Large Language Models

Today's Stories

1. 1

   Bank of America falls 1.3% as AI agents loom over deposits

   Bank of America shares fell about 1.3% to $53.72 on October 1 after Reuters Breakingviews raised the possibility that AI agents could make customers quicker to move cash into higher-yielding accounts. If customers move cash faster, that pressure lands on a valuable banking advantage: customers who leave their money parked, which supports funding economics.

   This is a scenario, not evidence of an AI-driven deposit exodus, so the investor test is deposit pricing, checking-account retention and net interest income.

2. 2

   Apple's RLTL;DR turns failed attempts into a 14–31% Pass@1

   Apple researchers' RLTL;DR method has a policy write a TL;DR insight after each failed attempt, then backpropagates so the model internalizes a direct task-to-insight mapping. On tool-calling and coding datasets filtered to Pass@128 = 0, it achieved a Pass@1 of 14–31% with insights in context, and 12–13% without. Standard GRPO training stayed flat at a Pass@1 of 0% to 1% on these datasets, so letting the model generate and internalize its own feedback appears to overcome a learning barrier where no teacher models or example solutions exist.

   The paper attributes the gain to task-to-insight internalization, and a reduced SFTL;DR variant training on only 4k (task, insight) tuples recovers almost the full performance, suggesting a compacted training paradigm worth watching.

3. 3

   Wichita Eagle: AI gives bad travel advice

   The Wichita Eagle reported that AI chatbots are giving travelers bad travel advice, including bogus bar recommendations and visa failures. Travelers relying on AI for trip planning may end up with incorrect or unusable advice, potentially affecting their plans and documents.

   It is unclear which AI systems or destinations were involved, so travelers should verify AI-generated travel advice with official sources. Watch for further reporting from the Wichita Eagle.

4. 4

   Ben Affleck says he used AI "for lots of things" in Netflix's Animals

   Ben Affleck said at a Bloomberg Screentime Q&A that he "did use" AI "for lots of things" in Animals, which he co-wrote, directed and stars in, premiering Oct. 9 in select theaters and on Netflix. Affleck said the AI use helped the film "move in some cases more quickly, really mostly in post-production," even though it is "a very human movie" without visual effects or a space setting.

   Affleck was "a little bit hesitant to point out" where AI was used, citing the need to "foster the illusion of reality," so how much of the tooling he eventually details is the test.

5. 5

   CrowdStrike taps CoreWeave for AI cyber defense push

   CrowdStrike and CoreWeave are partnering so CoreWeave supplies training and inference capacity for CrowdStrike's SafeMind and serves as a design partner for its Cyber Superintelligence Lab, executives said at theCUBE's Fully Connected event. The tie-up pairs CrowdStrike's security data with CoreWeave's compute infrastructure, a setup the companies say is needed to train defensive AI models on attack data and keep protection running while production systems stay live.

   CrowdStrike says these automated defenses must win the trust of CoreWeave and its customers before they count as everyday automation, so adoption hinges on whether that confidence is earned in practice.

6. 6

   Ecolab's Wijesinghe: top AI model cost more than humans

   Ecolab Chief AI Officer AJ Wijesinghe said putting the best model on high-volume work sent token costs soaring, sometimes making AI more expensive than humans, before optimization cut those costs by about 70% to 80%. Wijesinghe's cost point suggests the price of running top AI models on high-volume work can outweigh the labor it replaces, so the savings case for such deployments may hinge on cost discipline.

   Ecolab's optimization shows the gap is not fixed, but Wijesinghe said the value comes from tying AI to a company-wide transformation, with Ecolab expecting $325 million in annual run-rate savings by 2027.