Good morning {{first_name| Artificial intelligence lovers}}. Amid intense competition with Anthropy, improved models, and efforts to eliminate “side gigs,” OpenAI appears to be looking for something closer to home – an AI-powered agent phone.
Analyst Ming-Chi Kuo says the company is working on accelerating this device to 2027, with some notable capabilities. Good news for those who want a more powerful AI experience in their pocket. The question is: Where does this leave ongoing work with Jony Ive? Or is this the same device?
-
OpenAI is fast-tracking its “AI agent phone.”
-
Anthropic’s AI agents for financial work
-
Make your Notion agents more autonomous
-
Home “micro” AI data centers are coming
-
4 new tools for AI, community workflows, and more
Obinay
Image source: Images 2.0 / Rundown
Rundown: It is said that OpenAI acceleration Developing its first AI phone, it now aims for mass production in the first half of 2027, a full year earlier than previously reported, according to supply chain analyst Ming-Chi Kuo.
-
Kuo says the shift in timeline is likely driven by OAI’s IPO ambitions (powerful hardware could boost investor pitch) and increased competition in AI phones.
-
The phone’s standout specifications will be an image signal processor, with an improved HDR line to improve true optical sensing for AI clients.
-
MediaTek is positioned as the sole chip supplier, with the device using two AI processors to handle vision and language tasks simultaneously.
-
Kuo also added that 2027-2028 combined OpenAI shipments of this phone could reach 30 million, if development stays on track.
Why it matters: Controlling the hardware and operating system can be the key to a true proxy phone. But if OpenAI’s AI phone is closer than we thought, where does that leave a device built using Jony Ive’s io system? OpenAI acquired io was released last year to great fanfare to cross screens, but nothing concrete has emerged yet Except for some rumours.
Along with lambda
Rundown: Most large-scale training runs at only 35-45% of the FLOPS model usage, which means teams are paying for more than twice the compute they actually use. Lambda engineers benchmarked Llama 3.1 models from 8B to 405B on NVIDIA Blackwell GPUs to trace the loss of efficiency to its root causes.
Efficiency losses were attributed to:
-
Memory overhead determines effective throughput
-
Parallelization strategies are hardware incompatible
-
Serial communications stops GPU cycles
Anthropic
Rundown: Anthropic only unveil 10 ready-to-play AI agents aimed directly at financial services and insurance – capable of handling work ranging from creating promotional brochures and checking KYC files to reviewing earnings and valuations.
-
Each agent comes with skills and instructions specific to the task area, links to relevant data sources, and additional CLOUD forms for subtasks.
-
Companies can adapt any agent of their choice to their modeling agreements, risk policies, and approval flows – while staying informed 24/7.
-
Agents can be used as plugins within Claude Cowork or Claude Code on desktop, or as cookbooks, running as managed agents on the Claude platform.
-
Claude also gets Microsoft 365 add-on as well as data connectors from Dun and Bradstreet, Verisk, IBISWorld, and other financial services partners.
Why it matters: development, Cyber security, designAnd now financing. Anthropic jumps across industries, meeting companies where they are rather than selling a generic model and letting them figure it out. that it New joint venture worth $1.5 billion Along with Wall Street giants, it is promoting this strategy, further fueling its race with OpenAI.
Artificial intelligence training
Rundown: In this guide, you’ll learn about a hidden workflow that will make your Notion agents more independent and powerful than they are by default. This allows you to wake up any agent, give it tasks, and then get a report on what it did.
-
Create a Notion agent from the “Agents” option in the sidebar. Then open a database of claims and task reports for the agent. We have what are called “reports”.
-
Click the New drop-down menu, open Templates, and create a @Today template (ours was Daily Summary). @Today makes iterators inherit the current date
-
Set the properties, type the task instructions in the body of the page, and @mention the agent, so that the repetition triggers it. Remember to stop the agent when you do this, so it doesn’t overwrite the template
-
From the template, click New Template, then Repeat, and choose a tempo. We turn on our website daily at 7am. A small blue icon next to the template shows that it is live
Pro Tip: Try setting up daily reports, weekly reports, and email automation. You can now route them all through the same planning agent.
Provided by IBM
Rundown: Most leaders know that AI will reshape their businesses, but few have a clear guide on how to do it. IBM Business Value Institute analysis reveals 5 plays CEOs must implement now to capture gains by 2030.
To lead in the AI-first landscape, the CEOs surveyed suggest the following:
-
Customize your AI mix, not just your AI models
-
Hire a Chief AI Officer if you haven’t already
-
Coordination of intelligence – artificial and human
Artificial intelligence data centers
Rundown: California startup Span is As a team With Nvidia to install mini-AI data centers on the walls of residential homes and small businesses, leveraging unused electrical capacity in local networks to meet growing demand for AI computing.
-
Span has developed XFRA, small computing nodes that are mounted on the exterior walls of homes, along with HVAC and electrical systems.
-
Nvidia supplies liquid-cooled RTX PRO 6000 Blackwell Server Edition GPUs to power each XFRA box, ensuring noise-free compute for AI workloads.
-
Spann said CNBC It can install 8,000 XFRA units six times faster and at one-fifth the cost of building a similar 100-megawatt central facility.
-
The company is currently working with PulteGroup, one of the largest homebuilders in the United States, to test the fund and its economics in newly built communities.
Why it matters: Network stress from data centers is real, and Span boxes can distribute the load while only exploiting unused capacity. But the general response is an open question – not everyone will like the idea of installing a data center box where kids play, especially when the alternatives are e.g circumference-And Satellite Data centers are also on the horizon.
-
🚀 Fund automation – Orchestrate workflow automation securely and at scale*
-
🎨 Pomelli – Create product-specific campaigns using Google’s AI marketing tool
-
🖼 UNI 1.1 – Create new Luma images and edit AI that is close to the limits
-
🎮 Astrocad – A platform for creating shareable games using artificial intelligence
OpenAI GPT-5.5-instant He started It is being released to all ChatGPT users, offering improved performance, stronger memory, and more personalized and concise responses.
Microsoft Expanded Its Copilot Cowork agent system for iOS and Android, adding built-in skills for common tasks and data plugins for business systems.
apple Agreed Apple paid a collective $250 million to some US iPhone buyers to settle a class-action lawsuit over misleading claims about its new AI Siri, but did not admit to any wrongdoing.
Confusion Amnesty International Fired Computer for Professional Finance provides licensed data and 35 custom workflows to its proxy system to help analysts handle routine work.
Anthropic It is said committed To spend $200 billion on Google’s cloud and chips over the next five years, now generating more than 40% of Google’s cumulative revenue.
Coinbase CEO Brian Armstrong He said The company is cutting 14% of its workforce, roughly 700 people, as it shifts to AI-driven teams, agent-driven workflows, and leaner operations.
In each newsletter we show how the reader is using AI to work smarter, save time or make life easier.
Today’s workflow comes from the reader Shi W. In Springfield, Illinois:
“I went to a music festival – multiple stages, music over several days. Many of the bands had more than one set time/location. I provided Claude with the festival schedule in PDF format and asked him to extract the information, but ignore the signings and music workshops. I then read a list of bands I wanted to see and followed up with a subset of those I absolutely did not want to miss. I asked for a schedule outlining where I needed to be to increase the number of bands I could reasonably see with a simple prompt.
I understood the task and in one fell swoop produced a great spreadsheet to optimize my time and indicate my priorities, phases, times, sub-sets, if the band had another set, and explained the rationale for making the decision.
How do you use artificial intelligence? Tell us here.
That’s all for today!
Before you go, we’d love to know what you thought of today’s newsletter to help us improve The Rundown experience for you.
Rowan, Joey, Zack, Shubham and Jennifer – the humans behind The Rundown
