About this episode This episode explores OpenAI's groundbreaking 'Dots' project, a revolutionary step toward transforming chatbots into autonomous AI agents capable of multi-tasking and decision-making. We uncover why this shift matters for industries relying on conversational AI, highlighting how it enhances efficiency and user experience. Listeners will gain insights into the potential threats to traditional chatbot providers and understand how AI agents are poised to reshape the future of digital interaction and automation.
0:00 openai just announced chatgpt has 1,200,000,000 weekly active users and it gave every single one of them an autonomous assistant in episode 23 we talked about ai models hitting new milestones in math and science that was about building a smarter tool this is different this is about giving the tool a job a budget and a set of keys0:23 to your office the era of the chatbot is officially over the era of the agent has begun openai's big dev day announcement is a feature called dots don't let the cute name fool you a dot is an ai agent powered by their new models that gets its own dedicated cloud computer it connects to over 4,000 apps plus0:45 slack and teams you give it a task or better yet a responsibility an ongoing objective you set the boundaries what it can do on its own what needs your approval and what it must never touch then it just goes to work one early tester reported their dot proactively argued with a service provider and cut $500 from their1:07 annual bills developers can hand off bug triage and failing bills it's not a helper it's a tireless autonomous employee that you don't have to pay overtime and the work your primary dot does it doesn't even draw from your plan's usage limits this is not a feature it's a shadow workforce to power this they launched gpt 6.1 sole1:32 the headline is near astra intelligence for a fifth of the price but that's not the real story the real story is the price itself $2 for input $10 for output per million tokens why does that matter because agents aren't single shot queries an agent has to think it interprets your request decides to use a tool inspects the1:53 result decides the next step maybe uses another tool each step burns tokens at previous prices that was economically insane for most tasks at this price it becomes viable it becomes the default openai just subsidized the entire agent economy into existence this is the context for everything else they announced a new ultra fast mode for low latency tasks2:20 a decisions api a marketplace for agents and something called chatgpt spaces a shared workspace for humans and ai agents to collaborate they aren't just shipping a model they are building the entire operating system for this new workforce while openai was building an army of digital workers anthropic was making a move on washington there was no flashy product2:44 launch no new model just a quiet strategic declaration reported by techradar anthropic wants to get its ai claude working across all arms of the government let's be very clear about what that means this is not a pitch to get a better search bar on a government website all arms means civilian agencies yes but it also implies defense3:07 intelligence high stakes high security high trust environments where the cost of being wrong is catastrophic anthropic is playing a different game openai is building a platform for the masses for developers for startups for you they are building a decentralized chaotic explosive cambrian explosion of ai agents it's a volume play anthropic is aiming for the center they are3:35 positioning claude not as a tool but as a trusted institutional partner the kind of system you would embed in critical infrastructure the kind of system you would consult before making a decision that affects millions of people they are selling safety reliability and alignment as their core features while openai is building the factory floor anthropic is trying to3:56 build the boardroom so forget the benchmarks for a minute forget which model writes a better poem the ground has shifted this is no longer a simple race for the best model it's a platform war fought on two completely different fronts openai is building an open sprawling digital economy of autonomous agents anthropic is trying to become the secure4:18 centralized intelligence layer for the state itself the question is no longer about who has the smartest ai it's about who you trust to run the system
