🤖 AI HOT Daily · Jun 28, 2026
Alibaba Qianwen IME launches on macOS with 300 chars/min AI voice input, Runway API introduces ad localization Recipe, SpaceX registers SpaceXAI trademark plans to merge xAI, Apple Vision VP jumps to OpenAI, China's electronics industry profits surge 103.9% Jan-May, Raise Us launches $1B retraining program for AI-displaced workers, US companies switch 100% to DeepSeek as AI bills spiral, DeepSeek open-sources DSpark speculative decoding framework boosting speed 60-85%, dissection of a failed nation-state attack, open-weight models that matter in June 2026.
1. Alibaba Qianwen IME Launches on macOS: 300 Chars/Min
AI voice input up to 300 chars/min, auto-polishing, speech-to-text, 9 dialects supported. Clean and ad-free. iOS, Android, Windows versions coming soon.
2. Runway API Launches Ad Localization Recipe
Single API call to translate static ads and graphic assets for localization.
3. SpaceX Registers SpaceXAI Trademark, Plans xAI Merger
Elon Musk announces xAI will dissolve and merge into SpaceXAI, becoming SpaceX's AI product division rather than an independent company.
4. Apple Vision VP Jumps to OpenAI
Apple's VP of Vision products Paul Meade leaves for OpenAI's hardware division. First touch OLED MacBook with M5 Pro/Max chip coming late 2026 to early 2027.
5. China Electronics Industry Profits Surge 103.9% Jan-May
Electronics profits surge 103.9% driven by global AI demand for high-performance compute and memory chips. High-tech manufacturing profits up 44.7%.
6. Raise Us Launches $1B Initiative for AI Job Displacement
Former US Commerce Secretary Raimondo launches nonprofit to raise $1B for worker retraining in the AI economy. $500M secured. Supported by Amazon, Anthropic, Microsoft, OpenAI.
7. US Companies Switch to DeepSeek as AI Bills Spiral
US companies overspend on AI. SF-based Lindy moves 100% traffic to DeepSeek, saving millions monthly. Enterprises adopt model routing to match tasks to the right model.
8. DeepSeek Open-Sources DSpark Speculative Decoding
DSpark achieves lossless acceleration via semi-autoregressive generation. V4-Flash and V4-Pro per-user speed improve 60-85% and 57-78% over baseline respectively.
9. Dissecting a Failed Nation-State Attack
Author receives fake VC interview email. Claude scans and discovers hidden base64-obfuscated backdoor 'PinpinRAT' in a TypeScript patch targeting the author's Rust package.
10. Open-Weight Models That Matter — June 2026
OpenRouter highlights the four most notable open-weight models as of June 2026 and their best use cases.
Sources and verification
This legacy briefing did not retain its original source URLs. Verify safety, financial, policy, and breaking-news claims with authoritative primary sources before acting on them.