Andreessen Horowitz

It’s “a brand new day” in crypto, Ben Horowitz says ??

Marketing is becoming one of the most exciting frontiers for generative AI. Why? It’s iterative, creative, and thrives on text, image, and video – the building blocks of GenAI. Seema Amble explains a key shift: AI is moving marketers from 'doers'—creating emails and campaigns—to 'reviewers,' overseeing AI-driven workflows. Watch below ??

Intelligent automation is reshaping internal operations, turning manual, repetitive tasks into scalable workflows powered by AI. The winning formula for startups in this space? 1?? Horizontal AI Enablers: Develop reusable, foundational tools like data parsing or integration solutions that solve industry-wide problems. 2?? Vertical Automation Solutions: Focus on industry-specific workflows, replacing manual processes with automation to generate immediate business value. Startups leveraging these strategies are poised to transform industries and unlock massive market opportunities. For further insights on intelligent automation’s potential, as well “LLMflation,” AGI, GPU access and more, check out the a16z Enterprise newsletter here ??

The next era of sales is unfolding, thanks to AI. AI-native systems of record are not just improving efficiency—they’re fundamentally transforming how organizations capture, analyze, and act on data. a16z’s Marc Andrusko and Joe Schmidt IV share insights on how AI-native systems are reshaping organizations by integrating multimodal feedback from tools like email, Slack, and NPS surveys. They explain how these systems break down silos—such as the sales vs. marketing debacle—to align teams around a shared goal: the customer. ??

Could Las Vegas become the safest city in America?? In this episode of "The Ben and Marc Show,” Marc Andreessen and Ben Horowitz sit down with Chief Mike Gennaro and Sheriff Kevin McMahill of the Las Vegas Metro Police Department (LVMPD) to discuss the intersection of technology and public safety. From crime prevention to building trust in the community, the four discuss how new technology -- such as drones and automatic license plate recognition -- have helped the LVMPD reduce crime, improve efficiency, and strengthen relationships with residents. They also discuss how police departments can break the cycle of recidivism, address addiction and mental health issues, and help former offenders rebuild their lives in productive ways. This is more than just using data and technology to transform policing and public safety. It's how community-focused policing has the power to change the culture in America.

Over the next few months, expect to hear a range of speculation and “hot takes” about what will happen in the Crypto industry, from both a legislative and regulatory perspective. It’s too early to tell how everything will shake out. But we’re very optimistic that now, the government will foster innovation and accelerate progress. This will enable a future where we can deliver on the many consumer benefits that we’ve been excited about: giving people ownership over their digital identities, new business models for creatives, low to no-fee cross border transactions with stablecoins, new ways for small businesses like restaurants to engage with their customers, the emergence of decentralized social networks, development of physical infrastructure like energy grids, as well as blockchains democratizing AI and games, and so much more that we can’t even imagine yet. More from Miles Jennings, Michele Korver and Brian Quintenz, at the post in our comments ??

The future of AI in operations goes far beyond automating repetitive tasks. We believe there is substantial opportunity not only for AI to turn labor into software but also to productize internal work within organizations, often referred to as “ops.” Ops roles involve essential but often repetitive and mundane tasks. Companies have historically attempted to automate these tasks with Robotic Process Automation (RPA), but RPA efforts didn’t always deliver the desired results. Unlike RPA, which rigidly follows specific actions, AI agents can understand goals and dynamically choose actions to achieve them. With generative AI, we believe true automation through adaptable agents is now possible. And we’ve already seen examples of companies using AI agents to fulfill RPA’s original promise, transforming internal operations into more flexible, scalable workflows. We believe this is the future of AI ops, and where the opportunity lies. More from Kimberly Tan: https://lnkd.in/eHGcj9fy

The past four years had a greater impact on tech than any other period during his lifetime, Marc Andreessen says. And the latest election results feel like “a boot off the throat,” he says. So now that President Donald Trump has been re-elected, what’s about to change? According to Marc and Ben, likely a lot—especially in AI, crypto, and defense. They break down what they expect to see, and provide a full post-election recap, in the latest episode of The Ben & Marc Show, out now. Link to full episode in comments.

The cost of high-quality LLM inference has been plummeting, a trend Guido Appenzeller calls "LLMflation." By tracking the per-token costs of comparable models, he identified a 10x annual cost decrease over the past 3 years. The numbers speak for themselves: in 2021, GPT-3 cost $60 per million tokens. Today, a model of comparable performance costs only $0.06 per million tokens—1,000 times less! The cost of newer, better models, comparable to GPT-4, seems to follow a similar curve. Why? A few major factors are driving this dramatic cost decrease: - Hardware advances: Improved GPU performance and structural optimizations have driven down costs. - Model quantization: Moving from 16-bit to 4-bit precision has increased efficiency and reduced resource demands. - Software optimizations: New methods reduce compute needs and memory bottlenecks. - Smaller, more powerful models: Training improvements mean smaller models now outperform past giants, lowering computational load. - Open-source competition: Open models from companies like Meta and Mistral AI have intensified competition, driving prices lower. The result? LLM inference is essentially free—opening up new AI use cases that would’ve been commercially unfeasible just a few years ago. More from Guido Appenzeller: https://lnkd.in/ebtcPeDT

Could AI have saved TD Bank $3 billion? Outdated and underfunded systems left critical transactions unmonitored and alerts missed, resulting in a record fine. As Gen AI advances, it’s clear: adopting new solutions isn’t just about compliance—it may be a competitive advantage. Check out more from James da Costa, in today’s Fintech newsletter ??