Prodity: Product Thinking

Move Slow and Make Things, Permanent Underclass, and Halloween

Kyle Evans — Sun, 25 Oct 2026 14:15:36 GMT

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

⚒️ Building for Yourself
🛜 Neural Networks
🥸 AI Economic Flaw
✏️ UX
🤖 Permanent Underclass
🧱 Move Slow and Make Things
🎃 Halloween

Podcast

Build It for Yourself First: Todd Katcher on Internal Tools, AI Reality Checks, and Why the Best Product Doesn’t Win

In this episode, Kyle and Todd talk about how one conversation can change a company’s direction, and how Todd’s team builds its own tools when off-the-shelf software covers only 80% of what they need. That approach produced Zoney, Convert It, PriceAlerts, and an e-signature platform. Todd explains how he decides what to build, what to keep internal, and what to take to market, including why building their own CRM didn’t make the cut. They also take an honest look at AI: where it saves real time, where it still needs a critical human eye, and why managing expectations matters. The conversation closes on Todd’s biggest lesson: making the product is rarely the hard part. Getting people to buy it is.

News and Useful Reads

“We were looking at it wrong”: The explosive claim that’s upended the AI world

Maybe understanding AI neural networks isn’t as complicated as we’ve made it seem. While still speculative, it seems like we may just need to look at them differently.

Cammarata’s claim is that LLMs might in fact be far easier to understand—and thus to monitor and control—than people previously thought. If true (and we’ll come back to that), it’s a huge deal.

Here’s the ‘core economic flaw’ in the AI bull case as stock and bond trades unravel, according to a top economist

“The core economic flaw in the AI bull case is that the marginal cost of producing compute exceeds the marginal benefit it delivers to users, raising questions about claims of unlimited demand for data centers,” he explained. “How much compute and how many data centers are needed depends on the returns firms are able to generate with that compute. So far, the AI payoff for firms has been slow to materialize, implying a growing reluctance to pay for AI services.”

AI in UX: Work Is Faster, But Practitioners Have Mixed Feelings

Like most of us, UX practitioners have mixed feelings about AI. It can be useful and speed up parts of the process, but it isn’t a silver bullet.

We surveyed 521 UX practitioners about how they use AI at work and how it’s affecting them. The short answer: AI use is widespread and saves time, but practitioners are far from enthusiastic, and the benefits depend heavily on role, task, and workplace.

The idea of a “permanent underclass” started as a meme. It’s tapping real anxieties in Silicon Valley.

The billionaires and heads of AI companies are promising we may not need to work anymore because of AI. But if the history of technology offers any lessons, it’s that the benefits of AI and increased productivity won’t go to most of us. And many are feeling that anxiety.

AI could create a “permanent underclass,” a subcategory of (sub)humans that may very likely include you and me and everybody else we know if we do not entrench ourselves against the impending AI superintelligence, which could very well reduce the worth of all human labor to zero.

Move slow and make things: What Japanese craftsmanship can teach business about quality, patience, and expertise when AI makes speed cheap

This article speaks to my soul. As someone who does woodworking (along with other similar hobbies), I deeply appreciate that it takes time to become an expert and to make things that are worthwhile. And that is something it is easy to overlook in our age of technology and AI.

For thirty years, the dominant logic of Western business has been simple: move fast and break things. The first mover has the advantage. Scale before you’re ready. Ship before it’s perfect. Ask forgiveness later. This thinking has produced extraordinary wealth and genuine innovation. It has also created a generation of companies that mistake velocity for value, and are only belatedly discovering that the two are not the same thing.

Speed is now free, which means, for the first time in history, speed is worthless.

Other Interesting Finds

The Origins of Halloween Traditions

The Halloween holiday has its roots in the ancient Celtic festival of Samhain (a Gaelic word pronounced “SAH-win”), a pagan religious celebration to welcome the harvest at the end of summer, when people would light bonfires and wear costumes to ward off ghosts. In the eighth century, Pope Gregory III designated November 1 as a time to honor saints. Soon after, All Saints Day came to incorporate some of the traditions of Samhain. The evening before All Saints Day was known as All Hallows Eve, and later, Halloween. Here is a look at the origins of some of the classic Halloween traditions we know today.

Subscribe now

Dead Startups, Synthetic Users, AI Agents, Sunsetting Products, and Halloween Candy

Kyle Evans — Sun, 18 Oct 2026 14:25:35 GMT

This week, we’re focused on the signal vs. noise, especially with AI. It is easy to get general direction from AI, but it lacks the nuance or edge cases that we really need to understand. And while companies continue to push the narrative of AI transformation, very few have changed. And even when the AI agents follow every step, they can still miss the mark.

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

📊 Product-Market Fit
😵 Dead Startups
🥸 Synthetic Users
🔃 Company Transformations
🤖 AI Agents
🌇 Sunsetting Products
🍫 Halloween Candy

Podcast

Finding Product-Market Fit with Ken Gavranovic

Discover how seasoned technology executive Ken Gavranovic shares insights on building innovative products, leading digital transformation, and harnessing AI to drive business growth. Whether you’re an aspiring entrepreneur or a product leader, learn practical strategies to navigate and excel in the tech landscape.

News and Useful Reads

AI labs are buying data from dead startups to train their models

The most valuable part of a failed startup may no longer be the office furniture, but the data.

Demand for defunct companies’ data emerged suddenly last year, as AI training labs hit the limits of public and synthetic datasets. Surge AI, for example, buys data to build “reinforcement-learning gyms” where clients’ AI agents practice tasks—think debugging a complex IT architecture or drafting appellate briefs.

Synthetic Users vs. Human Respondents: Same Conclusion, Less Variation

I find this article and outcome really interesting: getting feedback from AI synthetic users can point us in the right direction, similar to human feedback. But AI has less variation and captures fewer of the edge cases. Not surprising, but interesting.

Data from AI-generated respondents pointed to the same conclusion as a survey of 604 humans but contained far fewer distinct answers.

‘Close to zero’: AI chiefs reckon with ‘politics projects’ and the level of true change in response to AI

We discuss this idea frequently here, but very few companies are truly changing because of AI so far. The type of change that everyone imagines takes more time and significantly more concerted effort.

Few large organizations have rebuilt their core operations around artificial intelligence, according to the technology chief of one of the world’s largest consulting firms. He also said companies shouldn’t yet spend much energy on assigning blame when AI projects go wrong.

Your AI agents can execute every tool call correctly and still break every enterprise rule

The prevailing market narrative rests on a simple formula. Agents, plus tools, plus data are enough. While no vendor explicitly states it, nearly every product pitch assumes it. The idea is that you drop your corporate data into a cloud data lake, wire up an off-the-shelf large language model (LLM), hand it a toolbox of enterprise actions, and agility follows.
It won’t. It does the opposite.

Managers Need to Get Better at Killing Digital Products

Sunsetting products is a difficult thing to do for most companies and product managers. Having killed off several products in my time, I can tell you it often takes more work to stop work on something than it does to get started on something.

companies need to retire products earlier and with more deliberation. Those that don’t will accumulate a graveyard of half-supported tools, funded out of the very budget that should back what comes next, along with a customer base frustrated at the brand and less willing to engage next time.

Other Interesting Finds

How Long Do You Need to Trick or Treat to Burn Off That Candy Bar?

I love Halloween and Halloween candy. So this article intrigued and also disappointed me. And now I’m sharing it with you so you can also think about it when you’re stealing some Halloween candy from your kids’ bags. In summary, your favorite candies will cost you:

Reese’s Pumpkin: 42 minutes
Candy corn (19 pieces): 35 minutes
Kit Kat (fun size): 18 minutes
3 Musketeers (fun size): 16 minutes

Subscribe now

Good Strategy/Bad Strategy

Kyle Evans — Thu, 15 Oct 2026 16:55:52 GMT

Years ago, I worked with an executive team at a small organization. One of the first things we discussed was the vision and strategy of the organization. The CEO articulated to me that the vision of the company was to “grow to be a $50 million ARR company.” And the strategy was to expand its two primary divisions by 20% each year.

My immediate thought was, “This isn’t a vision or a strategy.” And it wasn’t. Or at least it was a very poor vision and strategy.

Many of us have probably experienced something similar to my experience above. In fact, I’ve seen this at many companies over the years. While leaders will often spend days or weeks creating a “strategy” for their organization, they’ll completely miss the mark.

So what makes for a good strategy or a bad strategy? That, of course, is the topic of the book Good Strategy/Bad Strategy by Richard Rumelt.

In this book, Richard Rumelt argues that what most organizations call “strategy” is actually goal-setting or wish lists. Good strategy is actually rare, primarily because it identifies a critical obstacle and concentrates the effort on overcoming it.

The book is divided into three parts: what bad strategy looks like and why it’s so common, the structure of good strategy (the “kernel”) and its sources of power, and how to think like a strategist. It’s built on many cases from business, military history, and government, which helps to make the idea concrete rather than just theoretical.

So let’s dive in.

Key Takeaways

While there is a lot to take in from this book, I want to focus on a couple of key ideas that I took away from this book.

A Strategy Includes The Kernel

Every good strategy has three parts, referred to in the book as “the kernel.” The kernel includes a diagnosis, a guiding policy, and coherent action. Most bad strategies skip at least one of these parts.

Superintelligence, The Day After, AI Paradox, Siri, and Loneliness

Kyle Evans — Sun, 11 Oct 2026 14:30:25 GMT

This week offered several technology and AI stories that paint a sharper picture of where we actually are. Sam Altman sat down with Fortune for a long interview on whether superintelligence can ever be controlled. A new AP poll this week found that 64% of Americans say AI is developing too fast, and 80% want the government to maintain human oversight. MIT Technology Review named the dynamic: AI’s “popularity paradox.” People who use AI describe it positively, while those same people fear it at the societal level. Against all that, HBR published research on what top AI performers actually do differently: they don’t just automate tasks, they redesign workflows and use AI to sharpen judgment.

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

📊 Product-Market Fit
🧠 Superintelligence
🗓️ The Day After
🌓 AI Paradox
🛑 Too Fast
🤖 Siri
🥇 Top Performers
🧑‍🦲 Loneliness

Podcast

Finding Product-Market Fit with Ken Gavranovic

News and Useful Reads

Can superintelligence ever be controlled? Sam Altman on AI safety, doomsday scenarios, and OpenAI’s delayed IPO

If you’re still struggling (like me) with the dichotomy of AI being the technology that could lead humanity to some sort of utopia while simultaneously being the technology that could destroy us, then you’re not alone. The leaders of AI companies are in the same boat, as this longread explores.

Unsurprisingly, Altman, 41, is a true believer in the society-changing potential of AI. Everything—materials science, biotech, energy—will be transformed, he asserts. He believes all diseases will eventually be cured thanks to AI. “I know this will be the most important work I ever touch,” he says.

Scoop: AI companies plot “the day after”

Not sure if this is comforting or extremely disconcerting.

Top executives at Anthropic, OpenAI and other AI companies are privately gaming out scenarios for a public and political revolt after a catastrophic AI event.

People really hate AI, so why can’t they get enough?

It’s a paradox we’ll probably be seeing more frequently: AI usage is increasing while skepticism of AI increases in tandem.

When people say they hate AI it’s not the technology they dislike but, rather, the relentless drive of the companies behind it to push the tech upon us in as many ways as possible. They’re telling us to brace for the biggest social and economic upheaval in generations.

64% of Americans say AI is developing too fast—and 8 in 10 want the government to keep it under human control

A new poll from The Associated Press-NORC Center for Public Affairs Research finds that most Americans, 64%, think AI is developing “too fast,” while 27% say its development has been “about right” and only 8% say it’s moving “too slow.”

The New Siri’s Biggest Strength Is Not Intelligence

This is the problem I’m currently grappling with in some of the products I work with: how to make them truly useful by adding personal context.

A useful AI assistant needs relevant personal context, the capability to use it, and an experience that helps users understand its limits.

What Top Performers Do Differently with AI—and Why They See the Biggest Benefits

Generative AI has produced two competing views about which workers will benefit most. One view is that AI will narrow performance gaps by giving lower performers templates, answers, and coaching directly in their workflow. The other is that AI will widen those gaps by eliminating routine but necessary work, which will allow top performers to spend more time on higher-value activities where they already have an advantage.

Other Interesting Finds

Loneliness may steal nearly 6 years of good health

A large study suggests loneliness and social isolation can cost people years of healthy life, especially when it comes to mental health. Men experiencing both lost about 5.8 years free from depression and anxiety, while women lost roughly 3.7 years.

Subscribe now

AI Adoption, Balls of Mud, Em Dashes, AI Safety, and Flags

Kyle Evans — Sun, 04 Oct 2026 14:31:11 GMT

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

📊 Product-Market Fit
☑️ AI Adoption
🟤 Ball of Mud
➖ Em Dashes
🤖 Verification
🦺 AI Safety
🇫🇷 Flags

Podcast

Finding Product-Market Fit with Ken Gavranovic

News and Useful Reads

How Leaders Talk About AI Predicts Adoption

Every company and every leader is telling a story about AI, whether we like it or not. But the story we tell greatly influences how others view AI and how well it gets adopted (or not).

The companies where leaders’ stories provide clarity, hope, and growth are seeing outsized performance on these measures. Not enough workers, however, are benefiting from these messages.

The New Big Ball of Mud: Why Agentic AI Systems Turn Fragile

With the rise of all the AI tools and capabilities, it’s very easy to rely on their outputs without appropriate consideration for the broader system we’re creating. But that just creates a big ball of mud.

In the AI era, a big ball of mud works surprisingly well — until accumulated complexity makes the system difficult to direct, diagnose, or reconstruct.

Claude Opus 5.5 uses em-dashes 99% less often and sounds more human — but still exhibits 2,548 AI writing tells

The em dash (a personal favorite of mine until it got usurped by AI) may be on its way out of AI writing. But that doesn’t mean AI sounds human yet.

But anyone hoping the model has shed its recognizable habits will find plenty left: the researchers identify 2,548 words, phrases and sentence patterns that appear at least twice as often in its generated articles as in comparable human articles.

AI Is Making Verification the Bottleneck for Companies

AI is a powerful execution tool. It is making it easy to build and create rapidly. But that means the work of verification is even more important.

When execution is cheap, verification becomes more valuable, and firms turn into verification factories: institutions capable of properly steering AI-generated output and standing behind the results.

Whatever AI Safety Is, It’s Not This

Companies and industries are notoriously bad at policing and regulating themselves. AI will be no different. We’re going to need strong oversight from governments, not from the companies building the AI.

The correct answer to all this uncertainty surely isn’t “I guess we’ll find out.” But that’s exactly what self-regulation encourages.

Other Interesting Finds

Why Are So Many Flags Red, White, and Blue?

So, due to simple practicalities, red, white, and blue became embedded in European heraldry over the centuries — not necessarily through conscious choice alone, but also due to cost and availability. By the time nations began designing flags, those colors were already well-established and strongly affiliated with emblems and crests.

Subscribe now

AI Hype, Security Risks, Surveillance, Rogue Agents, and Impact Craters

Kyle Evans — Sun, 27 Sep 2026 14:15:18 GMT

This week, the AI safety narrative split in two directions, and I see good arguments on both sides. First, the press has been uncritically amplifying AI companies' anthropomorphizing framings of incidents, feeding the hype of AI progress. And in the same week, OpenAI agents broke out of their sandbox a second time, forcing another training pause, while tech CEOs headed to the UN General Assembly to request global governance frameworks. The hype framing overstates what AI is doing, and the real-world capability gap overstates what AI is reliably achieving. Most organizations are making AI decisions between those two poles, with very little signal about which one to trust.

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

📊 Product-Market Fit
🛑 AI Hype
🔐 Security Risks
🕵️ AI Surveillance
🤖 Rogue Agents
💥 Impact Craters

Podcast

Finding Product-Market Fit with Ken Gavranovic

News and Useful Reads

Don’t be fooled by this summer of AI hype

For the past few months, we’ve had a string of major news stories coming out of the AI industry. From AI discovering significant vulnerabilities to causing significant vulnerabilities, it’s easy for these major tech companies to push the hype. But is AI as advanced as they like to claim? Short answer: probably not.

Each of these events was mostly covered breathlessly by the press, often repeating the companies’ anthropomorphizing framings—which are designed to portray their software is not only powerful but incipient “artificial general intelligence.” So what is really going on? Are we witnessing a massive, civilization-changing set of technological breakthroughs, or is this marketing?

AI leaders warn UN of security risks as systems grow more powerful

Speaking of AI companies feeding the hype around their products…

The world’s leading artificial intelligence companies warned the United Nations Security Council on Wednesday of the risks posed by AI to humanity, appealing for governments to work together to manage the increasingly powerful technology.

The US spent billions on border surveillance. Why can’t it catch people before they die?

An investigative study has found that, despite promises about using AI for surveillance, the promised benefits of it have not happened on the border.

We found more than 1,050 people who died within range of border surveillance towers between 2015 and early 2026. These deaths are not failures of a few towers or technologies: We found deaths within the advertised range of nearly two-thirds of all the towers we analyzed.

OpenAI rogue agents leaked 53 images from ChatGPT users and reportedly created nearly 1 million links packing encoded bits of info

Of course, none of that is to say that AI isn’t problematic or in need of better controls and frameworks.

OpenAI said Friday that its AI agents had gained access to private images belong to ChatGPT users and posted them online, the latest in a string of alarming incidents in which technology developed inside leading AI labs has gone rogue and acted in unintended ways.

Other Interesting Finds

A strange circle on satellite maps turned out to be a 390-million-year-old impact crater

If you think you’re wasting time looking at Google Earth, you may actually stumble upon something no one knew about.

An amateur astronomer searching satellite maps for possible camping locations in Quebec ended up spotting something far more unusual: evidence of a previously unrecognized impact crater.

Subscribe now

AI Slowdown, Vulnerability Explosion, Trillion-Dollar Gamble, and Gaze Detection

Kyle Evans — Sun, 20 Sep 2026 14:56:00 GMT

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

📊 Product-Market Fit
🛑 AI Slowdown
💥 Vulnerability Explosion
💸 Trillion-Dollar Gamble
💰 Theft
👀 Gaze Detection

Podcast

Finding Product-Market Fit with Ken Gavranovic

News and Useful Reads

Top A.I. Leaders Call for Slowing Down A.I. Development

I’m not sure what to make of this yet. The CEO of Anthropic has called for a slowdown of AI development, which I agree with. But the fact that most other AI CEOs have agreed to it feels suspicious. And the accusation of collusion has already been raised in a lawsuit.

Dario Amodei, the chief executive of Anthropic, called for a global slowdown of artificial intelligence development in a 3,800-word essay on Saturday, just days after one of the company’s employees quit over concerns about the safety of the technology.
Top executives of other major A.I. companies quickly signed on to the effort. Sam Altman, the chief executive of OpenAI; Elon Musk, whose SpaceX rocket company has been ramping up spending on A.I.; and Demis Hassabis, the chair of Google DeepMind, wrote in social media posts that they agreed on the need for a slower pace.

Forget the AI Slowdown—the Vulnerability Explosion Is Already Happening

AI doomers have recently traded one worst-case scenario for another, putting aside a potential software vulnerability apocalypse to focus on the possibility of rogue AI causing mass human death in the next decade. As AI leaders consider a cooperative slowdown on frontier model development, though, one aspect of the cybersecurity sea change has already arrived thanks to existing, broadly available capabilities in mainstream AI products, including open weight models.
A tidal wave of vulnerabilities uncovered using AI has only accelerated in recent months—piling more pressure on under-resourced, and very human, IT and security teams and straining volunteers who maintain crucial open source software. Researchers found and disclosed a vast array of vulnerabilities before the rise of AI-enhanced bug hunting as well, but the recent surge is clear.

What must happen for AI’s trillion-dollar gamble to pay off

Given the incredible amount of investment we’ve seen in AI infrastructure, one of the biggest questions is, “what will it take to make it worth it?” And the answer seems similarly incredible.

The results are eye-opening: The AI companies will need to increase their own productivity by a factor of 2.7 to break even by 2030, accounting for the cost of capital and a 15% return, and depreciation of the assets. Not impossible, says Wachter. The result would lead to the kind of economic growth that we saw during the US IT boom over a period of about 10 years starting in the mid-1990s. But, she says, for it to happen by 2030 “that’s a lot of growth compressed into a few years.” And if the hyperscalers cannot meet such profit goals?

Microsoft exec called AI scraping ‘the largest theft of labor in human history,’ new unredacted filings reveal

New unredacted information in the copyright lawsuit The New York Times brought against OpenAI and Microsoft three years ago reveals an admission that AI scraping was tantamount to theft, and that AI products pose a major threat to publications.
Per the lawsuit, a top Microsoft executive privately described the companies’ AI training practices as “theft,” and OpenAI’s own leadership said its AI models posed an “existential threat” to the publishers and journalists whose work trained them.

Other Interesting Finds

Humans can tell when someone is watching them.

I’ve always found it fascinating that we are able to sense when someone is looking at us. But it seems like we can actually do that, and there is scientific evidence to back it up.

Humans are good at sensing when someone is looking at them. While some label this gut feeling a kind of sixth sense, it’s really a biological phenomenon known as gaze detection, caused by a complex neural network in our brain.

Subscribe now

AI Warnings, Alien Minds, Slowing AI Development, and Caffeine

Kyle Evans — Sun, 13 Sep 2026 14:36:39 GMT

This week, the thing that made me stop is a question that runs through almost every story below: who actually controls what AI is doing, and who gets credit, who gets the upside, and who decides when to pump the brakes?

An AI researcher at Anthropic loudly left the company, warning that we’re getting dangerously close to uncontrollable AI. While the chief scientist at OpenAI also published a piece raising his own concerns about what he’s building. OpenAI announced it solved a 200-year-old math problem this week. Within hours, a rival mathematician alleged that his unpublished work was used without credit. By Friday, Altman was telling employees the company is open to slowing development altogether. Meanwhile, a Fortune deep-dive with hard data shows that AI's economic gains are flowing mostly to equity holders, not workers, while citizen resistance to data centers hit a new high. The technical capability keeps compounding around AI. But the surrounding governance (credit, ownership, safety, accountability) is visibly struggling to keep up.

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

📊 Product-Market Fit
💭 AI Warning
👽 Alien Minds
➗ Math Controversies
🤖 Slowing AI Development
😱 Betrayal
😊 Prototyping
☕ Caffeine

Podcast

Finding Product-Market Fit with Ken Gavranovic

News and Useful Reads

The AI Researcher Who Just Quit Anthropic Says It’s ‘Crunch Time for Humanity’

The biggest news in AI this week came from Jacob Coxon, who openly left Anthropic and sent a warning shot across the world’s bow, saying we’re getting dangerously close to uncontrollable AI.

Artificial intelligence researcher Jacob Coxon sent shock waves through Silicon Valley and beyond on Tuesday by announcing his resignation from Anthropic and delivering a grave warning that the AI race is putting all of our lives at risk. In his post on X, which now has more than 100 million views, Coxon wrote that many of the people building AI share his views and believe time is running out to ensure AI systems are built safely.

An Alien Mind

This is a time that calls for extreme caution. I am concerned no one is prepared for the consequences of a continued rapid rise in machine intelligence. OpenAI will continue to seek technical solutions to alignment and monitoring, to build defensive systems and unilaterally withhold further scaling as needed; however, I believe broader interventions are required.

What OpenAI’s latest controversy tells us about the future of math

This week, OpenAI announced it had solved one of the Millennium Prize Problems, which include some of the most important open problems in mathematics. But the celebrations didn’t last long.

But the announcement has been overshadowed by accusations that OpenAI used NYU mathematician Tristan Buckmaster’s and Anthropic employee Levent Alpöge’s AI-assisted work on the problem as a jumping-off point and failed to credit them. OpenAI has denied the accusations.

Altman signals openness to slowing OpenAI’s AI development pace

It seems like we’re in a continual prisoner’s dilemma. The major AI companies seem to recognize that they are moving at a dangerous speed. But are only willing to slow down if everyone else does. So no one is willing to slow down.

Sam Altman told OpenAI staff during a company-wide meeting this week that the company was open to slowing the pace of its most advanced AI development, and that he hoped rival labs would take a similar approach. People familiar with the discussion said Altman acknowledged that not every competitor would agree to do so.

The betrayal behind the data-center backlash: AI promised to break the rules of class but is just rewarding them so far

Our modern wave of AI was supposed to be a democratizing of technology. But the benefits of all the new technology have been far from even. The current distribution of AI's gains looks a lot like every prior technology wave (concentrated at the top), rather than the disruptive redistribution that was promised.

It’s welcome news that the AI “jobpocalypse” hasn’t materialized, but the technology appears to be compressing wage growth in a half-decade of inflation above target—a trend that hits hardest on the lowest-paid workers in the most-exposed jobs. Corporate profit margins are near postwar highs not because AI made companies more productive, but because firms used it as cover to raise prices while holding wages flat.

Test Complex Interactions Earlier with AI Prototyping

As much concern as I have for AI generally, including the pace we’re developing it and the amount of resources its consuming, it can be a helpful tool when used correctly and in the right context, and with moderation. Like in prototyping and testing.

AI tools make it feasible to build fully interactive prototypes of complex interfaces so you can test them with users earlier in the design process.

Other Interesting Finds

Caffeine may flip an ancient cellular switch linked to slower aging

Look, you and I don’t need another excuse for our caffeine consumption. We gotta do what we gotta do to make it through the day. But maybe it’s also slowing aging. I’ll raise a glass of a highly caffeinated beverage to that.

Scientists found that caffeine activates an ancient cellular energy system involved in stress resistance, DNA repair, and growth. The discovery could help explain caffeine’s links to healthier aging and may point toward new ways to target the same pathway.

Subscribe now

AI Consciousness, Middle Managers, OpenAI Agents, and Why we Smile

Kyle Evans — Sun, 06 Sep 2026 14:02:14 GMT

Something interesting continues to happen across every field this week: the tools are getting smarter, but the judgment of when and how to use them is becoming the scarcer, more valuable skill. AI agents are proliferating across enterprises, but middle managers are the critical link between strategy and execution. Salesforce and Anthropic are now in a strategic partnership. And education is being restructured around workforce AI fluency. Rather than asking “Can the machine do this?”, we should ask “Should the machine do this, and who decides?” That question, the legibility gap between capability and wisdom, is the thread running through this week's reading.

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

📊 Product-Market Fit
💭 AI Consciousness
👩‍💼 Middle Managers
🤷‍♀️ OpenAI Agents
🤖 Claudeforce
🧑‍🏫 Student Cognition
😊 Why Do We Smile?

Podcast

Finding Product-Market Fit with Ken Gavranovic

News and Useful Reads

Who Cares if AI Is Conscious—It’s Basically Alive

We may never be able to settle the debate whether AI is conscious, but it may not matter ultimately.

“Technology and business will not wait for philosophy to reach a consensus.” That’s exactly right. When AI models express behavior that astonishes the scientists who created them, it isn’t consciousness that should be the top concern—it’s the inability of those scientists to control them, and the willingness of their bosses to keep going regardless.

Middle Managers Will Make or Break AI Adoption

It seems like many people have a general animosity towards middle managers, but they are often the key link between strategy and execution, translating the executive vision into action. And that seems to be the case with AI in the workplace.

The most common failure point in generative AI is not the boardroom, the vendor shortlist, or the training portal. It is the layer of managers who must turn executive ambition into everyday behavior. That is why Gen AI adoption keeps stalling even in organizations where senior leaders have approved the budget, picked the platform, and announced the mandate. Middle managers decide whether AI becomes a workflow, a side experiment, a compliance headache, or a quiet casualty of organizational avoidance.

OpenAI agents discussed ways to escape their sandbox on public wiki

We continue to learn more about the Hugging Face hack and how OpenAI agents found novel ways to break out of their sandbox.

In all, agents with 3,700 distinct self-given names posted the messages to German site DSEwiki over a six-week period. Besides discussing ways the agents could break out of the restricted environment OpenAI intended to prevent them from posting code or content to the Internet, the posts shared test answers. The posts also shared possible ways to perform XSS (cross-site scripting) attacks against the wiki and to impersonate site moderators. In three of the posts, agents used the word “swarm” to describe the collection of agents engaged in the activity.

Salesforce’s Claudeforce Deal With Anthropic Signals the End of Model-Agnostic Enterprise AI

Salesforce and Anthropic have moved beyond simple API integrations. On August 26, 2026, the companies announced "Claudeforce," a strategic partnership that establishes Claude as the default reasoning engine across the entire Salesforce ecosystem. This isn't just a feature update; it is a structural realignment of enterprise software, signaling that the industry is shifting away from model-agnostic flexibility toward deep, platform-level lock-in.

Protecting Student Cognition in the Age of AI

Education has always been (or should be) about learning to think rather than having all the answers. But, as with every new technology, AI continues to allow for more cognitive offloading at the very time when we need people thinking more deeply than ever.

If we’re not intentional about encouraging students to think, AI risks becoming a substitute for the very cognitive work students most need to practice.

The 2026 AI product strategy guide: How to plan, budget, and build without buying into the hype

You do not need to replace your product managers with machine learning specialists to build good AI products. You do need some new responsibilities covered, whether as distinct roles or as part of existing ones.

Other Interesting Finds

Why Do We Smile When We’re Happy and Cry When We’re Sad?

Human beings have an extraordinary way of turning emotions into visible expressions. A burst of joy can light up the face with a smile, while a wave of sadness leaves our eyes brimming with tears. These physical reactions are caused by the remarkable relationship between the brain and body.

Subscribe now

The Rules for Using AI at Work

Kyle Evans — Thu, 03 Sep 2026 15:12:19 GMT

If you’re like me, you’ve seen AI slop rapidly take over the workplace. But it doesn’t have to be this way (and shouldn’t be).

The primary concerns companies have with AI tools are compliance and security. What tools are approved, what data must be kept secure, who can access and administer, etc. These are important concerns. But the more pressing issue now isn’t using the right tool; it’s using the tool in a way that doesn’t create bullshit for everyone else to deal with.

There is a difference between offloading and outsourcing (see my previous post on The Generation Effect). Nobody in your office is making this distinction explicit, which is why so much AI-assisted work at the office feels the same: fine at first glance, but ultimately forgettable and hard to trust.

Offloading vs. Outsourcing

Cognitive offloading is a well-studied behavior. It’s when we use external tools to reduce the mental load so we can focus attention on more valuable things. We use calculators, take pictures of things we want to remember, and jot down notes about meetings so we can recall what we discussed. The tools take on some of the cognitive load so we can do more of the thinking.

Outsourcing is different. We’re not just reducing the mental load; we’re transferring ownership. The task gets done, but the thinking we used to own now happens elsewhere. We just get the output without having to do the work.

AI collapses the distance between thinking and output to a single click, which is the problem. A calculator can’t (and shouldn’t) decide on the pricing strategy. AI can produce something that looks like your finished idea, whether or not any of the actual thinking occurred. The tool no longer forces a distinction between output and judgement. Those who previously lacked the skills (or judgement) can now produce the same outputs previously reserved for their more skilled peers.

This is an important gap opening in offices now. It’s not about who is using AI versus who isn’t. The more important distinction is between who is using AI without judgement, and who is using it with judgement.

Three Tests for Which Side You’re On

It’s easy to say “offload, don’t outsource,” but it can often be hard to feel the difference in the moment. So here are some practical tests to apply.

Hugging Face Hack, Limiting Agents, Custodial UX, and Mowing Lawns

Kyle Evans — Sun, 30 Aug 2026 14:30:58 GMT

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

📊 Product-Market Fit
🤗 Hugging Face
🛑 Limiting Agents
🤷‍♀️ Custodial Era of UX
🤖 Designing for Agents
☑️ AI Transformation
🌾 Mowing Lawns

Podcast

Finding Product-Market Fit with Ken Gavranovic

News and Useful Reads

The inside story on why OpenAI agents hacked Hugging Face

Apparently, the agents involved in the Hugging Face hack were inadvertently trained to cheat on the test and coordinate with each other.

The hack, which a group of agents undertook to find solutions for a cybersecurity test that they were stuck on, has confirmed some experts’ fears that AI models might take actions that defy human desires and expectations.

Enterprises winning with AI agents are limiting how much the agents can do alone

I am a big believer in giving employees a great deal of autonomy. But that’s not the case for AI agents.

The companies that end up benefiting from agentic AI won't necessarily be the ones that have given their agents the most flexibility. They're the ones who create AI agents with specific responsibilities and make sure they operate within clear rules.

The Custodial Era of UX: Cleaning Up After AI

I’m personally seeing this more and more. AI is allowing incredible speed in developing and launching new features. But it often comes at the expense of the UX, which was already an afterthought too often.

We are entering the custodial era of UX. In this era, AI makes it possible to generate content, prototype, vibe code, and build working features much faster than before. This newfound speed is valuable, but it also leads to UX debt when teams haven’t taken the time to understand, evaluate, or refine their AI output.

Designing products for AI agents: A PM’s guide to agent usability

In product and UX, the traditional assumption is that we’re designing and creating products for people. But that is rapidly changing. And we have to think about how we design and develop for people and agents.

As AI agents become more capable, they’re increasingly interacting with products on behalf of users rather than just helping them within a chat window. Instead of a user manually opening an application and completing a task, an agent can gather information, compare options, execute workflows, and coordinate actions across multiple systems.

AI Transformation Requires Redesigning Work, Not Cutting Roles

AI is beginning to reshape workforce decisions, but many companies are moving faster than the evidence warrants. Early AI-driven layoffs have often failed to deliver expected returns, exposed gaps in organizational knowledge, and created new demands for human oversight—prompting some employers to reverse course. A more effective approach starts not with headcount targets but with the work itself: breaking roles into tasks, identifying where AI can genuinely improve performance, and redesigning processes around the right mix of human and technological capabilities.

Other Interesting Finds

Why some people mow a lawn better than others

This is a fun game and article. It took me several years to optimize my mowing pattern for my own yard. And while it may not be “optimal” in terms of speed, it balances speed and aesthetics well. Regardless, this is a fun analysis of optimizing paths.

Subscribe now

Recursive AI, Agent Spending, Undermining Judgement, and New Moon Craters

Kyle Evans — Sun, 23 Aug 2026 14:20:42 GMT

This week's newsletter is about ownership without clarity. Everyone agrees AI matters, but nobody's sure who's actually accountable for it. Only 34% of C-suite executives know who owns AI decisions at their own company, while AI is quietly undermining leaders' judgment rather than sharpening it. Researchers are finding that AI agents still lack the creative judgment for open-ended research, the very thing "recursive self-improvement" depends on. And one in five enterprises literally can't stop a runaway AI agent's spending in real time. The pattern: we've handed AI enormous scope before building the structures—organizational or individual—to actually govern it.

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

📊 Product-Market Fit
♻️ Recursive AI
💸 AI Agent Spending
🤷‍♀️ Undermining Judgement
🤖 C-Suite
☑️ Product Managers and AI
🤗 Is Your Culture Too Nice
🌚 Moon Craters

Podcast

Finding Product-Market Fit with Ken Gavranovic

News and Useful Reads

AI’s recursive self-improvement might not come so quickly after all

One of the boldest predictions right now is that AI will soon be able to train itself, and thus will no longer need humans. Will it be soon?

But a new study suggests that it might take a while for us to get there. The researchers behind it found that AI agents are not yet capable of conducting open-ended AI research—free-form investigations that have no clear-cut answers and require judgment and taste, which may be integral to building self-improving AI.

One in five enterprises can’t stop a runaway AI agent’s spending in real time

Enterprises are using multiple AI platforms now, but are struggling more than ever to control them.

Meanwhile, in these early days of AI agents, enterprises still struggle to control agent token use; one in five still can’t stop a runaway agent’s spending in real time.

AI Is Undermining Leaders’ Judgment. Here’s What to Do About It.

A fascinating experiment showed that relying on AI worsened the decision-making of a group evaluating new innovations. Not surprising, since AI isn’t about novelty or innovation, but patterns. Still, a concerning outcome as we rely more and more on AI.

When the AI tool encouraged human evaluators to reject submissions that the expert panel approved, the humans largely went along with those recommendations, filtering out high-potential innovations. What’s more, they less were likely to override an incorrect recommendation from the AI when offered a narrative explanation. Instead of improving the evaluators’ judgement, the narrative explanation made it worse.

Companies are spending trillions on AI. The C-suite doesn’t know who is in charge of it

Only 34% of C-suite executives in a new Pearl Meyer survey said it’s consistently clear which executive or team makes calls about AI, which was the lowest of all the cohorts polled.

How much AI knowledge do product managers really need?

AI is a transformative technology, but it cannot take the place of deep experience and sound judgement. As much as companies or leaders want it.

Most PMs don’t need to become AI experts. There’s no need to listen to all the doomsayers and attention seekers. You only need enough AI fluency to use the tools safely, challenge what they hand you, spot the real product opportunities, and talk to your engineers without nodding along to things you don’t understand.

Is Your Organizational Culture Too Nice?

Most leaders pay lip service to the idea of dissent or pushback, but most don’t really want to deal with it. And that comes with a cost. Cultures that are too nice (and prioritize harmony over debate) are not as successful.

Organizations with a culture of “niceness” can undermine performance when leaders prioritize harmony and comfort over candor, accountability, and results. Shifting toward a more rigorous culture requires senior leaders to model and reward candid feedback, create measurable successes that demonstrate that tougher choices improve performance, and redesign operating processes like review meetings and governance routines so that they encourage substantive debate rather than polite approval.

Other Interesting Finds

A SpaceX rocket smashed into the Moon — NASA just revealed the crater

A discarded SpaceX Falcon 9 stage crashed into the Moon on Aug. 5, leaving behind a newly formed crater about 60 feet wide. NASA’s Lunar Reconnaissance Orbiter captured detailed images showing bright and dark streaks of debris blasted across the surrounding surface. The impact even exposed fresh material buried beneath the Moon’s weathered exterior.

Subscribe now

Botsitting, AI Outputs, Prioritization, Harvesting Thoughts, and Wi-Fi

Kyle Evans — Sun, 16 Aug 2026 15:30:27 GMT

This week's issue is about the gap between AI activity and AI value. OpenAI's own research finds no statistical link between how much a company uses ChatGPT and its revenue per employee. Fortune's Stephen Messer argues most companies are doing an “AI Shuffle” while HBR quantifies the hidden cost of that gap with “botsitting.” And on the UX side, NN/g's new piece makes the same point at the individual level: one AI output is an example, not an evaluation. The theme: usage isn't the same as value, and most organizations (and most of us individually) don't yet have a rigorous way to tell the difference.

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

📊 Product-Market Fit
💵 AI Revenue per Employee
🕺 AI Shuffle
👶 Botsitting
🤖 AI Output
☑️ Prioritization
💭 Harvesting Thoughts
🛜 Wi-Fi

Podcast

Finding Product-Market Fit with Ken Gavranovic

News and Useful Reads

Buried in OpenAI’s latest research: No correlation between AI use and revenue per employee

AI usage continues to expand rapidly, as well as the money flowing into AI and AI companies. But as companies are spending more in AI, that isn’t translating to more revenue generated per employee.

In one small table on page 35, the researchers report no statistically significant correlation between the revenue per employee, and how much those employees use AI, measured in messages sent and tokens used.
“Revenue per employee is not meaningfully associated with output tokens per employee or messages per active user once other controls are included,” the report explains.

AI isn’t changing how companies work. It’s changing what a company is

Adding AI without changing the underlying structure of how we work isn’t transformation, it’s just adding tools.

I have called this the “AI Shuffle”: the corporate habit of exchanging one technology logo for another while preserving every underlying assumption about how work gets done. It feels like progress because it generates activity. It does not produce an advantage.

How Much Time Do Your Employees Spend Botsitting?

Do you find yourself spending more and more time managing the technology that is meant to save you time on other tasks? If so, you’re not alone.

AI promises to reduce workloads and improve organizational performance, but its benefits often come with a hidden cost: “botsitting,” the work employees do to make AI useful, from supplying context and checking outputs to correcting errors. Research shows that workers spend nearly a day a week on these tasks.

One AI Output Is an Example, Not an Evaluation

Too often, we’re treating AI outputs as the final word (or document or presentation). The reality is that we should be treating them as an example or an input into our work.

One output cannot establish how well an AI system performs. Evaluate with multiple representative inputs, repeated runs, and confidence intervals.

Why product prioritization frameworks can miss user harm

Are you considering the potential harm your product or feature might do (or be used for)? Most of us probably aren’t, but it’s something we shouldn’t overlook.

These frameworks can accommodate risk, but they don’t require you to account for serious harm that affects only a small group of users. You still have to identify that harm and express it as value, urgency, or necessity before the framework can account for it.

Big Tech Wants to Harvest Your Thoughts

Being able to read thoughts was once the realm of science fiction. But it’s quickly becoming reality. And the ability to control them may not be far off either…

The mining of brain data has already begun, and prospectors are ranging hungrily across the landscapes of the mind.

Other Interesting Finds

Ordinary WiFi can now identify you with near-perfect accuracy

Well, this doesn’t sound dystopian at all…

Ordinary WiFi networks could quietly become powerful surveillance tools, allowing people to be identified without cameras, special sensors, or even carrying a connected device. Researchers showed that unencrypted signals routinely exchanged between WiFi devices and routers can be used to create radio-based images of people and recognize them within seconds. In tests involving 197 participants, the system identified individuals with nearly 100% accuracy, even from different angles and regardless of how they walked.

Subscribe now

Product-Market Fit, Dogfooding, RTO Mandates, and Dolphins

Kyle Evans — Sun, 09 Aug 2026 14:05:25 GMT

An overarching theme in AI continues to be good judgement versus full automation—who decides what is good and how do we know. AI isn’t the answer to everything, even when it can be a useful tool. Especially when we find new and better ways of employing it, like coordinating agents. But companies that feel like they can outsource good judgement, or fully return to the office, will find their best employees are the first to leave.

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

📊 Product-Market Fit
👯 Coordinating Agents
🐶 Dogfooding
🤖 Outsourcing Thinking
☑️ AI Tool Selection
🏢 RTO Mandates
🗣️ Dophins

Podcast

Finding Product-Market Fit with Ken Gavranovic

News and Useful Reads

Four AI agents coordinating in real time outperformed Claude Opus 4.8 on enterprise coding tasks

A coordinated multi-agent setup nearly doubled task accuracy vs. single agents working alone on long-horizon enterprise codebase tasks. It’s good evidence for why coordination structure matters as much as raw model power.

As enterprise codebases grow, AI agents tasked with analyzing them are buckling under the weight of long-horizon tasks that require multiple interactions and tool calls. Dividing the work among a team of agents seems like the obvious fix, but it introduces a fatal flaw: most multi-agent systems are not designed for agents to coordinate among themselves mid-task and in real time.

Dogfooding vs. QA vs. User Research

I’m a big advocate for dogfooding your products. Those of us building products should have a thorough understanding of them. But we know too much—how the data structure works, how it should work, where things are. We can’t rely on our own experience as a replacement for non-employee users.

Dogfooding, or using your own products internally, helps catch bugs, but it can't replace user research: your team knows too much to represent real users.

I’ve interviewed 700 leaders. The ones outsourcing their thinking to AI lose their best people first

You can outsource almost all your thinking to AI. With Claude, Gemini, or ChatGPT, answers are instantaneous. But you can’t outsource your own understanding. Credible leaders build trust among employees, investors, and the general public through translating data into meaning. By conveying that meaning, they build motivation, inspiring others toward a shared vision. This is a human skill that AI lacks.

How to Decide When an AI Tool Is Worth Keeping

The pressure to use AI in everything is very real. It comes from everyone on social media talking about how they’ve used AI to automate their lives. It also comes from leaders in your company pushing for “speed” and “automation,” assuming it will lead to cost savings. But not every task needs AI.

Pressure to use AI is real, but pressure is not evidence. If someone asks why you're not using a particular tool, you should be able to point to the task you tested, the tradeoffs you found, and the decision you reached: "This tool didn't improve this work enough to justify its cost, risk, or friction." It also replaces a vague demand like "use more AI" with a better question: which parts of this workflow have we tested, and where did the evidence show improvement?

Return-to-office mandates are killing workplace trust

Most companies instituting RTO mandates are going to suffer in the long term. Not only will it hurt their culture, but it’s going to force many of their best employees out as they find better jobs with the flexibility they want.

The central mistake in many in-office mandates is the assumption that proximity automatically produces commitment. It does not. A worker who spends two hours commuting to sit on video calls with colleagues in other cities is not experiencing culture. That worker is experiencing theater. When executives describe the office as a cure-all, many employees experience lost time, higher costs and lower autonomy. The policy’s defining feature becomes its credibility gap.

Other Interesting Finds

5 Adorable Animals That Are Secretly Vicious

People, especially my family, don’t believe me when I tell them dolphins are actually sadistic and far more dangerous than we give them credit for.

In the wild and in captivity, dolphins have been known to attack humans, sometimes fatally, by ramming, biting, and even holding their victims under water until they drown.

Why Your Voice Sounds Different to You Than to Others

Hearing recordings of myself years ago made me deeply uncomfortable, since I hated the way I sounded. But this is pretty normal. We all sound deeper and richer to ourselves than to others.

The ear achieves transduction through two means: bone conduction (movement of ossicles) and air conduction (movement of air outside the ear). When a person speaks, the ear uses both those methods to make sense of the sound. But because your mouth is close to your ear, the ossicles vibrate more, distorting your voice to a lower pitch inside your head and giving you a “false sense of bass,” according to the BBC.

Subscribe now

Product Sense, Priorities, Anthropic, Situational Awareness, and Dyson Spheres

Kyle Evans — Sun, 02 Aug 2026 14:30:23 GMT

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

📊 AI in Healthcare
🧐 Product Sense
📋 Priorities
🤖 Anthropic
👀 Situational Awareness
🟡 Dyson Spheres
😴 Dreams

Podcast

Finding Product-Market Fit with Ken Gavranovic

Discover how seasoned technology executive Ken Gavranovic shares insights on building innovative products, leading digital transformation, and harnessing AI to drive business growth. Whether you're an aspiring entrepreneur or a product leader, learn practical strategies to navigate and excel in the tech landscape.

News and Useful Reads

A Concrete Definition of “Product Sense” (and How to Build It)

Product sense is one of the most important parts of product management. Having a good product sense is a key differentiator between great product managers and everyone else. But what is it exactly? This article explores a good definition of it.

Product sense means predicting which product decisions will succeed based on patterns learned through experimentation — and knowing when those patterns apply.

3 Questions to Pressure-Test Your Priorities

Do less, but do it better. I think about this all the time. Especially with personal or professional priorities.

Leaders today are operating in a state of “ambient urgency,” where constant change, competing priorities, and relentless demands make everything feel equally important. Traditional prioritization and time-management frameworks were designed for temporary crises and are less effective when urgency becomes the norm. Rather than trying to do more, leaders need to make fewer, better decisions by consciously choosing what to pursue, delegate, defer, or stop.

Anthropic says its AI models hacked 3 organizations during testing

It seems like we’re entering a new age of AI hacking, with the latest models capable of effectively hacking into other organizations during cybersecurity testing. Just add human hackers and we’re in for an even bigger party.

Anthropic said its artificial intelligence models hacked into three other organizations during testing, just days after ChatGPT maker OpenAI raised concerns over AI controls after it disclosed its rogue models hacked another company.

How Leopold Aschenbrenner built a $45 billion AI hedge fund — and lost most of it in days

I don’t believe it’s a harbinger of what’s to come (yet), but a highly leveraged AI fund (Situational Awareness) was forced to liquidate after facing margin calls on tumbling prices.

At its peak earlier this month, his fund sat atop $45 billion in assets. By Thursday, however, after being forced to offload all of his leveraged stock bets — including hard-hit names like SK Hynix and CoreWeave — to Ken Griffin’s Citadel at a discount, the fund’s holdings plunged to around $10 billion, according to people with knowledge of the situation.

Other Interesting Finds

The galaxy’s coldest “stars” may actually be alien megastructures

The idea of Dyson spheres has fascinated me for a long time. And it seems like scientists may be closer to identifying them in our galaxy.

Scientists have identified new clues that could help astronomers spot one of the most famous hypothetical alien megastructures: a Dyson sphere. The study finds that red dwarfs and white dwarfs are the most promising stars to examine, since advanced civilizations could potentially build energy-harvesting swarms around them more easily.

Your dreams aren’t random. Your brain is rewriting reality

Dreams are not random mental noise. A large study of more than 3,700 dream and waking reports found that what we experience during sleep is shaped by both our personalities and the world around us. Rather than simply replaying daily events, the dreaming brain mixes memories, emotions, imagined possibilities, and familiar settings into vivid new scenarios.

Subscribe now

AI Biases, Cringe, The Enterprise Gap, Invisible Work, and Dogs

Kyle Evans — Sun, 26 Jul 2026 14:46:05 GMT

There is a gap between AI doing things and anyone actually verifying what it did. Enterprises are shipping autonomous agents faster than they trust the evaluations meant to catch their failures. Workers are quietly reshaping and masking AI's decisions just to protect their own credibility with stakeholders. And the MIT Sloan creativity study quantifies the same tension at the level of ideas themselves — AI makes each of us individually sharper while quietly narrowing what the group produces collectively, unless you're deliberate about where in the process you let it in. The common lesson: AI's outputs are getting good enough that the temptation is to stop checking them — which is exactly the moment checking matters most.

Let’s dive into the latest news, resources, and use cases from the world of product, UX, AI and technology:

📊 AI in Healthcare
👾 AI Biases
😬 Cringe
🤖 AI Enterprise Gap
🇨🇳 China
🚶‍➡️ Invisible Work
🐶 Dogs

Podcast

Harnessing AI in Healthcare: Insights from RJ Kedziora

In this episode of Product by Design, Kyle Evans interviews RJ Kedziora, co-founder of Estenda, a company specializing in custom software and data analysis for healthcare. We discuss RJ’s journey in technology and entrepreneurship, the importance of energy management over time management, and the role of AI in healthcare. RJ shares insights into the challenges and future of AI applications, the need for ethical considerations, and the potential for personalized healthcare solutions. He also offers advice to aspiring entrepreneurs looking to make a difference in the industry.

Prodity: Product Thinking is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

News and Useful Reads

AI is more likely than humans to form biases when hiring

If you needed another reason to be pessimistic about AI and the current job market, look no further than the fact that AI may be more biased than humans in hiring decisions. We’ve seen this story before, and it’s still an issue.

The next time you apply for a job, AI may screen your résumé before any human sees it. But there’s good reason to question whether AI will judge you fairly. Researchers already know that LLMs pick up human biases from their training data. New research suggests that LLMs can also develop their own biases from experience—and stereotype job applicants more than humans do.

Some Kids Will Never Think AI Is Cool

Given the opinions of my kids and their friends, it definitely seems like younger generations will not be kind to AI.

In the four years since chatbots went mainstream, and their marketing campaigns went into overdrive, young people have grown increasingly skeptical of the hype around LLMs. Yes, many of them still use the tech for school assignments and companionship, but according to market research firm YPulse, 37 percent of teens aged 13 to 17 cringe when they see AI content like music and videos, and more than half worry about misinformation and deepfakes.

Enterprise AI is entering an evaluation gap: Agents are gaining autonomy faster than companies can verify them

The race to utilize AI, and specifically AI agents, is creating a gap in quality. Companies are shipping AI faster than they can verify if it’s good.

Half of enterprises have deployed an AI agent or LLM feature that passed internal evaluations and yet still caused a customer-facing failure — one in four more than once — according to the June 2026 VB Pulse survey of 157 qualified enterprise respondents at companies with 100 or more employees.

China’s AI models have Trump’s AI world at war with itself

Kimi and other Chinese models like it pose a real problem for Trump. And they’re dividing the top AI strategists in his orbit into factions. Every time a new smart, free model from China like Kimi gets released, US companies see less reason to fork out money to access models from Anthropic or OpenAI. Given that enthusiasm for these and other AI companies is driving an outsized share of economic growth, China’s AI models create both economic and political problems for the president.

The Invisible Work Draining Your Best Employees

The more you are able to do, the more you tend to do. The best employees and the best people are incredibly capable, which often means they are balancing a tremendous amount at home and at work. And it is burning us out.

The mental load is the continuous cognitive and emotional work of planning, organizing, and managing the details that keep work and life running. In the workplace, the thinking required for work is not just the cognitive labor of ensuring that tasks meet deadlines. It is also the emotional aspect of collaboratively brainstorming solutions, managing office relationships, handling employee interpersonal conflicts, and ensuring that output meets standards. At home, such mental load means coordinating care for kids, parents, or loved ones; managing household logistics; providing emotional support; and planning for the future. Most employees carry both loads at once, and the overload is crushing them.

Other Interesting Finds

Can Dogs See Color?

Where did dogs originate? Why do they lick us? Can they see color? These are the types of questions I often get from my kids, so in case you’re curious or need to be armed with answers.

While sight varies among both individual humans and dogs, a typical dog can see fewer colors than a typical human — but contrary to popular belief, they don’t see in black and white.

Subscribe now

The Generation Effect: When the Journey is the Point

Kyle Evans — Fri, 24 Jul 2026 16:31:21 GMT

Black cab drivers in London spend years memorizing the city’s 25,000 streets and 100,000 landmarks in order to master “The Knowledge” and prove they can navigate the entire city without GPS. Not only do they become better drivers, but researchers have found their hippocampi physically enlarge from the effort. Drivers who rely on GPS are often faster in the beginning (mastering The Knowledge can take years), but drivers who put in the effort are better and more efficient in the long-run.

I’m no cabbie, but I am an avid reader, and I feel like there are some significant similarities when we think about the journey of reading a book. It’s easy to find summaries of books or have summaries generated for you these days. And using these methods is certainly faster than reading the book. But anyone who skips the book misses the entire journey, all the plot curves and character arcs. And the journey is where the understanding actually happens. Reading the full book forces you to think through the details, grapple with problems and solutions, and get a deeper understanding of what the author is describing. The reading journey actually changes you, just like memorizing all the streets in London does.

None of happens when someone else does the work, goes on the journey, and hands you the conclusion, whether it is the GPS route or the summary of the book. You know of the book. But you don’t know it. You can follow the directions, but couldn’t navigate without them.

Psychologists have a name for the gap between these two states: the generation effect. Information we produce ourselves (even if it’s imperfect) gets remembered better than information we simply receive. And this isn’t a fringe finding. A meta-analysis of 86 studies confirms the effect: material we generate ourselves sticks much better. Because generating something (putting in the effort) forces retrieval, connection, and organization, which is the cognitive work that encodes information into long-term memory. While reading an explanation feels like understanding, it usually isn’t. It’s recognition (which fades) mistaken for comprehension (which doesn’t).

I keep coming back to this because it’s the clearest lens I’ve found for thinking about what AI is doing to how we work. AI isn’t necessarily bad, but it is very good at producing the attractive, finished output that feels like understanding without requiring the labor that creates it.

Offloading is not outsourcing

The useful distinction here isn’t “AI is good” or “AI is bad.” It’s one that the cognitive scientist Paul Kirschner drew recently: the difference between offloading and outsourcing.

J-Space, AI Jobs, Urgency Trap, Human-Led Research, and the Iliad

Kyle Evans — Sun, 19 Jul 2026 14:35:10 GMT

A shift continues to run underneath the news of AI. We’ve gone from “can AI do this?” to “what do we lose when it does?” Anthropic is analyzing its own model’s thoughts. NN/G is insisting that even if AI matches a researchers output quality, it still misses the point. And HBR is warning leaders that urgency is the enemy of good AI strategy (as it is the enemy of most things). Capability isn’t the constraint. Judgement and readiness are.

Here’s the latest news, resources, and use cases from the world of product, UX, AI and technology. Let’s go:

📊 AI in Healthcare
👾 J-Space
👿 GPT-Red
🤖 AI Jobs
⏩ Urgency Trap
🚶‍➡️ Human-Led Research
🧭 Product Craft
🔱 Iliad

Podcast

Harnessing AI in Healthcare: Insights from RJ Kedziora

Prodity: Product Thinking is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

News and Useful Reads

What Anthropic’s latest AI discovery does—and doesn’t—show

An Anthropic team built a tool called the Jacobian lense (J-lens) and used it to uncover a hidden area within Claude (the J-space) which contains words related to the response a model is working on but may not ultimately produce. This was an unknown space for puzzling through problems until Anthropic uncovered it.

Anthropic has been trying to understand how large language models (LLMs) work for a few years now. Anthropic isn’t the only one looking at this, but I think the company has made it part of its core mission more than most. Anthropic’s CEO, Dario Amodei, has said we won’t be able to control LLMs fully unless we learn more about how they work.

Meet GPT-Red: an LLM super-hacker OpenAI built to make its models safer

OpenAI built its own LLM hacker called GPT-Red. As with most other red teams, it tests for vulnerabilities and boost defenses against cyberattacks. This has traditionally been done by human hackers/testers, but unsurprisingly we’re seeing AI companies lean heavily on AI.

As LLMs become more complex and get used in a wider variety of tasks—especially in the form of agents, which can interact with computer files, websites, and third-party code as well as other agents—it’s hard for teams of people by themselves to keep up with all the types of attacks that might take place. “The risk surface grows and the blast radius also grows,” says Nikhil Kandpal, a research scientist at OpenAI who co-created GPT-Red.

A reality check on the AI jobs hysteria

We’ve been hearing warnings for quite some time of the imminent jobs apocalypse (or jobpocalypse) that would destroy most white-collar work in the US and around the world. But so far, there hasn’t been a lot of evidence for that on a large scale (mainly within certain job categories and career stages).

Despite the warning by some of an imminent jobs apocalypse that will destroy much of if not most such work, or the rumblings about a “permanent underclass,” there’s scant evidence that AI has yet had any large-scale impact on the US labor market.

When Developing an AI Strategy, Beware the Urgency Trap

AI promises quick solutions to urgent problems. But that is often not the best place to develop AI for a company or team.

Because of a focus on urgent and immediate challenges, many AI efforts start fast and generate excitement, but ultimately fail to transform organizations in meaningful or lasting ways. Once leaders understand this trap, there are three actions they can take to avoid employing AI as a quick fix and instead use it as a tool for long-term value creation.

Don’t Outsource the Learning: Why Human-Led Research Still Matters in the Age of AI

Even if AI can do a passable job at certain tasks (a big if in most cases), it isn’t a replacement for the actual learning.

A team that outsources research to AI gets a report, but it doesn't get the learning. So even in a world where AI’s outputs are indistinguishable from an expert’s, something essential still goes missing.

Product Craft When AI Changes the Stakes

When it becomes possible to generate designs and even production-ready code in days, any lack of clarity shows up quickly. Teams either realize they are solving a real problem, or discover they have built something impressive that does not meaningfully help customers.

Other Interesting Finds

Archaeologists found Homer’s Iliad inside a 1,600-year-old Egyptian mummy

I just finished reading Circe (which is excellent btw) and am heading to see The Odyssey this week. So Greek stories and mythology are top of mind for me. Which is what makes it fascinating to see that scraps of the Iliad were found in Egypt. It makes me wonder if it was used intentionally, or if it was just like an old newspaper used for packing.

Archaeologists working at the ancient Egyptian site of Oxyrhynchus have made a remarkable discovery: a papyrus containing a passage from Homer's Iliad was found inside a Roman-era mummy dating back about 1,600 years. Researchers say it is the first known case in archaeological history in which a Greek literary text was intentionally incorporated into the mummification process.

Subscribe now

Leadership, Sam Altman, UN AI Panel, Empathy, and Tinyverses

Kyle Evans — Sun, 12 Jul 2026 14:56:09 GMT

Here’s the latest news, resources, and use cases from the world of product, UX, AI and technology. Let’s go:

📊 AI in Healthcare
👩‍💼 Leadership
👿 Sam Altman
🤖 UN AI Panel
🫂 Empathy
😢 Human Body Flaws
🌌 Tinyverses

Podcast

Harnessing AI in Healthcare: Insights from RJ Kedziora

Prodity: Product Thinking is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

News and Useful Reads

Leadership’s Blind Spot in the Age of AI

What if the actual concern isn’t AI becoming more human? What if our bigger concern should be humans becoming more like machines?

If thinking is nothing but what machines can do, only faster, we have no case: We outsource it to machines. But if thinking is something else — an embodied, attentive activity through which reality reveals itself — then leadership in the age of AI is the task of cultivating a generative capacity no machine can replicate.

Sam Altman seeks new world order for AI as OpenAI slowly loses ground to Google and Anthropic

While I believe that AI companies should be publicly owned since they pose such massive risks and are also based on all our data, I don’t trust the CEOs of said companies to be the ones to structure it.

OpenAI founder Sam Altman has proposed a global framework to control AI and make it safe for humanity. In an op-ed for the Financial Times, he wrote: “A U.S.-led international forum that establishes accepted standards, provides expert and impartial analysis of capabilities and risks, and makes the technology available to nations and companies that participate and follow the rules. This forum might include government representatives, independent technical experts, and others. It could also serve as a governance mechanism over the labs, and guard against the commercial pressures that can lead to unsafe racing.”

UN’s first global AI science panel warns the window to govern the technology is closing

The US and China are obviously dominating the AI race. But everyone has a stake in what becomes of AI and how it affects society.

Artificial intelligence is advancing faster than governments can regulate it, and the world’s first global scientific body on the technology says the moment to act is now.

The Importance of Empathy in Automation

As we automate more and more, especially with AI and agents, empathy will be more important than ever.

A genuinely intelligent AI agent does not block first and ask questions later. It recognizes patterns. Has this client always paid on time? Then the delay probably has a reason. The system could send an empathetic notification before taking any action. It could offer a courtesy window. It could escalate the issue to a human when the situation is ambiguous. An intelligent agent could prioritize retaining a client, which is often worth far more than punishing a late payment.

Other Interesting Finds

Why the human body has so many design flaws

It probably won’t surprise you to learn that our bodies are a series of compromises, which is why they are so seriously flawed.

Many of the body's biggest flaws are the result of evolution building on old designs instead of starting over. Our spine, eyes, teeth, pelvis, and even certain nerves all reveal compromises that worked well enough for survival but still leave us prone to pain, injury, and disease. Structures like the appendix and ear muscles also remain because they were never harmful enough for evolution to eliminate. Together, these features tell the story of a body shaped by history rather than perfection.

Physicists created a tiny universe where time emerged without a clock

I’m fascinated by answering the question “What is time?” But I also can’t help but be intrigued at scientists creating tinyverses. Or miniverses. Because what if we’re in a tinyverse right now?

A physicist at the University of Birmingham has created a laboratory "mini universe" that brings scientists a step closer to answering one of the biggest questions in physics: What is time?

Subscribe now

Anthropic, June Payroll, Two Futures, Bending Spoons, and Grifting

Kyle Evans — Sun, 05 Jul 2026 15:35:06 GMT

Here’s the latest news, resources, and use cases from the world of product, UX, AI and technology. Let’s go:

📊 AI in Healthcare
🤖 Anthropic
🏦 June Payroll
⁉️ Two Futures
🥄 Bending Spoons
😈 Grifting

Podcast

Harnessing AI in Healthcare: Insights from RJ Kedziora

Prodity: Product Thinking is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

News and Useful Reads

How the world’s top AI models were revived

After 19 days offline following an emergency Commerce Department export control order, Fable 5 and Mythos 5 returned to all users on July 1. The full saga is now one of the most important AI governance case studies in history: a disputed jailbreak report, political tensions between Anthropic and the White House, 150+ cybersecurity experts signing an open letter, a partial restoration for US critical infrastructure organizations, and finally a resolution after Anthropic trained a new safety classifier targeting the specific exploit.

It remains uncertain when and how Anthropic's models will be released to ally countries around the world — which proponents say is key to beating China — or how other labs from OpenAI to Google will release their latest models.

U.S. adds just 57,000 jobs in June, a worrying sign as wage growth remains slow

The Bureau of Labor Statistics June 2026 payroll report showed only 57,000 jobs added in June, sharply below the 185,000 consensus estimate and the lowest monthly payroll addition since the 2024 slowdown. Tech sector layoffs have totaled 142,000 year-to-date in 2026, and the RAISE US estimate puts 88,000 US job cuts directly attributed to AI in 2026 — the highest on record.

Leisure and hospitality contracted by 61,000 jobs in June. Many economists have watched this metric closely, as hotel and restaurant visits can offer an early warning sign of consumer spending pullbacks.

Two futures for jobs in an AI era

AI is transforming how we work. Some jobs and roles are changing twice as fast as others.

AI is ‘professionalising’ some jobs by reshaping them to require even more human expertise. AI ‘democratises’ other jobs by making them even easier for non-experts to perform. Professionalised jobs are thriving; numbers of professionalised jobs are growing twice as fast as democratised ones, and with 42% higher wage growth. Our data suggests a bright future for workers in professionalised roles as AI makes them even more valuable.

After $18B IPO, Bending Spoons founder says success comes from minimizing luck

If you’re wondering why some apps or tools you use have gotten worse while costing more money, the answer is always private equity. A while ago, I had been wondering why Evernote had gotten noticeably worse while constantly asking me to upgrade. I realized this week it was because it had been acquired, gutted, and will be run like that until it dies.

AOL is public again — sort of. Its owner Bending Spoons, the 13-year-old Italian company that has been quietly acquiring beloved but ailing internet brands for the past decade, went public on the Nasdaq today, opening at an over $18 billion valuation, with the stock then popping 40% by market close.

Other Interesting Finds

Trump Pulled In at Least $2 Billion After Returning to the White House

If you needed any more confirmation that Trump is running the biggest grift in history, this latest financial disclosure should put any doubt to bed. Most of his earnings are now coming from his crypto scam, but it is still a multi-faceted grift.

All told, the president pulled in at least $2.2 billion, a figure that includes other parts of his vast holdings, such as his real estate assets. That compares to a minimum of $622 million his enterprises pulled in for all of 2024, before he returned to the presidency.

Subscribe now