July 1, 2026
AI News
The AI news worth knowing this week.
Anthropic releases Claude Sonnet 5, beating its bigger sibling
Anthropic released Claude Sonnet 5, a new model that beats its predecessor Sonnet 4.6 across benchmarks and even edges past the larger, pricier Opus 4.8 on a key knowledge-work test — delivering near-flagship quality in a lower-priced tier.
Read the original →Gemini's personalized AI image generation is now free for US users
Google made Gemini's personalized image generation free for all US users. Powered by its Nano Banana model, it can draw on your own Google data — Photos, Gmail and more — so a request like "design my dream house" reflects your real tastes. The feature is opt-in.
Read the original →DeepSeek open-sources DSpark, speeding up AI by up to 85%
Chinese AI lab DeepSeek, with Peking University, open-sourced DSpark — a framework that makes large language models answer 60–85% faster without lowering output quality. Released under a permissive license, it can sharply cut the cost and lag of running AI.
Read the original →Return of the ‘greybeards’: AI backfired – so Ford had to rehire humans
Ford replaced its AI-powered camera system with human inspectors after discovering the technology was unreliable and prone to errors.
Read the original →OpenClaw is finally available on Android and iOS
OpenClaw, a free open-source AI agent, can now be downloaded on Android and iOS devices.
Read the original →Dubbing & Open source
Open-source voice and dubbing tools worth a look.
jamiepine/voicebox
Voicebox is an open-source, self-hostable voice studio for cloning voices, dictation, and speech creation — a free, local alternative to paid voice tools. It has quickly become one of the most-starred new AI projects on GitHub.
Read the original →MemPalace/mempalace
MemPalace is a free, open-source memory system that lets AI assistants remember information across conversations. Billed as a top-benchmarked open-source memory layer, it is one of the fastest-growing AI projects right now.
Read the original →OpenMOSS/MOSS-TTS-Nano
MOSS-TTS-Nano is a tiny (0.1B-parameter) open-source multilingual text-to-speech model that runs in real time on an ordinary CPU — no GPU required — making local voice generation simple to deploy.
Read the original →Get the next issue by email
In any of 8 languages. Unsubscribe anytime.