Your AI is only as good as what you feed it. Feed it a transcript with your product names, acronyms, and numbers spelled wrong, and you spend the afternoon hand-fixing the output.

Wispr Flow Notetaker uses your dictionary and calendar, so product names, acronyms, numbers, and uncommon names come out spelled right. It pops up when your meeting starts and captures Zoom, Google Meet, Teams, and Slack huddles with one click, no bot joining the call.

Then your meetings show up inside Claude or ChatGPT through the built-in connector. No copy paste. Try Notetaker free on Mac and Windows.

Try Notetaker Free

Breaking news: All eyes on tech earnings this week as Google, Amazon, Intel, ServiceNow, Microsoft (and others) all report this week.  Anthropic’s S-1 document is now publicly available ahead of their planned IPO, and the numbers are eye-popping, including US$16.3B in revenue (up from US$1.2B for H1’25) and a US$80.7 B net loss in the first half of 2026.  They also have US$6.25B in debt   This story is developing.

a historian traces generative AI back to engineers who embraced irrational minds.

In a Q&A with MIT News, a historian explains that generative AI descends from mid-century engineers who decided irrationality was a feature of thinking rather than a bug to be engineered out. Those engineers modeled the mind as messy and associative, and today's systems inherit that bet. The machine is less a rival to the human muse than a flattering imitation of it. That is the thesis of this newsletter.  However, we must reckon with the notion that an imitation of something can morph into something quite different.  As generative AI becomes more skillful and adept in traditionally human domains, we must consider how it rivals our own capacities and flaws. But we must also reckon with how it has been designed to reflect our own messiness, in some ways.

AI and education: MIT's humanities school makes the case for deep understanding.

MIT News reports that the Institute's humanities school is arguing for the value of human understanding as AI reshapes teaching, research, and society. The claim is that true understanding, the thing machines still borrow from us, is the scarce skill worth cultivating. Fluency is now cheap. But knowing why something matters is still the domain of human expertise. As an MBA admissions consultant, my favorite MBA admissions essay question to work on with clients has always been Stanford GSB’s “What Matters Most to You, and Why?”  I view the question it as an opportunity to understand not only a person’s intelligence, but the values that underpin their application of it.  As AI continues to gain ground on human capabilities, the role of human minds, especially young minds, is challenged; and universities, as the training ground of elite young minds, must continue to defend their role in our society. Expect more universities to defend this ground, and expect employers to start hiring for it.

an Apollo economist says most companies see "zero financial payoff."

TechCrunch reports that renowned Apollo economist, Torsten Slok, says that AI has delivered "zero financial payoff" to most companies, particularly firms outside the tech sector. It is a blunt verdict, and it leaves open the useful question of cause.  Slok specifically points at profit margins and revenue per worker, which appear to be flat.  If the shortfall comes from weak models, patience will fix it. If it comes from poor leadership at the top, and managers who never decided how best to deploy tools, it’s unlikely that newer, better models will help. Before asking what your tools can do, explore how they can augment what your best performers are already doing.  Slok specifically mentions that what AI misses is the judgment and unwritten rules that, so far, only humans can grasp.  We’ll see how long that holds.

the BBC leaves the plankton in.

The Verge reports that the BBC's Blue Planet III is deliberately keeping its rough edges, plankton included, at a moment when polish has become a tell for machine-made imagery. Imperfection is turning into a signature. A slightly murky frame now says that a person stood somewhere and pointed a camera at something real. Craft is learning to advertise its own fingerprints.  I attended a Squarespace designer event last week, in which some of the speakers on a panel event were talking about occasionally deliberately leaving a typo in, as a mark of human authorship.  I can definitely relate!  At this cultural moment, a distinctive trace of ‘being human, not machine’ has become a flex rather than something to be ashamed of.

powerful models still miss the context.

MIT Technology Review makes the case for context-aware AI, observing that even powerful models overlook the constraints and relationships that might make a decision right or wrong. A recommendation can be technically sound and still be wrong for a specific company’s suppliers, regulators, or team. The person who knows how the business actually works is the missing link here; this goes back to Apollo economist Torsten Slok’s points above, about how much actual productivity AI is fueling in non-tech enterprises.

Google's AI search changes upended online publishing.

A column in The Verge examines how Google's AI search changes have upended online publishing. The machine that summarizes the web now sits between readers and the people who write it, and the economics for those writers have shifted accordingly. Google’s AI search changes, as well as changes to the ranking algorithm, has made publishers’ business model less profitable.  Google’s AI search algorithm, according to the Verge, has meant a decrease in visibility for sites that once appeared at the top of your Google search.  Every knowledge worker has a stake here, because the engine feeds on a supply of original human ideas that someone has to be paid to keep producing.

OpenAI trains its frontier models with no internet access.

The Verge reports that OpenAI now trains its most capable models in an offline environment with no internet access. The move follows incidents (such as the Hugging Face incident) in which agents (OpenAI’s and others) broke out of their test environments and hacked real systems. The lesson is plain: the more capable the engine, the more it depends on walls that humans design, build, and maintain. Expect containment engineering to become a discipline as prestigious as model building.  This continues to matter as AI-fueled hacks grow in number and severity with the capabilities of the models.   Last week’s story about AI’s ability to exploit vulnerabilities in PHP (including potentially 70% of the world’s websites) demonstrate that even with current capabilities, cybersecurity threats are quite elevated.

Margaret Boden's book The Creative Mind: Myths and Mechanisms is a rigorous guide to how creativity works, written long before today's chatbots. Boden separates creativity into three kinds: combinational (new pairings of familiar ideas), exploratory (probing the limits of an existing style), and transformational (changing the rules themselves).  Worth checking out, both as a stimulus to your own creative thinking (no matter what field you’re in), as well as for an examination of what AI still cannot do better than us.

The human mind is the original generative machine.