“FlashREINFORCE: Critic-Free Single-Rollout Asynchronous RL for Agentic LMs” This paper argues that scalable agent RL doesn’t need multiple rollouts of the same prompt, but can learn from each trajectory as soon as it finishes. They combine reward centering across independent prompts, sequence-level trust regions for stale trajectories, and sample-mean optimization so long failures don’t dominate training, creating stable critic-free learning from just one rollout per prompt. This moves from group-based RL that spends compute repeatedly sampling the same prompts to asynchronous RL where every rollout covers a new prompt, immediately becomes training data, and lets agents learn from more diverse experiences with less rollout compute.

.@flashnet’s cross-chain swaps depend on winning the wallet quote race Fynd returns the best price in 82% of those races (with ~58ms solves) Flashnet stays reliable while other aggregator APIs hit a rate-limit here's @foflexity on how Flashnet uses Fynd

FLASHBACK: In 1987, President Trump ran this message in The New York Times, The Washington Post, and The Boston Globe. The Democrats and useful idiots like Rick Wilson want you to think this is proof he supports Russia. The fact is, it proves he has been America First his entire life. Here is the full article: DONALD JOHN TRUMP To The American People: For decades, Japan and other nations have been taking advantage of the United States. The saga continues unabated as we defend the Persian Gulf, an area of only marginal significance to the United States for its oil supplies, but one upon which Japan and others are almost totally dependent. Why are these nations not paying the United States for the human lives and billions of dollars we are losing to protect their interests? Saudi Arabia, a country whose very existence is in the hands of the United States, last week refused to allow us to use their mine sweepers (which are, sadly, far more advanced than ours) to police the Gulf. The world is laughing at America’s politicians as we protect ships we don’t own, carrying oil we don’t need, destined for allies who won’t help. Over the years, the Japanese, unimpeded by the huge costs of defending themselves (as long as the United States will do it for free), have built a strong and vibrant economy with unprecedented surpluses. They have brilliantly managed to maintain a weak yen against a strong dollar. This, coupled with our monumental spending for their, and others, defense, has moved Japan to the forefront of world economies. Now that the tides are turning and the yen is becoming strong against the dollar, the Japanese are openly complaining and, in typical fashion, our politicians are reacting to these unjustified complaints. It’s time for us to end our vast deficits by making Japan, and others who can afford it, pay. Our world protection is worth hundreds of billions of dollars to these countries, and their stake in their protection is far greater than ours. Make Japan, Saudi Arabia, and others pay for the protection we extend as allies. Let’s help our farmers, our sick, our homeless by taking from some of the greatest profit machines ever created — machines created and nurtured by us. “Tax” these wealthy nations, not America. End our huge deficits, reduce our taxes, and let America’s economy grow unencumbered by the cost of defending those who can easily afford to pay us for the defense of their freedom. Let’s not let our great country be laughed at anymore. Sincerely, Donald J. Trump