SUMM

Elon Musk questioned Apple's apparent favoring of ChatGPT in Apple's App Store over his own apps, X and Grok, which he claims lack similar editorial support despite ranking high in usage.
Musk accused Apple of creating an antitrust situation by making it nearly impossible for any AI company other than OpenAI to top the App Store.
Community fact-checkers responded by citing Deepseek and Perplexity, which reached number one in the App Store after Apple's OpenAI partnership, undercutting Musk's claims.
Sam Altman countered Musk, alleging Musk manipulates X to benefit his own interests and suggesting legal scrutiny would reveal more.
The public exchange devolved into both parties urging each other to provide legal affidavits to support their claims about algorithmic manipulation.
Both used their own and rival AI systems (ChatGPT, Grok) to “decide” who is more trustworthy, with inconsistent answers, highlighting the futility and subjectivity of such arguments.
The video presenter describes the feud as juvenile and a waste of time, seeing Musk as possibly trying to sway public opinion rather than prove a legal point.

Lindy released a major update with features like “autopilot,” allowing its agents to perform human-like actions on user devices and platforms.
Example use cases include monitoring accounts for spam and automatically blocking spam accounts.
Lindy agents can now interact with multiple platforms to execute complex tasks, moving closer to the idea of a fully functional AI employee.
The video offers a promotion for Lindy credits for new users.

Anthropic’s Claude Sonnet 4 now supports a 1 million token context window, making it one of the first models after Gemini 2.5 Pro to reach this milestone.
The extended context enables handling about 75,000 lines of code or hundreds of documents in a single prompt.
This enhancement benefits users needing to load large codebases into the model for analysis or generation.
The feature is in public beta for API users with high-tier or custom rate limits, and is available on Amazon Bedrock and soon Google Cloud.
Pricing structures change based on prompt size, with increased costs for prompts over 200,000 tokens.

Perplexity now offers video generation features across devices, though generation quotas are relatively low (5–15 videos/month depending on plan) due to high computational costs.
Genie 3 inspired a new open-source real-time interactive world model called Skywork Matrix Game 2.0.
Matrix Game 2.0 can generate real-time, long-form, interactive videos at 25 frames per second, with scenarios like city, wild, temple run, and GTA-like environments.
It was trained on 1,350 hours of video from Unreal Engine and GTA 5, and is fully open source, available for use on a single GPU.

Ex-OpenAI researcher Leopold Ashenbrener, known for the situational awareness paper, raised $1.5 billion for his AI-focused hedge fund.
The fund’s strategy targets global stocks benefiting from AI, e.g., semiconductors and infrastructure, and also invests selectively in AI startups like Anthropic.
The fund leverages short bets against industries that may be negatively impacted by AI.
The fund reportedly achieved a 47% gain after fees in the first half of the year.

OpenAI’s unnamed reasoning model scored high enough to win a gold medal at the International Olympiad in Informatics, outperforming all but five human competitors and placing first among AI participants.
The model performed without external internet or retrieval augmentation, contrasting to its previous year’s performance where it fell just short of a medal.
Mistral AI released Mistral Medium 3.1, offering performance boosts, improved tone, and smarter web searches, with benchmark improvements in several areas.

AI News: Sam vs Elon, Claude 1m Context, Situational Awareness $1.5B