Signals and Field Notes #16
Learning and Innovation Observed....Gratias vobis ago quod legitis
Full disclosure - I’m not Catholic but I think its rather notable when a leader of one of the world’s largest religions is quoting one of the greatest fantasy authors of all time to talk to us about how we spend our time and what we need to think about fighting for. The whole Encyclical is here, and is worth a read.
See also: “A society grows great when old men plant trees in whose shade they shall never sit.”
Real enterprise transformation with AI requires six foundations, not one. Here’s how to build them all: I think this is right and the coordination of these efforts will be the real challenge and I mean like the longest mile, the tallest pole in the tent, whatever analogy you prefer - herding the cats of this transformation will be the cover charge for those orgs that want access to that concentrated winning > > “One of the results of this concentration has been that the businesses that failed to harness the new technology have been left in the dust by those that succeeded. We can expect to see the same pattern emerge with AI—not a broad-based transformation but a concentrated one in which the companies that adapt quickly will grow their market share while those that do not will fade into irrelevance.”
AI Agents Plunged the Tech World Into Chaos. Here’s Exactly How That Happened: One thing, I don’t know why we don’t get the chance to work in home offices or co-work spaces > > “Countless coders spent the holidays in basements and dens, madly trying out this new toy that let them build software as if they’d unleashed a hundred clones.” I will say this, this kind of reporting, almost at the moment of it happening will be critical to building a deeper history of this time further into the future. “Journalism is the first, rough draft of history” seems an appropriate sentiment from Washington Post publisher Philip L. Graham.
To Land a Job in AI, Try Reading Kant: This is cute but won’t be true at scale - although the world would be a better place if it was. I’d also add sociologists, anthropologists, and historians. Anything that helps tech bros realize that the humanities exist. Speaking about anthropology, this is a timely read: In Praise of Human Knowing > > “Before frontier labs and machines that promised to compute intelligence as a single ascending quantity, there was a discipline that walked to the far shore of the human and came back with the news that there was no single human to be found. It is the one field of human knowing that took, as its founding humility, the possibility that we do not yet know what a person is — and that we might learn, if only we were willing to sit still on someone else’s ground long enough to be made strange to ourselves. This humility is worth praising now at a moment that has grown impatient with the question. I am referring to anthropology.”
Microsoft’s quiet Claude Code retreat and the real cost of enterprise AI: Big signal on the business side “The Claude pullback is the most credible signal yet that the unit economics of enterprise AI coding do not, at current token prices, work. Not because the tools are bad. The opposite: they are good enough that engineers use them constantly, and the constant use is what breaks the maths.”
Exclusive: Coursera, Udemy complete merger to build AI-era skills giant: I think this is the right move for these companies but I think the importance of skills will continue to face a challenge or bump up against the limits to which orgs change the nature of work. “Coursera and Udemy have completed their merger, creating a massive online learning platform built for workers and employers, just as AI changes the skills needed for nearly every job. Why it matters: Coursera says someone has enrolled in a generative AI course every three seconds, on average, so far in 2026 — up from every four seconds in 2025.”
When enterprise AI finally works, it won’t look like AI: Keep beating this drum > > “AI use is broad, but most organizations still have not embedded it deeply enough into workflows and processes to create material enterprise-level benefits. It also finds that workflow redesign is one of the strongest contributors to meaningful business impact.”
Claude’s Human Resources plugins: In case you missed it, Claude dropped plugins for HR including for - Performance Reviews, Compensation Analysis, Onboarding, and Policy Guidance. I mean, that’s all, no worries.
Why the ‘Middle Path’ of AI Literacy May Be the Future of English Class: I love this approach, especially at the 10th and 11th grade levels. You invite AI in but also include a rigorous examination of its output especially compared to human output. > > “As we read, we examined how large language models’ recycled novel “analyses” mis-read and oversimplified complex literature, producing distillations that often lacked nuance compared with the creative, discursive yet defensible readings that the students themselves generated. They learned to discern actual analysis from simplistic summaries, and to suspect the allure of AI’s instant “correct answers.”
Why Confidential AI is the next big thing for enterprise: “Enterprise AI faces a trust problem that better models alone will not solve. Once AI systems begin handling source code, customer records, internal documents or regulated business logic, the question is no longer just whether the model performs well. Security teams and auditors want to know where inference ran, who could access data while it was in use and what evidence remains after the fact.” > > I firmly believe its goign to get much, much, weirder than this. Wait till you have to negotiate with new hires who will want to bring their digital twin with them. Then you’ll have to be able to negotiate the value of their digital twin vs someone else’s and you’ll have to figure out how to pull apart the proprietary data that their digital twin learns while they are working for you so that they can legally go to another company. Can you imagine trying to enforce NDAs or non-competes against AIs? I can, but then again, I read a lot of science fiction. I wonder if this might play some role in that future > > “OpenAI launches Privacy Filter, an open source, on-device data sanitization model that removes personal information from enterprise datasets.”
Why high-growth companies should build decision cultures: Not disagreeing but GOOD LORD we’ve been saying this since the dawn of the commercial internet.
The Teamwork Graph Was Atlassian’s Secret Weapon. Now It’s Sharing It.: Great stuff as usual from Ken Yeung. I think this is a wicked smart move by Atlassian. It gets to the heart of what an org’s true moat is now - a longitudinal data set > > “The Teamwork Graph is not a simple index. It spans more than 150 billion objects and relationships—not just conversations and people, but active projects, open issues, Wiki pages, and external assets Atlassian ingests from across the enterprise: design files, code repositories, Google Drive, SharePoint, and more. Every time a customer links to an outside tool, Atlassian pulls that object into the graph, building a continuously expanding map of how work actually happens, and one that agents can now query directly.” > > This is also what I don’t understand about LinkedIn. They set on insane data sets and seem to be doing almost nothing with them. Can you imagine the skills data they have? The social graph of much of the professional world? Its crazy to think what could happen if they licensed access to these data sets.
Google just bought a stake in the maker of Eve Online to train its AI models: If you’ve never played or heard of EVE Online, that’s understandable. Its also known as “spreadsheets in space.” What distinguishes EVE is how long people will plan, strategize and work to make certain plays win. Notable events include “The Bloodbath of B-R5RB (2014)…over 7,500 players and dozens of alliances fought a 21-hour battle. It resulted in the destruction of 75 Titan-class ships, costing players an estimated $300,000 to $330,000 in real-world equivalent” and scams like “The EVE Intergalactic Bank (2006) in which a player founded the EVE Intergalactic Bank (EIB), operated it as a legitimate bank for months, taking deposits and paying out interest and then he drained the accounts of an estimated 790 billion ISK (in-game currency) and published a video confession. At the time, this equated to around $170,000 in real-world value. This is recognized by Guinness World Records as the Largest Virtual Theft in an MMORPG” > > These scams and battles take years to plan and pull off and involve hundreds or thousands of people. This is goign to be interesting to see what Google’s AI learns from this data set.
I Am Begging AI Companies to Stop Naming Features After Human Processes: Yes! They don’t “dream” or “hallucinate” - also storms aren’t angry ever. And please stop naming stuff after Lord of the Rings characters - you totally misread the book. See? Stop this > > Anthropic will let its managed agents dream.
Rethinking operating models for humans with agents: Thanks to David Mallon, et al, for being part of this message - I wish there was some survey looking at the adoption of plans to reengineer work…oh wait…look! Data! > > “Early adopters are finding that bolting autonomous agents onto operating models designed for human workers is like fitting a jet engine to a bicycle. According to Deloitte’s Tech Trends 2026, many companies are automating existing processes designed for human workers without rethinking how work should be done. Deloitte’s State of AI in the Enterprise 2026 survey finds that 84% of companies haven’t redesigned jobs to fit AI, even though automation expectations are high. The main obstacle, according to the executives surveyed, is a lack of worker skills, yet less than half of the survey respondents report that their organizations are changing their talent strategies thus far.” > > Its like buying a race car and not paving the roads.
How To Future-Proof Your Career In The Age Of AI: This has me feeling good for the Humanities > > “The most prized future workers will be those who can decode a sea of outputs, spot the meaningful signal and translate it into action that others understand and trust. They will be adept at cross-domain reasoning: seeing how a precedent in one field might illuminate a policy in another, how a consumer insight from marketing could recalibrate a product’s technical architecture, or how an ethical concern might redirect a technical roadmap. They will be comfortable with ambiguity and rapid iteration, able to move from hypothesis to test to refinement with confidence and humility. They will be capable of designing decision processes rather than merely executing them.”
The Quest to Build a Better AI Tutor: Good lord - everything old is truly new again…personal tutors help. > > “But there was one key difference. Half the students were randomly assigned to a fixed sequence of practice problems, progressing from easy to hard. The other half received a personalized sequence with the AI tutor continuously adjusting the difficulty of each problem based on how the student was performing and interacting with the chatbot. The researchers found that students in the personalized group did better on a final exam than students in the fixed problem group. The difference was characterized as the equivalent of 6 to 9 months of additional schooling, an eye-catching claim for an after-school online course that lasted only five months.”
From Chaos to Innovation: Understanding Products and People in a Non-Deterministic World: HUGE shout out to EPIC and these talented researchers, for bringing research like this to the fore. It will be absolutely critical in finding ways forward. > > “In an era tempted by rapid in-market iteration, this paper demonstrates the critical role of ethnographic methods for understanding complex human interactions with non-deterministic LLMs. Through a longitudinal Wizard-of-Oz study of “Nova,” a simulated AI family wellness assistant, we exposed limitations of traditional usability methods in high-stakes, multi-participant contexts. Our methodological approach documented organizational chaos in group AI interactions, identified optimal patterns in human-AI and Human-in-the-Loop (HITL) collaborations, and traced the nuanced process of AI relationship formation and its impacts on user reflection and behavior. Ethnographic insights led directly to innovations including dynamic topic segmentation technology and multiple patent applications. This work demonstrates the indispensability of ethnographic methods for understanding AI systems within authentic social contexts, where human expertise and support remain vital.”