 |
|
|
|||||
Is America in a Second Gilded Age? Evan Osnos thinks so. In this episode of Most Innovative Companies, Osnos unpacks how extreme wealth, corporate influence, and political inequality are transforming American life. If youve ever wondered how the 1% really operate, this is the deep dive you need.
Category:
E-Commerce
Cheap or free access to AI models keeps improving, with Google the latest firm to make its newest models available to all users, not just paying ones. But that access comes with one cost: the environment. In a new study, German researchers tested 14 large language models (LLMs) of various sizes from leading developers such as Meta, Alibaba, and others. Each model answered 1,000 difficult academic questions spanning topics from world history to advanced mathematics. The tests ran on a powerful, energy-intensive NVIDIA A100 GPU, using a specialized framework to precisely measure electricity consumption per answer. This data was then converted into carbon dioxide equivalent emissions, providing a clear comparison of each models environmental impact. The researchers found that many LLMs are far more powerful than needed for everyday queries. Smaller, less energy-hungry models can answer many factual questions just as well. The carbon and water footprints of a single prompt vary dramatically depending on model size and task type. Prompts requiring reasoning, which force models to think aloud, are especially polluting because they generate many more tokens. One model, Cogito, topped the accuracy tableanswering nearly 85% of questions correctlybut produced three times more emissions than similar-sized models, highlighting a trade-off rarely visible to AI developers or users. (Cogito did not respond to a request for comment.) Do we really need a 400-billion parameter GPT model to answer when World War II was, for example, says Maximilian Dauner, a researcher at Hochschule München University of Applied Sciences and one of the studys authors. The results underscored the balance between accuracy and emissions. The least-polluting model tested, Qwen 7B, answered just one in three questions correctly but emitted only 27.7 grams of carbon dioxide equivalent. In contrast, Deepseeks R1 70B reasoning model answered nearly eight in 10 questions correctlywhile producing more than 70 times the emissions for the same workload. The type of question also affects environmental impact. Algebra or philosophy prompts produced emissions up to six times higher than what a high school student would generate getting homework help. Companies should be more transparent about the real emissions and water consumptions from prompts, says Dauner. But at the same time, users ought to be more awareand more judiciousabout their AI use.
Category:
E-Commerce
For two decades, conventional startup wisdom followed a simple mantra: Build one killer feature, win a devoted audience, and expand later. Startup luminaries like Paul Graham and Sam Altman championed this approach. They argued it was better to start small, focus narrowly, and earn the right to grow. Many of todays most iconic companies followed this path, launching with a single tool that eventually evolved into a suite. But what if the problem youre solving doesnt fit neatly into one feature? What if starting small just means setting yourself up to rebuild everything later? In an increasingly complex world, customers no longer have patience for point solutions or disconnected workflows. They expect products to understand how their systems work and to match that reality with integrated, end-to-end experiences. Thats why a new model is emerging. The rise of the compound startup The compound startup, a term popularized by Rippling CEO Parker Conrad, describes a company that builds multiple, deeply integrated products from day one. In a January interview with Y Combinator CEO Garry Tan, Conrad explained that the goal of a compound startup is to address systems of problems, not just isolated pain points. It is a model built for how people and businesses actually operate. Most business functions dont exist in a vacuum. In Ripplings case, payroll connects to benefits, onboarding, compliance, IT provisioning, and more. Customers dont want to piece together tools to manage each of those functions. They want a system that works together out of the box. At april, we didnt just stitch together a few tax tools. We built an entire suite of products from the ground up. Filing, forecasting, planning, optimizationall designed to work in sync, and all tailored to distinct taxpayer segments like investors, small business owners, gig workers, and everyday banking customers. We chose to build in tax, one of the most complex, fragmented, and regulated categories in fintech, because the problem demanded a compound solution. Tax laws shift constantly. Each state and jurisdiction operates differently. We could have licensed a white-label provider and shipped faster. Instead, we built our own tax engine, became the first new nationally licensed e-file provider in more than 15 years, and now operate a full-stack platform with fewer external dependencies. That decision has given us speed, adaptability, and product depth our partners cant find elsewhere. Why compound startups make more sense today The shift toward compound startups isnt just philosophical. Its practical. Todays challenges rarely sit in one lane. Managing personal finances touches tax, payroll, planning, and compliance. Running a business involves HR, inventory, scheduling, payments, and reportingall at once. Point solutions force users to become their own system integrators. They juggle multiple tools, manage disconnected data, and learn mismatched interfaces. Compound startups flip that script. They build coherence into the product architecture itself and unlock several key advantages: Unified data: Integrated platforms break down silos and allow smarter decision making across use cases. At april, data moves with consent across workflows, powering real-time tax insights, planning simulations, and filing automation. Shared UX patterns: A consistent interface builds user trust and reduces friction. Most april users complete their return in under 23 minutesa far cry from the 13-hour average reported by the IRS. Durable switching costs: When workflows span multiple integrated tools, the platform becomes stickier and more valuable as a whole. Platform-wide network effects: When more users adopt more of the suite, value compounds across use cases. Compound startups dont just solve tasks. They solve workflows. And that makes them more durable, more useful, and more differentiated in crowded markets. The long-term payoff Of course, this approach comes with tradeoffs. Building multiple products in parallel strains focus and burns capital faster. It forces earlier decisions around architecture, compliance, and team structure. Its not the right move for every startup. But for founders tackling systems-level problems, the risk of starting too small is greater. You cant increment your way to coherence. Weve seen the payoff at april. Our compound architecture has allowed us to respond faster, deliver richer experiences, and scale without compromise. The future is compound The startup playbook is evolving because the problems were solving have evolved. Systems are messier. Users expect more. Point solutions cant keep up. Founders shouldnt be afraid to build big from the start. The world doesnt need more single-purpose tools. It needs products that actually solve the full problem. The future isnt just compound. Its integrated, full-stack, and built to scale from day one. Ben Borodach is cofounder and CEO at april.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||