Signals and Field Notes #14
Learning and Innovation Observed....Gratias vobis ago quod legitis
Notion just turned its workspace into a hub for AI agents: Ever since B.J. Schone published this article last week, I’ve been paying a lot more attention to what Notion is up to. Points like this from a recent livestream shift Notion from a note taking app to something different:
“By building an orchestration layer — a system that coordinates AI work across multiple tools and data sources — Notion is positioning itself as more than a note-taker with AI features and instead as a hub where people and agents can collaborate across tools and databases.
Now Notion will allow teams to deploy their own custom code. With its new Workers, Notion’s cloud-based environment for running custom code, customers can write their logic and deploy it to a secure sandbox (i.e., an isolated environment that keeps the code from interfering with other systems). This allows teams to do things like sync their data into Notion, build custom tools, and trigger work with webhooks — which are automated signals that kick off actions when something happens in another app — without needing to rely on external infrastructure.
Another addition allows Notion’s users to chat directly with external AI agents they use, assign them work, and track their progress, as if they were one of Notion’s own custom agents. At launch, Notion says that Claude Code, Cursor, Codex, and Decagon are supported partner agents, but it plans to add more.”
In an era when workers are returning to offices, here’s how Dropbox is making remote jobs work: I love the work that Dropbox has put into being a virtual-first organization. It really shows that you can do things differently but not-so-ironically you actually have to put in conscious work to be different successfully. I respect the real work they’ve done here and that they’ve made their virtual-first toolkit openly available. And if you’re not following Allison Vendt, you should be.
AI pricing models are evolving—and it might cost companies more: Golly gee, file this under “Well duh” > > “It’s a move that should have been expected, said Harness Field CTO Adam Arellano. Old AI pricing models, including all-you-can-prompt for a set monthly price, were unsustainable, and that pain was going to be felt sooner or later. OpenAI already introduced token-based pricing for its popular ChatGPT model.”
Frontier AI models don’t just delete document content — they rewrite it, and the errors are nearly impossible to catch: “A new study by researchers at Microsoft shows that large language models silently corrupt documents that they work on by introducing errors. The researchers developed a benchmark that simulates multi-step autonomous workflows across 52 professional domains, using a method that automatically measures how much content degrades over time.” Kind of feel like this dealing with the symptom and not the cause > > “Interestingly, giving models an agentic harness with generic tools for code execution and file read/write access actually worsened their performance, adding an average of 6% more degradation. Laban explained that the failure lies in relying on generic tools rather than domain-specific ones. “Models lack the capability to write effective programs on the fly that can manipulate files across diverse domains without mistakes,” he noted. “When they cannot do something programmatically, they resort to reading and rewriting entire files, which is less efficient and more error prone.” The solution for developers is to build tightly scoped tools (such as specific functions to calculate or move entries within .ledger files) to keep agents on track.”
Culture is infrastructure—and Stockholm is betting on it: This is largely about music but I can’t love the headline enough > > “City planners are increasingly aware that culture needs the right infrastructure to grow. An index compiled by AEA Consulting counted 267 cultural infrastructure projects announced in 2025, representing $13.6 billion of planned investment. That is the highest number of announced projects in the last decade. An institutionally established cultural sector allows a lively grassroots scene to flourish, supported by the right policies.” > > I think org change management plans need to think about this - that the org culture represents an infrastructure element. Framing it that way places culture in the foundational position it deserves.
Amazon workers are under pressure to up their AI usage—so they’re making up extraneous tasks: I love when I get to use the same meme twice in a single issue > > “As detailed in a new report by the Financial Times, Amazon employees are reportedly using the company’s new internal AI tool, MeshClaw, to create extraneous AI agents—not to increase productivity, but to drive up AI activity.”
Anthropic courts a new kind of customer: small business owners: This makes so much sense in terms of businesses that could truly get new capabilities or efficiencies that they just can’t get any other way (e.g. they can’t hire enough headcount to do it) > > “Anthropic is looking to court smaller companies. To that end, the company announced Wednesday the launch of Claude for Small Business, a new suite of services designed for customers who less resemble Walmart and Starbucks and more resemble the local hardware store or coffee shop…The new suite also includes integrations between Claude Cowork and a number of software products — like QuickBooks, Canva, Docusign, HubSpot, and PayPal.” See also > > Anthropic now has more business customers than OpenAI, according to Ramp data.
What I Learned about Connection and Trust from Black Farmers and the CIA: My friend Eugene Kim does what he does 2nd best which is write about his experiences finding commonality among disparate people (the thing he does best is making those connections happen IRL). I’m super happy that I got to play the smallest part in this journey > > “With whom we do things matter. Making positive change in the world is fundamentally hard, but both NuLeadership and Intellipedia reveal a path that can lead to success. It starts with finding your people. When we find each other and lift each other up, the whole becomes greater than the sum of its parts. I was struck by the role that community continues to play for both of these projects, and I feel blessed to have been able to observe and experience this firsthand.”
SAP Launches Unified AI, Automation Suite (gift link): Two things - another giant moves in this space - this, like SFDC, will have ripple effects. Another thing, the data that SAP sits on will be its defensible moat - that includes the insight it has into multiple processes and workflows across thousands of clients.
Deployment is L&D’s New Moat [2026 update on the FDLD model]: Another friend, this time Marc Ramos, writing smart stuff. He’s got an important point around how we think of “forward deployed” folks but read on down to find the skills portion of his article. My thinking is that he’s spot on about the skills and we also need to look at those skills within an ecosystem. How do we change undergrad and grad curriculum to build the foundations for those skills to be sharpened in the corporate space. How then do we changes systems of hiring and rating and assessment to hire for, value, and reward those skills? So Marc charts a course that hopefully, shows L&D a way not to just survive this new era but thrive.
Deal reached with hackers to delete data stolen from the Canvas educational platform: …and in related news…recruiting for CISOs among learning companies jumps (probably) > > “The company that operates online learning system Canvas said it struck a deal with hackers to delete the data they pilfered in a cyberattack that created chaos for students, many of them in the middle of finals. Instructure, the parent company of Canvas, said in an online post that it “reached an agreement with the unauthorized actor involved in this incident.”
All Those A.I. Note Takers? They’re Making Lawyers Very Nervous (gift link): Classic 2nd order effect - see also unintended consequences. “Executives and corporate boards generally expect conversations with their legal team about legal matters to have attorney-client privilege. They lose that protection if they share the same information with outside parties — and it’s possible that an A.I. note taker could have the same effect. Companies that make such tools, the argument goes, may have access to the transcripts and data related to them. While courts have not directly addressed the issue, they’ve considered parallel questions.”
The enterprise risk nobody is modeling: AI is replacing the very experts it needs to learn from: This is the key > > “There is a critical difference between a field being automated and a field being understood. We can automate a huge amount of structural engineering today, but the abstract knowledge of why certain approaches work lives in the heads of people who spent years doing it wrong first. If you eliminate the practice, you don’t just lose the practitioners. You lose the capacity to know what you’ve lost.” > > I 100% believe that the orgs that will not only survive but will be thriving in 24-36 months will be those that understand this - that the way to best deploy AI (specifically agentic) in the enterprise space, is as a mechanism to grow the capabilities of the org and not as excuse to cut labor costs. > > Here’s another aspect to this “‘How else are people going to learn to do the job’: MIT AI expert warns against automating Gen Z entry-level jobs” From the article “MIT researcher Andrew McAfee has warned companies could risk damaging their future workforce if they replace entry-level workers with AI today. By filling some of the lowest pay grades with AI, Gen Z workers will not only be left without jobs, but they won’t have the experience to be able to progress into higher paying jobs, ultimately causing a major disruption across the talent pipeline.” Are short term bumps in lower labor costs (which might not even be recognized because of severance costs), worth it to hamstring your org in the long run?
ArXiv to Ban Researchers for a Year if They Submit AI Slop: This won’t stop it but its a move in the right direction > > “ArXiv, the open-access repository of preprint academic research, will ban authors of papers for a year if they submit obviously AI-generated work.”
Claude’s next enterprise battle is not models: it’s the agent control plane: “For the last two years, the enterprise AI race has mostly been framed as a model war…but the next strategic fight may not be over which model answers a prompt best. It may be over who controls the layer where agents plan, call tools, access data, run workflows and prove to security teams that they did not do anything they were not supposed to do.” > > This has been a strong signal getting stronger for a while now, the next interesting question for me is, what’s the battle after that look like? Where it be fought? Maybe here > > Runway started by helping filmmakers — now it wants to beat Google at AI > > “Runway co-founder and co-CEO Anastasis Germanidis said training models directly on observational data from the world is the next frontier of AI. The companies that get there first, he argues, won’t be the ones that perfected language.”
The World’s First 240Hz Video Smart Glasses for Gaming Aren’t Cheap: What two industries have traditionally led in the deployment of new tech? That’s right, porn and gaming. Look at these stats > > “You get the same 171-inch (1,920 x 1,080) virtual display through the micro OLED panel, 57-degree field of view, Bose-tuned sound, 3DoF (three degrees of freedom) tracking technology used for anchoring virtual windows, and more.” Now I have to say that if I’m going to be in an open office environment, a rig like this would be so helpful. I don’t want to walk around with them on and I don’t want to be out and around others with them on but wowza could they make an office environment amazing.
OpenAI is reportedly launching a phone for ChatGPT: Just don’t. Here’s the thing about phones - its not just a piece of tech - its personal. Its with you almost always. It takes pictures of trips and loved ones, it has all your banking info, it has your silly games - but what it really is, is a box of apps curated by you. Now one of those apps might be ChatGPT but it might not be. I can break up with ATT whenever I want and take my phone and its apps to another carrier…what happens when I want to break up with OpenAI. Hire an anthropologist. A phone is much more personal than a laptop. Quit trying to make fetch happen.
Microsoft’s new research finds an AI ‘paradox’ holding companies back: Is anyone keeping count of how many times this will have to be said (and then it’ll still be ignored)? “A new Microsoft study of 20,000 artificial intelligence users in workplaces around the world concludes that the biggest barrier to getting real value from AI isn’t the technology or the workers themselves — it’s the ingrained culture of the organizations where they work. That “Transformation Paradox” is one of the central findings from Microsoft’s annual Work Trend Index, released Tuesday morning, which paints a picture of employees eager to reshape their jobs and organizations that aren’t really in a position to make it happen.” I mean ‘duh’ - doesn’t really even cut it anymore. Companies are so eager to bring in the new tech and so unwilling to do the actual work of re-factoring how their orgs are put together and how work and jobs are organized.