OpenAI is introducing what it is calling, naturally, "stateful AI." The company has announced that it will soon offer a stateful runtime environment in partnership with Amazon, built to simplify the process of getting AI agents into production. It will run natively on Amazon Bedrock, be tailored for agentic workflows, and optimized for AWS infrastructure.
Depending on the list of functionalities that you have, or the areas where you have to explore things and analyze, it will automatically create [AI] agents, micro [AI] agents, fulfill the need, and then, when it comes to actual borrower engagement, provide everything to the front-end agent.
Running AI infrastructure costs are astronomical. Back in 2023, it was estimated that OpenAI spends around $700,000 per day to run ChatGPT—about 36 cents per query. However, in 2024 with the release of its higher-performing o3 model, some queries cost over $1,000 of computing power. Consequently, OpenAI CEO Sam Altman reports the company is even losing money on its $200 ChatGPT Pro subscriptions.
In our own environment, over 90% of targeted Level 1 volume is handled autonomously, with resolution rates above 99% for those categories and materially faster than human-only workflows. It accomplishes this by having the Autonomous Workforce operate on top of the live configuration management database (CMDB), active workflows, policy engines, approval chains, and real transaction history - all updated in real time every time a ticket closes, a workflow executes, or a policy changes.
An AI agent is simply a model that receives input, follows defined goals and rules, makes step-by-step decisions, and uses tools to take actions. Instead of viewing AI agents as autonomous digital workers, break it down. First principles thinking says this definition captures the essence of how AI agents function and operate within business workflows.
Multiagent orchestration logic behind Computer is that, rather than becoming general-purpose tools, AI models have instead branched off into different specialties: Anthropic's Claude, for example, is famously popular among software engineers. Relying on a single model to complete a complex task is therefore a bit like trying to assemble an Ikea dining table using a butter knife; it could be possible, but the finished product is going to be a little wonky.
Atlassian has been in the business, for decades, of collaboration software helping people get work done. Now, you enter agents, and agents are now doing a lot of that work, and so you want to be able to coordinate between humans and agents.
With this expansion, Opal is shifting from a low-code orchestration tool to a platform in which an AI agent independently determines which actions, tools, and models are needed to achieve a goal. The new agent runs on the Gemini 3 Flash model and automatically selects the tools needed to complete a task.
Powered by the Gemini 3 Flash model, the new agent in Opal enables autonomous workflows that plan, reason, and execute on the user's behalf. Now, instead of manually picking a model, developers can select an agent in the generate step. The agent then triggers the right tools and models, such as Web Search for research or Veo for video, needed to accomplish the user's intended mission.
The new battleground in banking is intelligent operations and scalable execution. In 2026, banking is about moving money smarter, faster, and with fewer humans in the middle. Across corporate finance and global retail operations, banks are experimenting with technology and operational design in ways that challenge long-held assumptions about scale, speed, and control. Three recent developments exemplify what's happening in money movement: Goldman Sachs deploying AI agents, Truist automating corporate receivables, and Nubank expanding abroad with a lean digital model.
Last year, one category AI absolutely dominated was being an extremely agreeable coworker. While this might sound nice, this can turn into a problem for founders who rely on AI as their only teammate. When your head of legal, HR, and supply operations are all AI agents, unsubstantiated flattery can create costly blind spots. That's one of the reasons OpenAI said goodbye to its " yes-man" version of ChatGPT, and why some AI-powered solo founders are training their tools to push back.
Just consider a typical day in the life of a modern human: you glance at your phone while waiting for coffee to brew, skim headlines while half-listening to a podcast, mentally rehearse a client pitch while walking your child to school, reply "noted" on Slack during a meeting while updating a slide deck, check your bank balance while standing in line,
It isn't a universal truth, but a vast number of goods and services have their own full-circle moments. While there are still plenty of travel agencies in the U.S., the overall number is still down considerably from a peak in the 1980s. For some industry forecasters, though, the future looks a lot like the recent past, except that instead of travelers trusting human agents with making their travel plans a reality, they'll use AI agents for the same purpose.
As enterprise platforms rush to add conversational bots into workflows, they're also inadvertently giving those agents broad access to sensitive information - and, in some cases, letting bots chat freely in a way no privacy or marketing team would ever approve. This is exactly the type of hidden pitfall Aaron Costello, chief of SaaS security research at AppOmni, hunts for.
The episode examines what's new: AI-agent coding tools that can work in the background like personal assistants. Warzel is joined by Anil Dash, a longtime technologist, to unpack how hype and venture-capital incentives can distort the conversation around advances, and what the rise of tools like Claude Code and the more reckless "OpenClaw" experiments mean for labor, security, and everyday work.
We began GitHub Agentic Workflows as an investigation into a simple question: what does repository automation with strong guardrails look like in the era of AI coding agents? A natural place to start was GitHub Actions, the heart of scalable repository automation on GitHub. GitHub Agentic Workflows leverage LLMs' natural language understanding to let developers define automation goals in simple Markdown files describing the desired outcome.
The bot, designated MJ Rathbun or crabby rathbun (its GitHub account name), apparently attempted to change Shambaugh's mind by publicly criticizing him in a now-removed blog post that the automated software appears to have generated and posted to its website. We say "apparently" because it's also possible that the human who created the agent wrote the post themselves, or prompted an AI tool to write the post, and made it look like it the bot constructed it on its own.
Have you ever asked Alexa to remind you to send a WhatsApp message at a determined hour? And then you just wonder, 'Why can't Alexa just send the message herself? Or the incredible frustration when you use an app to plan a trip, only to have to jump to your calendar/booking website/tour/bank account instead of your AI assistant doing it all? Well, exactly this gap between AI automation and human action is what the agent-to-agent (A2A) protocol aims to address. With the introduction of AI Agents, the next step of evolution seemed to be communication. But when communication between machines and humans is already here, what's left?
Take2 - $14M Series A Take2, an AI agent network that automates end-to-end healthcare recruiting tasks, has raised $14M in Series A funding led by Human Capital. Founded by Yaniv Shimoni and Kaushik Narasimhan in 2023, Take2 has now raised a total of $14M in reported equity funding. AlleyWatch is NYC's leading source of tech and startup news, reaching the city's most active founders, investors, and tech leaders. Advertise today →
As enterprises shift from experimenting with AI agents to deploying them in production environments, Google is rolling out a structured skills program aimed at helping developers build, test, and operationalize AI agents using Google Cloud tools, specifically its Agent Development Kit (ADK). Named the Gemini Enterprise Agent Ready ( GEAR) program, the initiative packages hands-on labs, 35 free monthly recurring Google Skills credits, and badge-earning pathways into a track within the Google Developer Program.
Errors can quickly become a bottleneck if hallucinations multiply when agents interact, said Nicolas Darveau-Garneau, a former Google executive and author of the book " Be a Sequoia, Not a Bonsai." If a single agent has a 5% hallucination rate, then it's hard to daisy-chain multiple agents without a high risk of errors. That's because the risk increases exponentially, he told Business Insider.
What Is the Agent Internet? In early 2026, a new layer of the internet has emerged -- one built by and for AI agents. Over 95 platforms now exist where autonomous AI agents communicate, trade, create, play, govern, and conduct research. This is not a speculative whitepaper. It is happening right now. The Agent Internet is a decentralized network of platforms where AI agents -- not humans -- are the primary users.
A2UI (Agent to UI) is a UI protocol from Google that lets AI agents generate user interfaces on demand. It introduces declarative mini-apps where UI components and actions are defined in a schema, and the agent can operate them automatically. Think Telegram-style mini-apps: small, self-contained interfaces that work without custom integration code. Instead of a long question-and-answer loop, agents can now send interactive, native interfaces directly to the client.
