Effective asynchronous planning, or the ability to efficiently reason and plan over states and actions that must happen in parallel or sequentially, is essential for agents that must account for time delays, reason over diverse long-horizon tasks, and collaborate with other agents. While large language model (LLM) agents show promise in high-level task planning, current benchmarks focus primarily on short-horizon tasks and do not evaluate such asynchronous planning capabilities. We introduce Robotouille, a challenging benchmark environment designed to test LLM agents' ability to handle long-horizon asynchronous scenarios. Our synchronous and asynchronous datasets capture increasingly complex planning challenges that go beyond existing benchmarks, requiring agents to manage overlapping tasks and interruptions. Our results show that ReAct (gpt4-o) achieves 47% on synchronous tasks but only 11% on asynchronous tasks, highlighting significant room for improvement. We further analyze failure modes, demonstrating the need for LLM agents to better incorporate long-horizon feedback and self-audit their reasoning during task execution. Code is available at https://github.com/portal-cornell/robotouille.

We are reaching peak fried chicken sandwich in Los Angeles. So prolific is fried chicken between slices of bread that it may one day surpass pastrami as the quintessential L.A. sandwich. But when done well -- chicken crisp, bread excellent, pickles in abundance -- there's always room for a newcomer. "It's a sandwich that rules them all," Hymanson said. There's a lot of chicken sandwiches in the neighborhood, and we thought it would be nice to be part of the club." Kismet's take is a Goliath, with wings of iceberg lettuce sticking out the sides and a knife plunged into it, announcing it as a whole-meal experience.

You just got home after a long commute from work, and you're starving. The pantry's empty -- you didn't get around to buying groceries -- and so you fire up a food delivery app and hop over to a list of favorites. You're in the mood for something different, but you're stuck with indecision; what if you order something you end up disliking? Los Angeles-based startup Halla aims to solve that problem once and for all with Halla I/O (which stands for "intelligent ordering"), a platform that uses artificial intelligence (AI) to generate Netflix-like recommendations for grocery, restaurant, and food delivery apps and websites. "We use psychographics and data to predict your preferences and cravings at any given moment," CEO and cofounder Spencer Price told VentureBeat in an interview, "and we apply predictive analytics and AI to taste and flavor attributes to gain an understanding of the food itself."

Jason Rowley is a venture capital and technology reporter for Crunchbase News. To the uninitiated, startup fundraising can be confusing. And even some of the resources designed to be approachable for the newcomer often raise more questions than they answer. So we've launched a series called "A Startup Takes Flight" to simply explain the dynamics of fundraising and deal terms. To do so, we're following two entrepreneurs who started a company and raised some money from investors.

In my last retail review, I explored how my learnings at NRF 2017 changed my view ofthe so-called omni-channel (Omni-channel may be science fiction, but a single source of truth matters). That leaves open the impact of predictive, "AI", and personalization tech on retail. CEO and Founder Alan O'Herlihy gave me a fast-paced rundown of how his company has become entrenched in five of the ten largest global retailers. How did they pull it off? Get ready for this one: instead of pursing the singularity, they focused their deep learning tech on a real world pain point: lost sales at the point of sale.