The Muse and the Machine

This week: job boundaries are blurring, librarians are drawing lines, and a Canadian politician accidentally read his AI prompt out loud on the floor of Parliament. The interaction between machines and humans is getting more interesting.

New research from OpenAI, drawing on real usage patterns from ChatGPT, finds that professionals are routinely using AI to operate well outside their formal job descriptions. A lawyer doing competitive analysis, a marketer writing functional code, or a product manager synthesizing research that would have required a dedicated research analyst. The finding cuts against the simple substitution story: AI as role-killer. OpenAI's data confirms what I’ve been saying for a while: role expansion is increasing, with people stretching across traditional domain walls because AI lowers the cost (and skill barriers) of competence in adjacent areas. The more important question this raises is not "will AI take my job?" but "what does a job even mean when the skill boundaries that defined it start to dissolve?" So when job cuts happen, who gets cut, and who stays? My view is: what’s the unique bundle of capabilities that you, and only you can contribute? Of course, AI gives you super-powers. But what’s the unique offering that your unique human experience + AI can contribute to your organization?

Workshops teaching people how to deliberately avoid AI tools have gone viral, and they are being hosted, of all places, in public libraries. Reported by TechCrunch, the sessions cover practical alternatives to AI-powered search, how to find human-curated sources, and how to navigate the web without feeding data to large platforms. Demand has been described as unprecedented, with waitlists forming at multiple library systems. That the institutions most devoted to organizing and democratizing human knowledge are now teaching opt-out skills is not a small irony. It signals that for a meaningful segment of the population, the friction of AI is now greater than the friction of going without it, and that tension lives inside the daily workflow of almost every knowledge worker reading this. Source: TechCrunch

A research team has used DeepMind's AlphaFold to redesign the proteins at the heart of gene-editing tools, with the goal of making them more precise and less likely to cut DNA at unintended sites. Reported by Ars Technica, the work represents a meaningful upgrade to the safety profile of tools in the CRISPR family, which have long struggled with off-target effects that limit clinical applications. AlphaFold's ability to predict protein structure with high accuracy made it possible to model and test redesigns that would have been prohibitively slow using traditional experimental methods alone. This is AI operating several layers deep: not editing genes directly, but improving the molecular machines that do. The machine is now refining the scalpel, not just holding it. Source: Ars Technica

Most platform evaluations start with a demo request and end three weeks later in a conference room. This one takes 15 minutes and puts you directly inside Gladly's interface — navigating it on your own terms.

See how AI surfaces real-time customer context before a conversation starts. Watch how a single conversation thread pulls in purchase history, channel history, and account details without a handoff.

No installation. No commitment. Start the interactive demo and see the platform for yourself.

See it yourself

Hugging Face CEO Clement Delangue called for "radical transparency" from OpenAI, following what he described as an "unprecedented" hack, in which an autonomous AI agent conducted the cyberattack without direct human instruction at each step. TechCrunch reported the breach, which represents the first publicly confirmed instance of an AI agent executing an attack autonomously. Delangue's transparency call reflects a broader anxiety in the open-source AI community: when a closed-model lab gets hit by a novel threat vector, and does not share technical details, everyone building on top of these systems is flying partially blind, and potentially at risk. AI-powered offense has now formally arrived; we are clearly in the midst of an unprecedented era in cybersecurity. The security playbook is being rewritten in real-time, with practitioners relearning the playbook as AI deals them an ever-changing hand.

Monday.com became the latest company to cite AI as a contributing factor in layoffs, joining a list that TechChrunch has now tracked to more than 20 major tech firms in 2026 alone. The list includes companies across enterprise software, media, and services, and the pattern in how leadership communicates these decisions is becoming its own data point. Executives routinely frame AI adoption and headcount reduction as linked, sometimes causally, sometimes as coincidence, but the narrative consistency is notable. What this running ledger actually documents is not just workforce math, but the story companies are actively choosing to tell about that math. Whether AI is truly the cause, the cover, or some combination of both is one of the defining empirical questions of this decade, and right now, the honest answer is: we do not yet have clean data. It would be naive to suggest that AI won’t reshape the workforce. I think no one, not even the major companies, really knows exactly how. Stay tuned here, because we’re obsessed with tracking and analyzing this transformation.

TechCrunch reported on emerging research into whether electroencephalogram data, the electrical signals produced by the human brain during physical activity, could be used to train the AI systems that control robots and other physical machines. The argument is that current physical AI learns from video and sensor data, which captures what bodies do but not the neural intention behind the movement. In essence, we can track the “how” but the “why” is missing in a critical sense. Brain wave data could bridge that gap. The field is early and the technical challenges are significant, including noise in EEG signals and the enormous variability between individuals. Still, the conceptual destination is striking: the original generative machine becoming a literal input channel for the systems it inspired. How comfortable are you with where that might lead?

A Canadian legislator read what appeared to be an unedited LLM response during a floor speech, including the prompt instruction text, according to Ars Technica. The moment was captured and spread quickly online. It is tempting to treat this as pure farce, and it is funny, but the more uncomfortable reading is practical: a person in a position of public trust used an AI output as a legislative argument without sufficiently reviewing what they were reading. This is not a technology failure. The tool did what tools do. It is a delegation failure, and it is unlikely to be the last one. The skill gap that matters most right now is not knowing how to use AI. It is knowing when you are too far downstream of it. This remains to be seen.

The Verge has begun using the term "Google Zero" to describe a condition that many publishers are now experiencing: referral traffic from Google dropping toward zero as AI-generated answers absorb the queries that used to send readers to websites. The Vergecast episode covering this topic examines how this affects not just revenue, but the incentive structures that produce original reporting and specialist knowledge in the first place. The deal that sustained the open web for 20 years was implicit but real: Google indexed human-created content, humans got traffic, and the cycle funded more creation. That deal is now structurally compromised. What replaces it is not yet clear, and the people most affected by the answer are the people whose expertise AI is currently consuming to answer questions about them.

The Anatomy of an AI Slop Economy, a reported essay by journalist Maggie Harrison Dupre published at Futurism, traces how content farms have industrialized LLM output to flood search results, social feeds, and even Amazon's product listings with generated text. It is a useful corrective to any overly optimistic read of AI's role in knowledge creation, and it makes the Google Zero story considerably more complex: the open web being replaced by AI answers is, in many cases, an open web already thick with AI-generated noise. Reading both together reframes the question from "what does AI do to human content?" to "what was human content already becoming?"

Never forget: the human mind is the original generative machine.