Ground Truth.
AI, checked against the source.
← 2026-07-052026-07-06later →

Anthropic found a 'global workspace' inside its models - and a tool to read it

2026-07-06

Anthropic showed that a small set of internal patterns in its models acts like a silent working memory the model can report on, steer, and reason through - and released a tool that reads it to catch the model lying.

interpretability · anthropic · ai-safety · consciousness · alignment

A $4-per-million open model is coming for the frontier's 90% margin

2026-07-06

GLM-5.2, an open-weights model priced at under a fifth of Opus, scores as the top open model and 4th overall - and a widely-shared essay argues it is the first real threat to frontier labs' ~90% inference margins.

open-weights · glm · economics · inference · china

Nvidia is now backstopping the sales of its own chips

2026-07-06

Nvidia formalized a financing program that guarantees a revenue floor for cloud firms buying its GPUs in exchange for a cut of their sales - a move analysts say could help drive AI debt past $7 trillion by 2029.

nvidia · gpu · ai-infrastructure · finance · economics

The RL 'mirage': the policy you optimize isn't the one you ship

2026-07-06

A top-ranked paper shows that in modern RL training, improving the model you optimize does not guarantee improving the model you actually deploy - because the two run on different engines that disagree on probabilities.

reinforcement-learning · rl-post-training · training · reasoning · llm

A 32B model reaches frontier level by learning what to remember

2026-07-06

New research reframes agent memory as a trainable skill rather than a growing transcript - and shows that optimizing memory alone lets a 32B open model rival Claude Opus, and a 4B model leap from 4% to 78% on a benchmark.

ai-agents · agent-memory · distillation · long-horizon · llm

A hobbyist turned an e-ink tablet into Tom Riddle's diary

2026-07-06

An open-source Rust app turns the reMarkable Paper Pro into a magical diary: write a question by hand, wait a beat, and a vision AI writes back in animated handwriting as your own ink fades away.

e-ink · remarkable · vision-llm · open-source · rust

Four rival AI labs propose a shared severity scale for jailbreaks

2026-07-06

Anthropic, Amazon, Microsoft, and Google jointly proposed a five-level scale for rating how dangerous an AI jailbreak really is - aiming to standardize a chaotic field where every 'jailbreak' currently sounds equally alarming.

ai-safety · jailbreak · anthropic · policy · cybersecurity

A 4B model on your device nearly matched a 72B one - by copying its memories

2026-07-06

Researchers distilled a large AI agent's memory skills onto a compact 4-billion-parameter model, raising its success rate on a household-task benchmark from 4% to 78% while running three times faster than its 72B teacher.

distillation · agent-memory · on-device · efficiency · ai-agents

← 2026-07-052026-07-06later →