Ground Truth.
AI, checked against the source.
← 2026-06-282026-06-29later →

DeepSeek's new open models give everyone a million-word memory by default

2026-06-29

DeepSeek previewed two free-to-download V4 models that can read a million tokens at once, no longer as a premium add-on but as the standard setting.

open-weights · long-context · deepseek · attention · models

Tidal will stop paying royalties on fully AI-made songs

2026-06-29

Starting July 15, the streaming service won't monetize tracks where every part was made by generative AI - the first major platform to demonetize rather than just label them.

ai-music · policy · streaming · creators · generative-ai

Microsoft's new memory system lets AI agents remember more by storing less

2026-06-29

Memora keeps the rich detail of a conversation but searches it using tiny six-word labels, cutting the cost of remembering by up to 98 percent. The code is public.

agent-memory · microsoft · agents · retrieval · open-source

South Korea bets over a trillion dollars on chips, data centers, and robots

2026-06-29

The government and its biggest companies committed more than $1 trillion to memory fabs, AI data centers, and a goal of building tens of thousands of humanoid robots a year by 2028.

industry · semiconductors · robotics · policy · infrastructure

Amazon and Anthropic's partnership is cracking over the price of Claude

2026-06-29

A renegotiated contract is expected to sharply raise Amazon's bill for Anthropic's AI, pushing Amazon toward OpenAI and its own models even though it's an Anthropic investor.

industry · anthropic · amazon · pricing · business

Qwen used human-feedback training to make its image AI follow directions better

2026-06-29

A new recipe applies the same reinforcement-learning approach that polished chatbots to an image generator, then merges separate skill models into one - improving how faithfully it follows prompts and edits.

image-generation · diffusion · reinforcement-learning · qwen · post-training

NVIDIA's new method stops AI dream-worlds from breaking the laws of physics

2026-06-29

PhysisForcing trains video-generating world models to keep objects solid and interactions believable, raising how often a robot's imagined plan actually works.

world-models · robotics · video-generation · nvidia · physics

A robot AI that adapts to a moved camera by wiggling, not retraining

2026-06-29

A new method lets robot policies figure out a changed setup from a few seconds of self-directed fiddling, so they keep working when the camera or robot body changes - with no retraining.

robotics · world-models · in-context-learning · adaptation · vla

← 2026-06-282026-06-29later →