already has a public proof of concept.
A new arXiv preprint claims a theorem showing that JEPA models trained on different modalities converge to the same representation. It reports that a 109M-parameter text model out-aligns 72B LLMs. The results are striking, but they are benchmark-specific and not yet peer-reviewed.
NOIR, a new MIT-licensed desktop tool, scans an Android phone over USB debugging and links each app's sensitive permissions to the tracking and ad domains hard-coded in its code. It went from zero to about 1,800 GitHub stars in four days. It's a useful map of risk, but its findings show what apps can do, not proof of what they actually sent.
Sekoia says APT29 used compromised sites, lookalike domains and a hijacked Plerdy analytics script to push zero-click exploit chains at visitors on outdated iOS and Android. Every bug was already patched, so up-to-date phones were safe. CISA has now added one of the iOS flaws to its exploited list.
