 |
|
|
|||||
For much of the AI era, intelligence has been on-demand: a user issues a prompt, and the model responds after reasoning through the request. But as AI systems grow more autonomous and expectations rise for real-time reasoning, low latency, and cost-efficiency, the definition of intelligence is shifting. Were entering a new phase where AI is expected to stay ready for the next requesteven during downtime. The key to unlocking this proactive AI future may lie in an unexpected moment: when the AI is “asleep,” a phase now called sleep-time compute. The term was coined in an April 2025 white paper by Letta, a Berkeley-born AI startup spun out of UC Berkeleys Sky Computing Lab, founded by researchers Charles Packer and Sarah Wooders. Developed in collaboration with Databricks and Anyscale cofounder Ion Stoica and others, the sleep-time compute framework aims to shift AI from reactive to proactive intelligence. Instead of waiting for prompts, AI agents use idle time to precompute answers, refine memory, and anticipate user needs. Wooders says the idea draws inspiration from neuroscience. Just as humans consolidate memories during sleep and reflect beyond immediate tasks, AI should be able to do the same. We might think about a conversation that we had with someone yesterday and make new conclusions from it, or spend time learning new things even if there is no immediate job to be done. AI Agents, on the other hand, dont spend any time thinking outside of the scope of a task, she tells Fast Company. With sleep-time, the idea is to give AI agents the same ability to think and process offline just like we do as humans. The result is an always-on AI system thats faster, more cost-efficient, and remarkably responsive. The paper reports accuracy gains of up to 18% in certain reasoning tasks, and a 2.5-times reduction in cost per query. By spreading computation across related queries and reducing redundant processing, response times and operational costs fall significantly. Why Wait When Your AI Can Think Ahead? Lettas approach uses a dual-agent model. One agent handles live interactions; the second, the sleep agent, activates during downtime to analyze past conversations, parse uploaded documents, and reorganize memory. This division allows the system to maintain context without reprocessing everything in real time. Wooders says the goal is to let agents learn offline by generating learned context, or consolidated insights from prior data. As context windows grow larger, an agent might have a ton of tokens dedicated to storing this learned contextincreasing the likelihood that any new task or question is about a topic that its already thought about, she says. For his part, Packer calls sleep-time compute a successor to test-time compute (TTC), and the next big direction for scaling AI. Rather than only adding compute during inference, systems can now scale intelligence during downtime. Sleep-time compute builds on the idea that the longer a model can reason . . . the better the final answer, says Packer. By staying active during downtime, AI agents can refine their memory, precompute likely responses, and redistribute compute resources more efficiently to improve both performance and cost. Stoica, the Anyscale cofounder and UC Berkeley professor, sees this shift as pivotal, noting that vast quantities of compute will be spent on reasoning at training time or sleep time to create shared context, unlocking greater efficiency when models are in use. Test-time compute, or inference, refers to a models ability to apply knowledge to generate outputs. Allocating more resources at this stage improves output quality but increases latency and back-end costs. Always-on tools like chatbots and coding assistants need fast, low-latency responses to serve users effectively. As these systems grow more complex, they require significantly more compute, says Anyscale cofounder Robert Nishihara, driven by sophisticated agentic systems that demand significant computational resources. Lettas research shows that sleep-time compute also boosts model power. In benchmark tasks like GSM-Symbolic and American Invitational Mathematics Examination (AIME), shifting computation to downtime reduced test-time workload by up to five times without hurting accuracy. Agents can update knowledge, refine memory, and improve performanceall without human input or added GPUs. Sleep-time Compute is Already Reshaping Billion-dollar Stacks The concept may sound theoretical, but major tech companies like OpenAI, Anthropic, Cursor, and Google are already building with sleep-time principles. During an interview panel in June, OpenAI CEO Sam Altman previewed how our future AI interactions will shift. Im excited about a future where multiple copies of AI models like o3 run constantly in the backgroundreading Slack, checking emails, acting like a team of helpful agents, said Altman. Id love to wake up to drafted email replies and a summary of unfinished tasks like heres what you didnt finish on your to-do list yesterday with suggested next steps. OpenAIs AI coding tool Codex now enables asynchronous code refactoring in cloud environments. Moreover, AI code editor Cursor recently launched background agents that operate in parallel cloud environments. Developers can deploy a fleet of agents that run test suites, refactor code, and generate new features in the background, guided by context. Anthropics Claude Code SDK offers similar functionality. Developers can deploy subprocesses that function like backstage assistants, handling testing or debugging without interrupting the main workflow. Googles “Project Naptime” and “Big Sleep,” an internal collaboration between Project Zero and DeepMind, is also exploring principles similar to sleep-time compute for code vulnerability detection. Building the Future of Ambient Intelligence Letta has embedded sleep-time compute in MemGPT 2.0, an open-source framework that equips AI agents with persistent, efficient long-term memorywhat it calls infinite context. By offloading memory tasks to sleep-time phases, the framework improves context management and reliability. Here, sleep-time compute acts like a silent housekeeper, running continuously to stay organized. Through asynchronous memory consolidation and simulated scenarios, Letta is advancing a long-standing goal in AI: agents that prepare for the future, not just react in the moment. Test-time scaling often slows down the user experience, with tasks lke Deep Research taking minutes to complete. “But with sleep-time compute, the time the agent can spend thinking is unlimited, says Wooders. Its about creating a new dimension of scaling compute, which historically has led to improvements in AIs capabilities. Letta says its framework is already making an impact, from financial chatbots summarizing earnings reports overnight to medical agents analyzing patient histories while the system is idle. Lettas model-agnostic infrastructure lets developers mix and match models within a single agent. For example, a chat agent might run on OpenAI while a sleep-time agent handles memory on Anthropic. This makes it easier to build AI that feels stateful, always-on, and proactive, says Packer. As AI evolves toward multi-agent systems, the ability to think ahead could define the next wave of tech breakthroughs. The most powerful systems wont just be those with the largest models, but those that know how to process information quietly and efficiently, even in their sleep. In the future, vast quantities of compute will be spent on reasoning at sleep time by agents to make sense of new information and context that the agents encounter, says Stoica, the UC Berkeley professor. I expect this direction to be a major driver of progress in AI. Engineering the right shared context through reasoning at training time or sleep time will allow for far more efficiency at test time.”
Category:
E-Commerce
If we havent been there ourselves, weve seen it happen: A well-respected team member, bursting with potential, is promoted into a new leadership role. There are congratulations and smiles all around, and the new chief digs in, scheduling meetings and even offering sneak peeks at their 90-day plan. But as the good vibes fade, and everyone sets their eyes on the workKPIs, deliverables, an upcoming board meetingit soon becomes apparent that somethings wrong. The new leader may find their ideas and ways of working arent quite . . . landing. Have senior leaders moved the goalposts? Are peers expressing less enthusiasm? Is the team holding back? Whatever the disconnect, frustration builds. Anticipated quick wins become slow-rolling fails. Senior leadership begins to express concern, offering feedback that only increases the pressure. The new leader feels awkward and isolated, losing sleep and second-guessing every move. Soon, their job may even be at risk. A widespread problem This kind of failure would be tragic enough if it were uncommon. But its not. Some 46% of leadership transitions underperform, according to research and advisory firm CEB Inc. (Now owned by Gartner Inc.) The stats are even worse for the C-Suite. New CEOs have a failure rate as high as 50%, according to McKinsey, and 90% of them wish they had handled their transition differently. According to McKinsey, a failed leadership transition can cost the company more than two times the executive’s annual compensation. But of course the cost of ripple effects from the failurelower productivity, higher turnover across the team, and missed opportunities can cost even more. All told, failed executive transitions are one of the most underrecognized systemic risks facing organizations today. So much of a companys momentum depends on getting transitions rightpromoting top performers, replicating success across departments, expanding the leadership bench. A leadership failures impact ripples outward, causing direct reports to hunker down, peers to pull back, and cultural damage to deepen. Leaders across the organization begin to doubt its ability to manage change. Growth itself can stall. But organizations can inoculate themselves against this negative feedback loopand accelerate the success of their new leadersby rethinking how they support transitions. Here are four against the grain rules that can help a new leader take root and thrive. 1. QUICK WINS, YES; HERO MODE, NO New leaders often arrive eager to prove themselvesscanning for early wins they can capture. But that motivation can come across as self-serving and raise a red flag for the rest of the team. Early moves should build trust: showing the leader is focused on enabling the team, not spotlighting their own capabilities. The organization can play a critical role here. The new execs own leader should define what success looks likesteering them away from hero moves and toward shared goals, with an emphasis on team visibility, engagement, and inclusion. This matters even more when the new leader is promoted from within. Others may have wanted the role or expected a different outcome. The new leader must address that dynamic directly, affirming each team members value. But its the organizations job to reinforce early and often that the teams success is the leaders success. Thats corporate culture at its best. 2. WHAT GOT YOU HERE WONT GET YOU THERE Most new leaders were promoted because they excelled in a previous role. But that success was likely built on different strengths: technical skill, individual output, or tactical problem-solving. These arent the skills needed to create and communicate a vision and strategy, lead across functions, and navigate complexity. Internal promotions must make a clean break from their old roles or risk blurring the focus on the most important thingwhats next. Without this increased self-awareness, its natural to fall back on whats familiar, especially as the pressure rises. Instead of stepping up to a broader mandate, newly elevated senior leaders tend to double down on execution. It may feel safe, but it stunts growth and signals a lack of readiness. Even worse, the new leader may have blind spots about their own skills and inclinations. Getting 360-degree feedback, even if informal (and if possible, additional psychometric data) is critical to building self-awareness. This is where a neutral confidante, a coach, or even a practice of asking others for their candid feedback, can be invaluable. 3. THE GOOD, THE BAD, AND THE POLITICAL Theres a taboo in most organizations about gossiping to a new boss or appearing to be highly political. But avoiding all such discussion can in fact be dangerous. New leadersespecially external hiresoften walk in blind to personality landmines, power dynamics, and team history. Instead, they need unvarnished insight into who theyll be working with: the stakeholders who are aligned and the ones who clash; the teams that collaborate and synergize; and those that dont. They need clarity on how things get done. Imagine how much more effective you could be in a new role if someone detailed all the team members and their relationshipsalmost like a pregame scouting report analyzing each players mental and physical strengths and vulnerabilities, and the plays they like to run when a games on the line. It would instantly improve your ability to collaborate and record successes. Even better if it will give you insight into how your own work style fits in. Consider a product leader recruited from a fast-moving startup into a legacy organization. In one version of his story, his trademark urgency, focus, and accountabilityincluding take-no-prisoners KPI reviewsalienates peers and they quietly sideline him. In another scenario, hes warned that his bulldog style could backfire, thanks to a hiring manager who was transparent about personalities and fit. Its the organizations jobideally the new bosss bossto reveal this before Day One and keep the conversation going. Every workplace has its unique culture, comprising people, politics, and pressure points. If those arent surfaced early, the new leader struggles to find their way, which at best delays success and in too many instances derails it completely. 4. THERES NO SUCH THING AS MAGIC New leaders are often handed a mandatesometimes clear, more often vague. Theyre expected to bring change and energy, and turn things around. But rarely do they get clarity on pacing and priorities, resourcing, and success measures. Who are their internal customers anyway, and what are their requirements? Consider a COO, hired into a high-growth company with a mandate to increase operational efficiencies. In one version of the story, she immediately overhauls the supply chain strategy and upgrades technologyonly to clash with a CEO whos laser-focused on short-term KPIs. In another version, she seeks clarity on priorities up front, aligns her approach to what the CEO needs to make quarterly numbers, and executes a plan that builds lasting momentum. Without early, specific alignment to stakeholder expectations, a new leader is left uncertain about the playbook, the true priorities, and how aggressively to move forward. Worse, they may act with confidence, but in the wrong directionaiming all their energy at an objective no one else shares. Dont assume talent plus title equals traction. Hiring or promoting the right person and letting the magic happen isnt a strategy. THE BUILDING BLOCKS OF GROWTH Successful leadership transitions are essential to a companys growth in the same way that new routers are essential to an expanding telecommunications network. Scaling up requires that each connection works. When they dont, the existing structure becomes overburdened, everything slows down, and the limits to growth become painfully clear. All leadership transitions bear risk. Most of the risks are identifiable and predictable, and afford opportunities to apply strategies that mitigate them. Taking deliberate steps to make transitions go well wont eliminate that risk, but it will shift the odds in our favor. The only thing more risky than changes in leadership is leaving leaders to face them alone.
Category:
E-Commerce
One hundred miles off the coast of New York City, there is an underwater canyon teeming with marine life. Seabirds soar overhead as whales, sharks, dolphins, sea turtles, and fish gather around Hudson Canyon. With so many species calling the canyon home, the Wildlife Conservation Society wants Hudson Canyon to be designated a National Marine Sanctuary. The designation, awarded by the National Oceanic and Atmospheric Administration, would protect the ecologically diverse area from companies hoping to mine the seabed for oil, gas, and minerals. Its not just the endangered species WCS is hoping to save from disruptive and dangerous miningit also wants to save the fish you eat for dinner. With a striking new campaign created by the advertising agency McKinney, WCS is calling on seafood lovers to sign its petition urging NOAA to protect Hudson Canyon, home to the creatures that stock seafood markets in New York City and beyond. [Image: courtesy Hudson Canyon] Were protecting the species out there, were protecting their health, but were also protecting the economic viability of our waters, says Christine Osekoski, executive director of the Wildlife Conservation Society. To help communicate the importance of Hudson Canyon to the people who enjoy the spoils of commercial fishing there, McKinney took an analog-first approach to the campaign. They printed the petition right onto the butcher paper that seafood markets wrap around the fish they sell. [Image: courtesy Hudson Canyon] What better way to get the actual cause, actual information, and actual petition into peoples hands . . . than at the moment you are consuming the very thing that is being threatened? asks Omid Amidi, chief creative officer at McKinney. To create an eye-catching design on the butcher paper, McKinneys team members used a Japanese printing technique called gyotaku, brushing the types of animals found in the Hudson Canyon with blue ink and pressing them onto paper. The process yields nearly perfect impressions of the very same creatures the campaign is trying to saveblack sea bass, scallops, and crabs, for example. [Image: courtesy Hudson Canyon] The fish prints are paired with maps of the Hudson Canyon, copies of the petition text, and QR codes to sign it. These elements, all in blue, are overlaid with blocky red letters reading Quit Floundering, Then Save the Canyon and Save the Scallops, Then Sear Them, among other sayings. The simple layouts and contrasting blue and red ink are meant to evoke the advertising and storefront design choices of old New York fish markets. The end product is a far cry from the plain brown butcher paper that markets traditionally use to wrap seafood. [Image: courtesy Hudson Canyon] The design itself is just meant to stop you in your tracks, Amidi says. Even though its a light piece of paper, it has the weight of all the work and all the care we put into it. Adding to the campaign, McKinney designed window clings and counter cards for participating markets, as well as created signage displayed at the New York Aquarium and online videos featuring local fishmongers supporting the effort. [Image: courtesy Hudson Canyon] The campaign launched June 9, the day after the United Nations World Oceans Day. Since then, participating seafood markets in the New York City area have wrapped their fish in WCSs petition and stirred up support mong customers. Six markets are participating in the campaign: Mt. Kisco Seafood, Greenpoint Fish and Lobster, Metro Seafood, Mermaids Garden Sustainable Seafood, Martys Gourmet Seafood, and Lobster Place at Chelsea Markets. We definitely have a crew of loyal customers who are into sustainability, says David Seigal, culinary director at Lobster Place at Chelsea Markets. But we also have a lot of customers who want to know where their food is coming from, and I think those are the people who are most interested in this. Some participating fish markets are already asking for more shipments of the paper, Osekoski says, as more people see the design and sign the petition. This show of support is an important step in the process toward Hudson Canyon being designated a National Marine Sanctuary. Soon, NOAA will release its draft of the designation documents and solicit comments from the public before ultimately choosing whether to make the area a sanctuary. By the end of the public comment period, WCS hopes its petition will have 25,000 signatoriesand the nonprofit is already one-third of the way there. For Seigal, also an avid fisherman who frequently travels to the Hudson Canyon, protecting the area is a cause especially close to his heart. Were in business with Mother Nature, when it comes down to it, he says. Any threat to Mother Nature is a threat to, at a minimum, our business, but really to our existence as a human race.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||