• Dotting the App Landscape
  • White House Saunters into AI Safety
  • Samsung Brings a Bread Knife to a Gun Fight

Dotting the App Landscape
While the new Dots agent platform is arguably the biggest announcement out of OpenAI's Dev Day, launching yet another agentic assistant platform isn't really big news in tech anymore. Everyone seems to have one now. Most of them are more or less sophisticated versions of the open-source OpenClaw project that caught the world by surprise earlier this year. Dots in themselves aren't super interesting; like most other agents, you can converse with them via messaging applications like Slack and Teams or initiate voice calls to chat with them. You can assign them to work on tasks that require multiple steps and even multiple days to complete. The most interesting aspect of the Dots launch is the tie-in with apps. Dots are said to have the ability to integrate with more than 4,000 apps. Users can choose the apps to which they want to grant Dots access and whether they want them to have direct access or operate in a separate cloud environment that won't require a users' laptop or computer to remain on full-time. Dots can maintain connections with apps over time, running background tasks and checking in on apps on a schedule or as delegated tasks require. OpenAI gave some examples of how using Dots in apps can provide value to a user:

  • Review customer feedback, identify issues, and help prepare code fixes to address them
  • Track changes to a product launch and prep revised drafts
  • Compare account history and customer requirements against product documentation, then flag outstanding concerns

Because Dots can run in their own cloud computer, you don't necessarily have to give them access to your personal computer. But you can also choose to give Dots access to certain files on your computer to complete tasks and revoke access when those tasks are complete.

If it seems like this is being aimed directly at enterprise customers, you'd be right on the money. Enterprise is an area where OpenAI has lagged behind Anthropic recently. The pitch is that Dots can be connected to the apps used in an enterprise and can therefore be tailored and optimized for company-specific workflows. This is where being able to integrate with Slack and Teams becomes important; Dots can function like co-workers or junior assistants to help with daily tasks or complete ongoing tasks on their own. While there does not seem to be a comprehensive list of the 4,000 apps that can work with Dots, the apps that already work through ChatGPT plugins are a known entity and include apps like SharePoint, OneDrive, Outlook, GitHub, Dropbox, Salesforce, Databricks, and many others. But for regular people interested in trying out Dots, there is also integration with the Google suite - Gmail, Calendar, Google Drive, etc. OpenAI really seems to be making a pitch here to developers as well that launching apps on their platform is as viable as launching them in app stores. Every app that is added to the ChatGPT platform is an app that can be accessed by Dots, and as apps are made available on the platform, ChatGPT can recommend those apps to users who can then make use of those apps while taking advantage of the information ChatGPT already has about them to personalize and optimize the experience. For some apps, there might not be any need to access them via the web or in desktop apps anymore. People can even sign into apps using ChatGPT much as many people use Google or Apple to log into apps today. As more apps are integrated into ChatGPT, the ability to expand as an app platform grows. Subscriptions and services could be initiated right inside ChatGPT, which could allow users to escape the Apple and Google taxes imposed by the mobile platform stores (at least while OpenAI isn’t applying a tax of their own).

This is going to be an interesting angle to watch. All the big tech companies want freedom from the phone and the tyranny of the app stores. Up until now, they have looked to the possibility of their own hardware to facilitate their escape from the walled gardens. OpenAI might be on to something here, for once taking advantage of the fact that they have the name most akin to Kleenex in this space.

White House Saunters Into AI Safety
On September 29, 2026, the US president met with the leaders of the frontier AI labs (i.e. Open AI's Greg Brockman, Anthropic's Dario Amodei, Google's Sundar Pichai, NVIDIA's Jensen Huang, Meta's Mark Zuckerberg, and SpaceXAI's Elon Musk) and got them to sign a "morally binding" security framework for advanced AI, which the agreement also says should now be called "super intelligence". There are four basic tenets to this agreement:

  • Internal controls to monitor model capabilities and behavior during training and deployment, including risks involving cybersecurity, biological and chemical threats, and unintended access to computer systems.
  • An internal review team to check that those controls work and that problems are fixed.
  • An independent external auditor or evaluator to assess the controls.
  • An independent board committee to receive internal and external reports and oversee fixes

It's really important to note that there are no consequences for the tech companies if they fail to adhere to this agreement. There are no timeframes for implementation, no penalties for failing to comply, and no reporting on their activities is required. A committee was suggested to provide guidance and ensure compliance, but it remains just an idea and not anything that is being moved forward at this time. Even the 4 tenets themselves are loosely defined. The internal controls include capability monitoring, alignment monitoring, cybersecurity safeguards, biosecurity safeguards, chemical threat safeguards, prevention of unintended system access, detection and remediation, and the creation of an internal oversight team. There are no given parameters or expectations for any of these controls, so in theory, each company could say they're living into them without changing much of what they're already doing today. The same vagueness applies to the tenets on the internal safety team and the independent external auditor. Even the term "independent" isn't defined and there are no specifications on how these groups should evaluate the companies or what access they should be granted to conduct effective reviews. And nothing is spelled out concerning next steps if internal or external teams should find any issues worth reporting.

Unless these companies actually choose to work together to flesh out some of these details (by the way, meeting regularly is part of the agreement as well, but of course the frequency and nature of the meetings is not delineated), this agreement doesn't seem to be worth the paper on which it was printed, save perhaps for having the signatures of all these men and the fact that there is a typo on the signature page - much like those rare stamps with the upside-down plane, that combination might be worth some money to some collector somewhere someday. But if anyone hoped that this meeting and any subsequent outcomes from it would lead to better AI safety, that person has been very much left wanting more.

Samsung Brings a Bread Knife to a Gun Fight
In last week's edition, I had this paragraph at the end of my story about Googlebooks:

This presents an interesting position for Samsung; they are not releasing a Googlebook at this time and all indications are that the release of the Galaxy Tab S12 Ultra will be only a few days after the first Googlebooks ship to those who pre-ordered them. Samsung surely wants to continue selling their tablets; they have put a lot of time and money into making them compelling devices with features and capabilities that few other Android tablets share. Will Samsung ever make a Googlebook? And if so, what would it mean for the S line of tablets? In the meantime, I've put my plans to upgrade to a Tab S12 Ultra on hold and have pre-ordered the Acer Googlebook convertible which has USI 2.0 pen support and might perhaps be a decent replacement for my Tab S11 Ultra. Samsung can't really afford for many other people to do the same, so they will quickly have to determine their position and placement in this new Android world.

The Tab S12 Ultra was officially announced this week and select reviewers and influencers who have had early access to the tablet have been releasing their YouTube videos on the differences from last year's Tab S11 Ultra. Much as expected, the tablets are essentially the same except for a newer chip in this year's tablet. In one respect this is a good thing because it means anyone who bought the $400+ Pro Keyboard for their Tab S11 Ultra can upgrade and know that accessory will work fine with the new tablet. But in another respect, it means the Tab S12 Ultra now has to compete with Googlebooks and its pricing makes this somewhat challenging. Although one of the key differentiators of the Tab S12 Ultra is that it is a tablet and can be used full-time as a thin and relatively light tablet, once you add on the keyboard accessory, the value proposition dips quite a bit. The Tab S12 Ultra starts at $1399 and with any Samsung keyboard accessory, it will be more expensive than any of the Googlebooks currently available for pre-order. For anyone who planned to use their giant tablet in a keyboard case more often than not, the comparison against a Googlebook is a little rough. You give up the ability to separate the tablet from its keyboard, but gain better keyboards and trackpads, get more RAM in the base models (16GB vs. 12GB in the S12 Ultra), more storage in the base models (512GB vs. 256GB in the Tab S12 Ultra), the ability to have a Linux environment for development and running Linux apps, an extra 3 years of platform support, and all of the unique Googlebook features such as Continue On, Screen Sharing, and Magic Pointer - none of which are available on the Tab S12 Ultra. In my mind, one buys the Tab S12 Ultra over a Googlebook because either you know you absolutely want the tablet form factor and not a convertible, you don't want to wait for the tablet Googlebooks we believe are coming based on the leaks we've already seen, and/or you know you want Wacom EMR stylus support that none of the current crop of Googlebooks offer. Even the Acer Chromebook convertible offers only USI 2.0 stylus support, which is not as responsive and optimized for the needs of digital artists as Wacom EMR (or the Apple Pencil). For anyone who doesn't fall into one of those camps, I would have a hard time telling them to buy the Tab S12 Ultra. Samsung didn't show off any new compelling software features that set the Tab S12 Ultra apart from the Googlebooks. It is almost as if Samsung does not believe Googlebook will go anywhere and in the end those of us who believe we can find similar satisfaction in Googlebook as we have in Tabs Ultra of the past will all eventually come crawling back to their tablet line.