




The US government has lifted its block on Anthropic's Claude Mythos 5 AI model, allowing the company to grant access to over 100 US organizations including large corporations and government agencies, according to multiple reports. The move marks a significant de-escalation in tensions between the Trump administration and the AI lab, which had escalated two weeks ago when export controls were imposed. Commerce Secretary Howard Lutnick informed Anthropic in a letter that the decision was made because "appropriate safeguards are in place," as Wired reported. Axios, which also saw the letter, noted that Lutnick cited "significant progress" from the company's engagement with the government. "appropriate safeguards are in place." The restrictions had been imposed two weeks ago, effectively shutting down Mythos 5 and its cousin Fable 5 after warnings from Amazon and other companies, according to Semafor and Hacker News. The abrupt removal created uncertainty in the industry and among allies whose cyber capabilities were impacted, Axios reported. "significant progress" However, the public-facing version, Fable 5, remains blocked with no clear timeline for rollout, according to The Verge. Talks about Fable 5 are ongoing, Hacker News reported, citing sources. The partial resolution follows weeks of tense negotiations between the White House and Anthropic, leaving the full future of the model still uncertain.

This article discusses a technique called Matrix Orthogonalization to improve associative recall (AR) in recurrent neural networks (RNNs), aiming to close the gap with transformers on this capability. The work is motivated by domains like long-horizon reinforcement learning where
Companies that laid off workers expecting AI to replace human expertise are reversing course and rehiring employees. Ford is a notable example, reportedly re-employing hundreds of experienced engineers to handle quality issues that automated systems couldn't resolve. The article









Shared by @flarup ↗


Shared by @TheZachMueller ↗
Shared by @THR ↗
Shared by @baseten ↗