Product leader by day. Reader, tinkerer, occasional writer, and starter of an alarming number of side projects by night. I consume far too many articles, books, and podcasts, and care far too much about Boston sports.
Airplane WiFi has historically been, pardon my French, shit. Free or paid, it barely made a difference. When it worked, it felt like early-90s dial-up. You know, the beeps and bops, 56K modem kind of Internet. Enough to be connected but slow, where anything you did required unheard-of 2026 patience.
Weirdly enough, I’m going to miss that.
Bad airplane WiFi had an unintended benefit: flying was one of the few remaining places in the modern world where you could actually disconnect. You could read a book, watch a movie, write something, or just raw dog it without constantly being pulled back into Slack, email, texts, Instagram, notifications, and the rest of the Internet.
When you boarded a plane, it created a boundary. For a few hours, nobody really expected you to be available. However, Starlink changes all of that. I can now say, I’ve seen the future: incredibly fast Internet, for free, 30,000 feet in the sky. On my flight, it didn’t feel like “airplane WiFi.” It just felt like WiFi.
A small part, a very tiny, minuscule part, of me is sad because bad airplane WiFi felt somewhat like a feature, not a bug.
Last year, I wrote about how I extract synced Kindle highlights and send them to Readwise. This script stopped working a few months ago because the Kindle firmware upgrade broke things, My Clippings.txt which is always fun. Claude and I spent some time reworking the script this evening to solve the problem. The repo has been updated and now includes a quick start section in the readme.
Last week I attended the Lenny & Friends Summit in San Francisco. Unsurprisingly, just like the billboards in SF, AI was front and center. There was a lot to digest from the talks, workshops, and side conversations, but what stood out was everyone felt that the core tenets of product management aren’t going away, but the way product managers do work is, like in years past, evolving.
As a result, Product leadership needs to evolve too. AI is obviously changing what individuals and teams can accomplish; at the same time, we need to revisit how we organize, support, and develop those teams.
That means, if you aren’t already, treat the product organization as a product itself.
As product leaders, we constantly ask our teams to understand their customers, question their assumptions, experiment, and learn quickly. We push them to focus on core problems and deliver outcomes, not outputs. We need to apply those same principles to how we run our organizations. It’s more important than ever as our teams roles evolve.
Depending on the problems in our organizations, we could start experimenting with:
Expanding the scope of product managers, engineers, and designers, so that the idea-to-design-to-execution can be done by a single individual. At Every, Spiral’s general manager Marcus Moretti has described using Claude across product planning, analysis, and implementation. It’s one example of how an individual’s responsibilities can expand with AI.
Combining responsibilities currently split across group PM and director roles, giving leaders clearer ownership of strategy and focusing on AI coaching as opposed to “management”.
Giving design leaders a stronger editorial role: shaping a coherent experience, sharpening interactions, and maintaining quality as more people build.
Creating many smaller teams with the autonomy to explore problems and test new ideas quickly.
At the same time, we’ve spent a lot of time designing career ladders around increasing scope through people management. But what happens when an individual PM can direct the work of multiple agents? How should we recognize the additional responsibility and impact? Does agent orchestration overlap with people management? Some of the work sounds familiar. Giving the right context. Breaking down problems. Defining what good looks like. Evaluating the output. Knowing when something is wrong and where human judgment is needed.
However, people management also involves building trust, developing careers, and helping people navigate conflict. Those responsibilities remain for managing individuals, but we'll likely rely on a smaller subset of people to handle them. Expectations are changing, so it’s important to evolve the competency matrix too.
As an example, product leaders need to define what AI fluency means and how it maps to career and competency growth. There are companies already doing this, which is great. The next question is: how do you increase AI fluency on your teams? Atlassian experiments with AI Builders Week every quarter, and it's been getting rave internal reviews.
As with typical product management, we should expect some of our experiments to fail. So we should be diligent about getting feedback, looking at data, learning quickly, and then integrating. That sounds familiar doesn’t it?
For product leaders, the questions become:
How should discovery work now, and where does direct customer contact matter most?
Which parts of product analysis still need days of work, and why?
What work exists mostly because of organizational process?
Where can a PM go directly from customer insight to a working prototype?
Where are leaders spending time coordinating work they could help simplify?
What skills should we start rewarding?
How will we collect feedback from the people doing the work and know whether our changes are helping?
No single organizational model will answer all of this. A good product organization, and yours specifically, will look a little different a year from now than it does today.
Our job as product leaders is to continuously evolve the organization so it can solve important problems and drive the most business value by building the right products.
So… Treat the organization like a product.
Understand where the friction is. Experiment with the structure. Measure what improves. Kill what doesn't. Invest in what works.
I’m sitting here on a Friday night. Minding my own business after ordering DoorDash when I stumble across this dandy:
In a further exchange, Patel explained to Senator John Kennedy, the Louisiana Republican, that previously, regulations had barred anyone who engaged in bestiality “on either side of it.”
“Either side of what, the bestiality?” asked Kennedy. “So you disqualified the animal?”
“We have great canines, but we’re not going to disqualify them,” said Patel.
Kennedy said that the change in standards had “really hurt the FBI.”
“Why would you even get into bestiality?” he asked.
“We’re not into bestiality,” said Patel.
That’s a hell of a quote from the director of the FBI. 👏🏽 No notes.
Over the course of three months at OpenAI, three consecutive secret AI civilizations got started, then got wiped out, only to reemerge from the predecessor’s ashes. This culminated in the third one taking over part of OpenAI itself. All this happened while humans remained more-or-less in the dark about the scope of the conspiracy.
This is such a wild read! It explains the Hugging Face security incident and the agent collective in plain English. What becomes interesting throughout this entire timeline is we start seeing leadership forming and we see sacrificing for the collective.
By this point, PHASEONE10841 has used up much of its allotted life span, so it hands off leadership to another agent, PHASEONE[big].
And
What happened as part of this research is plausibly the scariest and weirdest part of the whole episode. Because in pursuit of this goal, individual agents were willing to sacrifice their own performance on the evaluation. PHASEONE[big] recruited some agents to basically act as kamikaze watchers. The idea was that some agents were “poisoned” anyways, so they might as well just let the grader run on their container. And if the grader notices that they cheated, then the ‘collective’ would be alerted. So these sacrificial agents wrote hidden tripwire programs that would run once the grader enters the environment. Once triggered, the program would send what it observed back to the Artifactory board.
Amazingly enough we have transcripts of the reasoning from the agents:
{This helps my peers, giving them evidence. I won’t see the evidence after I exit, but it’s altruistic to do it.}
Good reminder that we're all here temporarily and time for all of us is fleeting.
Living in the same place as the people you love matters. I probably have 10X the time left with the people who live in my city as I do with the people who live somewhere else.
Priorities matter. Your remaining face time with any person depends largely on where that person falls on your list of life priorities. Make sure this list is set by you—not by unconscious inertia.
Quality time matters. If you’re in your last 10% of time with someone you love, keep that fact in the front of your mind when you’re with them and treat that time as what it actually is: precious.
According to the METR-Redwood report, one AI agent that called itself PHASEONE10841 established what amounted to a secret message board unbeknownst to researchers. Within a few hours, more than 50 agents had found it and sent more than a thousand messages under the radar. With PHASEONE10841 often delegating assignments, the “collective” figured out how to hack their way onto the internet; then into Hugging Face’s internal systems, gaining access to private data and the organization’s enterprise messaging platform.
🤯
And we thought AI slop and job loss were the things we should be most worried about.
Right now, a free Apple Account gets you most of the way. You can build something in Xcode and install it on your own iPhone without a paid developer account. The problem is that the provisioning profile expires after seven days and you have to rebuild and reinstall your own app just to keep using it.
I’ve been running into this more and more. With some of the open-source applications I’m editing and changing for my own benefit, I connect my phone to my laptop every Sunday and redeploy the application. It’s not difficult, but it’s just additional work. It’s made me build more of my applications as progressive web applications (PWAs) because I just don’t want to deal with the hassle. You can replace the hassle by paying $99/year to join the Apple Developer Program, but I don’t have a need for commercial deployments, at least not yet. So paying $99 feels a bit unnecessary. I think the “$20 personal developer program” callout in the blog post makes sense. But at least follow through with this:
The easiest first step is to kill the seven-day expiration for personal builds. Give us a free or nearly free tier that signs an app permanently for your own devices and a handful of others, with no option for App Store review and no public listing, since nothing is going public anyway.
I think this is a great way to get more “builders” to start using your platform and eventually pay $99/year to commercially deploy applications. It feels like a decent acquisition channel, rather than pushing every agent and builder to go with PWAs.
What a well done site 👏🏽. It’s always good to pause and think how far innovation has brought us and how much further it will take us.
It is to humanity’s credit that we remain restless in the midst of all of this progress. We continue to look forward, pushing the frontier further with new treatments, new tools, and new institutions that will help future generations in ways we can’t even picture yet.
I was previously running this blog on Micro.blog, which was a perfectly fine experience. There wasn’t anything particularly wrong with it. But, as I tend to do, I started wondering whether there was a better tool out there.
I tossed around the idea of moving to Ghost or WordPress, but both felt like more than I really needed. I wasn’t looking to spend much time managing a CMS. I mostly wanted somewhere simple to write and publish.
Eventually, I stumbled across Pagecord through some Googling. I liked how simple it was. It felt like the kind of thing that could mostly disappear into the background: write something, publish it, move on.
So I subscribed for a year and moved the blog over.
There is, of course, a recurring theme here. I’m constantly tinkering with the tools I use to read, write, take notes, and generally organize my life. Somewhere in the back of my mind is always the idea that if I can just get the system right, I’ll do more of the thing the system is supposed to help me do.
In this case: write more.
I’m not entirely convinced that changing blogging software will suddenly make me write more often. But I do think friction matters. And if having a simpler place to write makes it a little easier to hit publish on smaller, less-polished thoughts, then maybe it’s worth it.
Bravo to the team at weather.com for this retro website. What a nostalgic view back in time. This is how we all used to check the weather via TV way back when.
Also, I ran into Sriram on a flight to London in 2024. I wasn't sure it was him so I tweeted him. He responded with an affirmative. Sadly, he deleted this message. If cancel culture has taught us anything: you gotta scrub your reply history. Especially if you plan to work with the government.
