I had coffee this weekend with my good friend Michael Galpert, father to my godchildren. Too much coffee. We talked too much Claw. OpenClaw that is.

Michael has been running around the country organizing ClawCons. New York, Austin, Tokyo next. Not industry conferences. In the halcyon days of Web 2.0, we called them meetups and un-conferences. Sponsors are falling over themselves to get them attached to this new new thing, that borders on madness and hope.

Just people showing up in rooms because they want to talk about what OpenClaw makes possible. Developers, yes. But also small business owners, retirees, students, people who have never attended a tech event in their lives. The energy in those rooms, Michael says, is unlike anything he has seen in years.

I have known Michael long enough to know when he is genuinely excited about something versus just being hyperactive for a second. He is so ‘clawed’ into OpenClaw.

Makes you wonder: what is really going on here? Hope and madness. Whatever the reasons, I am fascinated with the very idea of OpenClaw, how it might be foretelling a future we are not seeing yet.

OpenClaw is a free, open-source AI agent built by Austrian developer Peter Steinberger. It started as a weekend project in November 2025, originally called Clawdbot, went through a rename to Moltbot after a trademark complaint from Anthropic, and landed as OpenClaw at the end of January 2026.

You connect it to whatever AI model you prefer. It runs on your own machine. You talk to it through the messaging apps you already use, including Slack, WhatsApp, Telegram, iMessage, and Discord. And then it does things. Typical chatbots answer questions. It acts. It clears your inbox, makes reservations, tracks your calendar, executes tasks on your behalf while you are doing something else.

By March 2026, it had crossed 247,000 stars on GitHub. Steinberger has since been hired by OpenAI to work on personal agents, and the project has moved to a foundation. The lobster, as they say, has molted. Ironic, the movement’s Jesus has ceded control to his apostles.

Across the planet, everyone is tinkering.

China tech is going bonkers over OpenClaw. On a Friday afternoon in Shenzhen, nearly a thousand people lined up outside Tencent headquarters to get OpenClaw installed on their laptops. Engineers from Tencent’s cloud unit were helping students, retirees, and office workers set it up. The Chinese have their own phrase for it now: raise a lobster, a reference to OpenClaw’s red lobster logo.

Every major Chinese cloud provider has released its own version. Tencent has WorkBuddy. MiniMax has MaxClaw. Moonshot has Kimi Claw. Local governments in Shenzhen and Wuxi announced grants for startups building on the platform. MiniMax, whose model runs inside many OpenClaw setups, has seen its stock rise over 600 percent since its IPO two months ago.

And then Beijing sent a diktat to state banks, government agencies, and the families of military personnel: Clip the Claw.

A thousand people lining up on Friday. A memo from the CCP. Same technology. Same week.

It’s not nothing.

The Mac Mini has become the unofficial hardware of the OpenClaw moment. People are buying the Apple desktop with a single purpose: a dedicated machine to run their agent, separate from their main computer, connected to a free or cheap open-source model. The SF Standard called the Mac Mini bro the new matcha latte guy. Oy vey!

The publication shares the story of Aaron Ng, a 35-year-old AI engineer in the Sunset. He did not want to give OpenClaw access to his own computer. So he bought a Mac Mini just for the agent. It handles his email, tracks updates about his newborn, controls his smart lights. He texts it baby logs because, in his words, the existing apps were terrible. Others are using it to post their bets on Polymarket to Twitter and game the results.

Mac Mini bro! Apple has backed into a cultural moment, without even trying. Jokes aside, this physical object represents a philosophy. The intelligence lives on your machine. You own it. You aim it. No subscription. No permission required.

I checked in with Hiten Shah, one of my close friends who is building hard with OpenClaw. He summed it up in a single line: (AI) Power to the people.

This is why retirees are lining up in Shenzhen. This is why people with no GitHub account are showing up at ClawCons. For the first time, they can feel AI’s intelligence, even if it is not very good. Yet. Not a demo. Not a keynote promise. Not big boys burning billion dollars a month. A thing that actually does things on their behalf. The gap between what you want done and what gets done has always required either your own time or someone else’s labor. OpenClaw makes that gap feel smaller. That feeling, even in its rough and half-broken form, is new.

It has been almost a month since I published How AI Goes To Work. “What OpenClaw shows is how AI will work in the background,” is what I wrote. “And that is what the ‘AI’ future looks like for normal people. Not a separate AI app. Intelligence woven into tools you already use. Doing work you used to do yourself. Or used to hire someone to do, done by software.”

I have seen this kind of excitement before. Many times. That is what keeps me around in Silicon Valley. The excitement of new technology is my addiction. I saw this with the big poppa, the internet itself. Web 2.0. Social. Cloud. In terms of product, the closest parallel to OpenClaw for me is my first time with WordPress.

WordPress did not invent PHP or MySQL or web publishing. It was a continuation of someone else’s idea. However, it assembled those pieces into something a non-technical person like me could actually use. The gap between having something to say and being able to publish it closed overnight. No developer. No hosting deal. No media company gatekeeping who got a voice. Yes, there were other platforms and other options, but it dovetailed with the rise of open source, the optimism of publishing on the internet, and the joy of sharing and being social.

The best part that most people forget about WordPress is that it was never really about us. It was about the rest of the world that could not afford American software economics. At Forbes.com, we spent millions on bad software, trying to publish professionally. Now you could do it for free. A teacher in Jakarta. A student in Manila. A journalist in Nairobi. Suddenly they had the same publishing infrastructure as the New York Times. That is what the excitement was really about. Not the product. It was the collapse of the permission structure.

OpenClaw has that same siren call. Two hundred dollars a month, or even twenty a month, to Anthropic or OpenAI is not a rounding error for most of the world. The world doesn’t pay $10 for lattes. This is the real barrier. But Chinese-made Qwen runs locally. Chinese-made DeepSeek runs locally. OpenClaw does not care which model you plug in. So someone in Manila or Cairo or Bogota can run the same agentic setup as a San Francisco startup, on a model that costs them nothing. If you want something from Claude or ChatGPT, and you can afford to pay for it, then plug that in too.

WordPress democratized voice. OpenClaw is pointing at something further: action. The gap between what you intend and what actually gets done has always required resources. A team. A budget. An organization. OpenClaw is the first very rough sketch of closing that gap for everyone.

WordPress operated in a forgiving domain. A bad plugin broke your homepage. Mine broke all the time. And there were no plugins. OpenClaw operates in the domain of real-world action. A bad agent deletes your inbox, spams your contacts, creates a dating profile without your knowledge. The blast radius is different. One of OpenClaw’s own maintainers warned publicly that if you cannot understand how to run a command line, this is far too dangerous for you to use safely. That is not a ringing endorsement. But it is also not the point.

OpenClaw is not ready for prime time. The security issues are real and the warts are visible. It is a symbol of what is coming, not the thing itself.

AI can be personal. Not a service you subscribe to. Not a platform you visit. A thing that runs on your machine, serves your intentions, uses the model you choose, and works through the apps you already live in.

In his keynote speech at GTC, Nvidia CEO and founder Jensen Huang described OpenClaw as:

What is Open Claw? It’s an agentic system. It calls and connects to large language models.It has resources that it manages. It could access tools, it could access file systems, it could access large language models. It’s able to do scheduling. it’s able to– do cron jobs, is able to decompose a problem that you gave it into step by step by step. It could spawn off and call upon other subagents. It has IO. You could talk to it in any modality you want. It’s an operating system. I’ve just used the same syntax that I would describe an operating system. Open Claw has open sourced essentially the operating system of agentic computers.

In 2023, Bill Gates wrote that “in the near future, anyone who’s online will be able to have a personal assistant powered by artificial intelligence” and that agents would “utterly change how we live.” OpenClaw, running on a free Chinese model through WhatsApp on a $599 Mac Mini in the Sunset, is the first time regular people can actually feel that becoming real.

That’s what I am Clawing about.

March 16, 2025. San Francisco

The Wall Street Journal recently reported that leadership wants OpenAI, the company, to focus. Seems like a plain old business strategy story. Nope!

First, in more prosaic terms, the all-hands and what was said was indicative of need for focus and urgency. I read it as mild panic stations. Second, step back enough and a clear and complete image should emerge. It reveals the real game being played. It is the grand hand at the big AI poker table. The IPO. Who gets there first, who sets the rules. And who really wins.

The IPO is not about one company. Instead it is about three American AI companies — Anthropic, OpenAI, and SpaceX, which owns xAI. It is about the scale of money to be raised from the market. It is also the urgency to do so. The Economist notes that if all three offer 15 percent of their shares, the combined sum would roughly equal every dollar raised across all American IPOs over the past decade.

It is against the backdrop of a global crisis, and this being a battle for winner take all. And most importantly, with the Middle East busy with bigger problems, the money spigot for American AI might be as tight as the Strait of Hormuz right now. The Gulf sovereign wealth funds that have been backstopping this AI frenzy have other fish to fry, at present.

Public market investors in New York and London will now have to carry the weight. Which means the IPO window is real, it is short, it is now, and it will not stay open forever. In short, this is a three horse race to the IPO market. And OpenAI isn’t the leading pony.

OpenAI was all over the map. Sora. A web browser called Atlas. A hardware device. TikTok-for-AI. All announced with the same breathless urgency, same press release energy, different product each time. Cashing in on the announcement economy.

Simo told staff last week they had to stop being distracted by “side quests.” That is a remarkable word for what was, in practice, an $840 billion company running several unrelated experiments inside itself while its most focused rival ate its lunch.

All startups, big and small, are messy. Some have more disorder than others. The admission of chaos is a way to show recognition, and eventual correction of the problems. The question to me, is why this admission of chaos? And that’s why you need to bring out the popcorn.

Let’s break down the WSJ news report itself. The fact that the Journal “reviewed” the transcript is a giveaway. The Journal didn’t say they had the transcript, or that it was leaked to them. “Reviewed by” is a tell. This is a controlled leak. Nothing wrong with it.

Companies do it often. Big publications like the Wall Street Journal get the scoops and exclusives. This is a game that has been played for the longest time. Every word attributed to Simo, from “side quests,” to “code red,” to the Anthropic “wake-up call,” was chosen for outside consumption, to shape the story.

Don’t get me wrong. While the actual new content of the story is thin, it is still real. The organizational dysfunction is real. The Sora team was housed under research while launching a consumer product — a good sign of that dysfunction. However, the Anthropic “wake-up call” framing is a classic Jedi mind trick. Admit competitive weakness to look clear-eyed to investors and customers, while simultaneously using a rival’s success as the proverbial carrot and the stick.

However, one has to read the story in a larger context. The WSJ report has to be read against other news this week. Reuters reports OpenAI is in advanced talks with TPG, Advent International, Bain Capital, and Brookfield to create a new joint venture valued at around $10 billion that would push its enterprise products through PE-backed portfolio companies across industries. It is forming “Frontier Alliances” with McKinsey, BCG, Accenture, and Capgemini, announced last month. This is a great narrative being put together for a company on a deadline, for a specific audience. OpenAI is great at building narratives. It might back its way into a long-term strategy, but for now, you have to see every piece of news coming out of OpenAI right now as part of a jigsaw puzzle.

There is one audience for this final picture. The bankers and institutional investors who will price the offering. OpenAI needs them to believe three simple things. It has its house in order. It has a real swing at corporate dollars. And that it is ahead in the AI race, both for consumers and corporations.

One thing odd, maybe just to me, is why OpenAI has been stuffing its ranks with former Facebookers who are known to juice growth, find edges, and keep people addicted. They have little background in getting enterprises to buy into a product. Simo herself ran the Facebook app. That organization’s genius is consumer engagement: behavioral hooks, dopamine loops, the relentless optimization of the feed. You can see that in the recent iterations of ChatGPT. It has become such a sycophant, and creates answers and options, that you end up engaging with it. That’s juicing growth. Facebook style.

The next shoe to drop would be news that OpenAI poached some big name executives from say, Salesforce or one of the big boy SaaS companies to take charge of the enterprise push. Right now, OpenAI’s own numbers show that its enterprise business generates $10 billion out of a total annualized revenue of $25 billion. In comparison, Anthropic, the smaller company, is widely regarded as ahead in enterprise adoption.

Why is all this important? Again, as I said, money. I know I am repeating myself. The public market money. That is the only real story. There are three AI companies heading toward public markets. Two have friends in Washington. One has a great satellite launch and Internet business. One has focus. One is perceived as an “all over the map” business.

Zoomed in, there is a reason why Grok and OpenAI are looking to go after the “code” business. Anthropic’s success is good enough proof that AI, despite all the talk of impending singularity and AGI, is all about software — the critical choke point in a digital-first world. That is the business. For now! Not orbital data centers. Not a social video app. Not a hardware device. All these may (or may not) be real money spinners in the future, but for now, you gotta focus on juicing your revenues to go public. Focus is the answer.

Anthropic’s revenue run rate has surged past $19 billion, up from $9 billion at the end of 2025 and roughly $14 billion just weeks before that. Amodei confirmed at Morgan Stanley that $6 billion was added in February alone, driven almost entirely by Claude Code. Revenue doubling in two months. That is the curve that makes a prospectus compelling without a PR campaign to explain it.

The US government has tried to change that calculus. Defense Secretary Hegseth declared Anthropic a supply-chain risk after it refused to give the Pentagon unrestricted access to its models. Altman positioned OpenAI as a bridge-builder while quietly moving to take Anthropic’s vacated government contracts.

Anthropic has no friends in Washington. What it has, in the immortal words of former Microsoft CEO Steve Ballmer, is “Developers, Developers, Developers.” Right now, developers are the ones who actually matter. Despite its somewhat unclear messaging and occasionally awkward public identity, Anthropic’s Claude Code is a product they want to use. Wallet always speaks the loudest.

I won’t be crying over OpenAI. They are doing $25 billion in annualized revenue, with 900 million weekly users, and a CEO who has proven he can bend a narrative to his will.

As I said before. When a grand hand is being played, focus is not such a bad thing. Especially when the billions you desperately need are on the line.

March 17, 2026. San Francisco.

As a photographer who is primarily a writer, I like reading about photography so I can learn from those who create. I do that by reading photo magazines and listening to podcasts. One such magazine is Aperture. They take me out of the noise and bring me right into the core of the work. Photographers I don’t follow, or don’t know. Always a pleasant surprise.

In a recent issue they ran an interview with Christophe Lemaire and Sarah-Linh Tran. I have followed Lemaire since his time at Hermès, so I was genuinely excited to read about how he and Tran think about creating. Lemaire is one of the few fashion brands that earns the word “philosophy.” These are people who think in multiple planes: fabric, cinema, sociology, memory.

The initial excitement soon turned into disappointment. Not with the subjects but with the actual interview, and more specifically with the person doing the interview. The interviewer, Alistair O’Neill, an academic from Central Saint Martins, was so much more interested in impressing the subjects that he could never get out of the way.

He finished their thoughts or volunteered his own references. He mentioned mid-interview that he had recently interviewed someone else, for a book chapter he is writing. He was, in the politest possible terms, more interested in being seen conversing with interesting people than in what those people actually had to say.

This is the interaction I see everywhere. The mutual admiration interview is a genre of its own. You see it in podcasts, in magazine profiles, in the long-form Q&As that cultural institutions produce to signal seriousness. The subject is interesting. The interviewer is credentialed. They agree with each other at length. Nothing is at stake. No one is surprised. You finish reading and feel vaguely full but can’t say what you ate. It is worse on technology podcasts. You know people are doing interviews to accumulate symbolic capital, not insight.

Everything gets so diluted that you lose the actual substance. The comfortable questions. Nothing proactive or one that scratches the veneer. The best interviews I read were in Rolling Stone, Creem, and Spin magazines. The interviewer didn’t guarantee to agree. Not hostile, or antagonistic. Not adversarial in the cable news sense. Just genuinely curious enough to risk the awkward follow-up. To leave a silence instead of filling it with your own reference.

Lemaire said something worth sitting with: “Style has a kind of mystery about it. You see someone in the street passing, and you’re like, who is this person?” O’Neill nodded and moved on to Irving Penn.

This was the heart of the conversation. The real reason for the interview. Sitting in that sentence. Pulling on it. Why does mystery matter? What does it mean to make clothes that resist easy categorization in an era when everything is optimized for instant recognition? What does that cost commercially? How do you survive it?

Instead: Penn.

FFS!

Five Things Not To Do

1. Finish their sentences. You are not the subject. They are. So let them lead. Otherwise you get the answer you expected, not the one that was coming.
2. Volunteer your own references. Every time you name-drop, you redirect the conversation back through yourself. The subject stops leading. You get zero revelation.
3. Mention your other work. Nobody you are interviewing cares about your other work. Unless they specifically ask or comment about it, it just gets in the way.
4. Ask only comfortable questions. Agreement is not a conversation. It is applause. The interesting stuff lives precisely where the conversation gets uncomfortable.
5. Fill the silence. Silence is the ultimate pressure. Pressure produces truth. The instinct to rescue the subject from a pause is the interviewer protecting themselves, not the reader.

What Rolling Stone Interviewers Got Right

The Rolling Stone interview has shaped my approach more than any other style. The long, languid, and learned approach is still my cheat sheet. The contrast with today’s technology podcasts could not be starker.

A few things from those interviews that have long since been forgotten, but I often use as part of my interview toolkit:

Take your time. An interview is not a two-hour session. You need to spend time to learn about your subject. When I met Brunello Cucinelli, I had no plans for an interview. It was a 30-minute meeting. It turned into a day with the man. It is still one of my most well-read interviews. And I have been a professional writer for the best part of four decades. If you saw the movie Almost Famous, the main character based on Cameron Crowe didn’t get just one hour in a hotel room. He got days on a tour bus. Proximity produces honesty. The subject stops performing eventually. But of course, you could be Lex Fridman. You could talk, and talk, for hours and numb your subjects, not to mention atrophy the brains of your listeners who actually listen.

Do the research. It is really important to know your subject’s work. And how they think. The Rolling Stone writers were music fans, but they were experts as well. They knew music, they knew the backstory, and they just had knowledge. Not just prepared questions. That knowledge meant subjects couldn’t hide behind vague eloquence. Charlie Rose did good interviews on television. Whatever you thought, or still think, of him, and there is plenty to think about, he arrived at every interview with what felt like a Ph.D. in his subject. Politicians, novelists, scientists, filmmakers: with the help of his team, he knew how to guide the conversation no matter what. His work and Rose the man are two different things. The lesson survives him.

Follow the contradiction. I recently wrote about the ever-changing arguments of Sam Altman around advertising. It is not unfair to ask him directly: what happened? Why did you change your thinking? When a subject says something that doesn’t square with what they’ve done or said before, pull on that thread. Politely. Persistently. It is an opportunity to explain for the subject, and for readers to learn. Your role as an interviewer is to bring an understanding about the subject to the reader. You are only interesting, and what you know or say is interesting, only if you are able to achieve that.

Let them be uncomfortable. You have to read the John Lennon interviews in Rolling Stone. They are hard to read. That’s the point. Nobody rescued him from the difficult moments. The reader deserved that discomfort. So did Lennon. Comfort is the enemy of truth in any conversation worth having.

Be present, not prominent. I go into every interview curious. I want to come out smarter than when I walked in. And the only way to do that is to listen. Really listen. You feel the great interviewers in the room. You never feel them performing for the room. That difference is everything.

A Report Card Of My Own Work

I am not going to say that I am good at following my own rules. I slip up, but with the right intentions. I have linked to three long-form interviews with Brunello Cucinelli, musician Nitin Sawhney, and writer and culture critic Jenna Wortham. In these interviews you will find moments where I insert myself.

I make a verbose speech with Nitin, and with Brunello I talk about myself more than I should. But if you read the interviews, there is a reason for breaking the rules. I am not trying to get off the hook. What I am trying to point out is that whenever I have inserted myself, it is because I am hoping to unlock and provoke a response. I use this technique not to be antagonistic, but to share my own foibles.

To push, to provoke, to offer a thought they can agree or disagree with. The conversation, at least for me, moves forward. When O’Neill inserted himself into the Lemaire interview, the conversation folded back on itself. It became about the interviewer’s taste, not the subject’s thinking.

That is the only test worth applying. Who is this serving? If it is the reader (or listener) then you are on the right track.

For what it is worth, here is how the Rolling Stone tests apply to three of my own interviews:

Take your time. The Brunello interview is all about taking the time. A 30-minute meeting that turned into a day.

Know the work cold. I have listened to Nitin Sawhney for years, so it was easy to follow and talk about the emotional and musical evolution of his work, and elicit responses.

Follow the contradiction. With Jenna Wortham, I push on the gap between internet idealism and structural racism directly and persistently.

Let them be uncomfortable. Jenna yes. Brunello, probably not as much. He is a philosopher-king who never gets rattled.

Be present, not prominent. The Nitin nostalgia speech and the Brunello ending are the two places where I cross the line briefly. I know it. Now you do too.

March 19, 2026.

Michael Smith used AI to create music, and then used AI to create bots to get the “plays” and took the smartest technology companies, including Spotify and Amazon, who should know better, for about $8 million. He is going to jail for his crimes. It is easy to dismiss this as one-and-done fraud. It is anything but. It is an early warning of how AI will disrupt the systems that power our digital society: how culture gets discovered, how commerce gets directed, and how conversations get shaped.

At present, most of our digital society is powered by tech that is, generically speaking, recommendation algorithms. Spotify’s Discover Weekly, YouTube’s suggestion engine, TikTok’s For You page, Amazon’s product feed, Facebook’s news feed. All of it runs on signals of human behavior. Stream counts, completion rates, saves, shares, playlist adds, clicks, purchases. These represent people making choices. The only way to fake that was to hire a back-office army to do it. This is cultural legitimacy laundering.

Scale can now be had for cheap. Almost free. Who is to say that with AI, there won’t be bots that learn and adapt (agents, in polite company) designed to game the system? What if real artists with real music have this army to goose up their stream counts. The algorithm then promotes them to real ears.

What if it is not just real artists, but AI music being created by music-making software models that get better every month. The algorithm promotes this too. Humans like it, save it, share it, and add it to their playlists. They are generating real signals on top of the fake ones. At what point does fraudulently-obtained popularity become real popularity? There’s no clean line.

Just like Uber was limos for everyone, this is payola for anyone, for $200 a month. Even cheaper, if open-source models have their way. You think musicians won’t do it? Look around. They are already buying bots to inflate their numbers using gray-market services. Smith might be the idiot going to jail, but we are facing a structural collapse of the discovery and taste-making apparatus.

In the case of music, since royalties are countable, you can make a case for fraud. In other arenas, things are not going to be as simple. What gets bought on Amazon, what trends on Facebook, what becomes culturally popular. All of it runs on the same logic as Spotify. Signals of human behavior, gamed by machines. AI is making authenticity optional.

Whether it is Spotify directing culture, Amazon directing commerce, or Facebook directing conversations, I have yet to hear from any of their leadership on how they plan to redesign what they do because of the coming onslaught of fake signals.

Smith used crude tools to steal $8 million. His fraud is the boring version of the problem. The interesting version hasn’t been prosecuted yet. It may never be.

Further Reading

It has been a week of inference. I started the week writing about OpenClaw and its growing popularity. And I end the week with a recap of Nvidia’s GTC event. A lot has been written about the event, especially about Nvidia CEO Jensen Huang’s big claims of quadrupling his revenues to above a trillion dollars, the question that wasn’t asked, why?

Sure, we can dismiss this as a bluster of a bull-market darling, but in reality, if we believe that AI is going to be the new way of interacting with information (as I do) and will impact all sorts of industries, then you have to give his claim some thought.

As I explain in my breakdown of Nvidia GTC, Jensen’s Trillion Dollar Token Factory, for CrazyStupidTech, his boast makes sense for two reasons. First, this is indeed an inference inflection. What that means is that AI goes from being in the “capex” phase to being in the “opex” phase. That key distinction is why you see Anthropic adding a billion or so in revenue every week or so. And that is why OpenAI is in a tizzy. And that is why Elon Musk has made “coding” a top priority for his Grok.

I am hoping to shift gears next week and write longer essays, eschewing news and daily goings-on. I have a few personal pieces that remain unfinished, including reviews of a new photography geegaw and of course the MacBook Neo.

Till next week. For now, enjoy these lovely articles.

• Why Tech Bros Are Now Obsessed with Taste. Kyle Chayka in the New Yorker on Silicon Valley’s latest borrowed virtue. Silicon Valley is a taste dead zone. I have written about this many times. Now the same people without taste are arguing that because AI levels the playing field, and anyone can build anything, so the only edge left is knowing what’s worth building. That’s “taste.” Don’t look for it in these parts. Chayka makes my point without me losing my hair. [New Yorker]
• We Have Learned Nothing. Jerry Neumann is my favorite venture capital and startup writer. None better. In his latest piece he makes a point to remind us that the whole edifice of the science of entrepreneurship is just that, an edifice. So, to paraphrase Steve Jobs, think different. And do different from whatever the current advice says. [Colossus]
• In Search of Banksy. Reuters spent a year using what they call a “reverse forensic” approach to unmask the world’s most famous anonymous artist. Wow. What a great story. I mean, make this a documentary. The journalism is meticulous. [Reuters]
• Russia’s Self-Inflicted Communication Crisis. Politico maps the race to cut Russia off from the global internet. Telegram, WhatsApp, YouTube, Signal, Discord, X, FaceTime, LinkedIn, all blocked or throttled. Russian President Putin signed a law letting the FSB shut down cellular and internet access on demand. The twist is that the Kremlin’s own military depends on Telegram for frontline communications and crowdfunding. They’re destroying the infrastructure their war machine runs on. What an amazing story. [Politico]
• The Repricing of Time: Equity in the Age of Agents. Jordi Visser makes the financial argument for what many of us feel intuitively. AI doesn’t just disrupt business models. It compresses time. When competitive half-lives shrink from a decade to months, equity stops representing ownership of a durable franchise and starts resembling a call option on execution velocity. This is the best thing I’ve read on what AI actually does to market structure. I wish I had written this piece. [Visser Labs]

Things I wrote this week, ICYMI:

• Lobster Boil
• OpenAI Has New Focus (on the IPO)
• How Not to Interview Interesting People
• Why Fraud Is the Boring Problem
• Jensen’s Trillion Dollar Token Factory

In Memoriam.

Miles Rose, who passed away on March 1, 2026, in Puerto Rico. He was 72. And he was a friend from the Internet’s halcyon days and ran SiliconAlley.com. I lost touch with him over the past decade or so, but he was and will remain a cherished friend and a reminder of happier, more innocent times in the evolution of the internet. Rest in peace, Miles.

Len Deighton, one of my favorite authors and a master of the spy genre, passed away. I enjoyed this recap of his work. In the end, this is the best way to remember someone who made words his everything.

March 22, 2026. San Francisco

When it comes to OpenAI, smart money is starting to do the math out loud. And something doesn’t add up. On surface, today’s news that OpenAI is offering 17.5% guaranteed returns to private equity firms looks like a shot at the Anthropic threat. Scratch the surface, and you start to see the story behind the story.

The PE deal is the kind of deal you do when you’ve borrowed against the future and the future is taking longer than expected.

Remember a few weeks ago when Nvidia CEO Jensen Huang (one of the backers of OpenAI) said that OpenAI was not a well-run business. Now Thoma Bravo founder Orlando Bravo is saying it out loud. Bravo walked away from the JV idea after questioning the long-term profit profile. It is not a coincidence that two smart money operators are arriving at the same conclusion. Just different words.

As I wrote in yesterday’s newsletter, (and in my breakdown of Nvidia GTC, Jensen’s Trillion Dollar Token Factory, for CrazyStupidTech) we are at an inference inflection. What that means is that AI goes from being in the “capex” phase to being in the “opex” phase. That shift is why you see Anthropic adding a billion or so in revenue every week or so. And that is why OpenAI is in a tizzy. And that is why Elon Musk has made “coding” a top priority for his Grok. All three companies are about “business” and revenues and not just promises.

OpenAI is good at promises. Especially the promises of a limitless future. And it has done a good job of borrowing against that future. Except, that future isn’t arriving or arriving fast enough.

It announced the $110 billion funding round, while only about $25 billion was near-term committed capital. Stargate was a $500 billion infrastructure promise. Oops. Each of these moments followed the same logic. Project confidence so large that the world treats it as reality. The reported PE deal is just the latest version of the same story.

The question to ask is why is OpenAI willing to guarantee 17.5 percent return? What gives. My harsh read of the situation is that the company needs money, and private equity is the last large pool of capital it can tap. And that tells you something about how many other doors OpenAI has already knocked on.

Here I go again with more questions than answers.

When I look at the news about the joint ventures, I read it as OpenAI cannot afford the cost of selling to big companies on its own. Deploying engineers to customize models for clients is expensive, slow, and margin-destroying. The joint venture shifts that cost onto PE firms while they think they’re buying access. They’re not. They’re picking up the actual cost of the work, dressed up as a partnership. It is a smart move, especially when you are preparing for a public offering.

Here is the line buried in the Reuters report that you should read again and again. The joint venture structure would provide “clearer segment reporting that can support the IPO narrative.”

Loosely translated (and financial analysts would be better to break this down), the JV structure allows OpenAI to create a line item that makes it look like it has recurring enterprise revenue, which is exactly what IPO bankers need to justify a valuation of a trillion dollars. The real product here is not AI. It is an IPO prospectus.

Maybe I am completely off base. There is almost certainly something we don’t know. There might be a product breakthrough that changes everything. OpenAI has surprised before.

However, it is hard to ignore that both Jensen and Bravo raised issues around OpenAI as a business. After all, no healthy business needs to offer terms that OpenAI is offering. You know why this is happening. It is because the need for cash is real. And the need to go public and get that cash is even more real. That 17.5 percent guarantee is just a magic number in magic math.

Previously on this topic:

• OpenAI Has New Focus (on the IPO)
• The Great AI Game vs AI Theater
• $110 Billion in Name Only
• Mad Money: The Big AI Race
• The Debt Beneath the Dream

Rudyard Kipling was right!

“San Francisco has only one drawback — ’tis hard to leave.”

I thought I would be here for two years, maybe. And then go back to New York. And here I am, still here. The city has given me more than I have given it. The rush of a modern gold rush. Friendships. Little chance meetings that transformed who I am as a person. People have come, people have gone.

Booms. Busts. Booms. Busts. Booms.

Still here!

Charles Moore, School of Architecture, UCLA, said that “one of the basic human requirements is the need to dwell, and one of the central human acts is the act of inhabiting, of connecting ourselves, however temporarily, with a place on the planet which belongs to us and to which we belong.”

23 years later, it looks like I have found that place. I don’t yearn to leave.

23 years ago my biggest regret was missing opening day at Yankee Stadium. Today, the Yankees are opening their season at the Giants’ stadium. Everything comes full circle.

March 25, 2026.

San Francisco as my “dream state.”

For thirty-odd years, broadband has been sold as a one-way pipe. The download speed was all that mattered. Bigger the number, better. No one cared about the upload speeds, because the internet was mostly something you consumed. Streaming video, loading pages, pulling files from the cloud. Upload was an afterthought.

All that changed when the pandemic came around, and with it came how we used and interacted on the network.

And if you use upload as a core part of the internet experience, cable broadband is losing again. Not only to some of the fiber providers, but to the upstart, tiny rivals, those pesky municipal broadband providers. They certainly have come a long way since I went to Truckee in 2004 to write about their nascent muni-broadband network being put in place by the local utility.

A new report from Ookla, studying 13 months of Speedtest data from 14 of the largest municipal broadband providers in the U.S., makes that case plainly. Eight of those 14 municipal ISPs beat their local competitors on median upload speeds.

In 2004, I had a core thesis: broadband would decentralize opportunity (it has,) let entrepreneurs build from anywhere (and they do,) and break Silicon Valley’s grip on who got to participate in the tech economy, and it has. I believed it. The municipal networks being built in places like Cedar Falls, Iowa and Truckee, California felt like the capillaries of a new nervous system.

Of course, thesis and reality are two different things, as I have learned time and again.

It didn’t quite happen that way. The infrastructure spread, but the change didn’t happen. At least not at scale. Startups still clustered in San Francisco and New York. The dream of Bend, Oregon and Overland Park, Kansas as legitimate alternatives to the Valley remained a pipe dream. It still is.

What finally forced the shift wasn’t better pipes. It was a pandemic that left people with no choice but to work from wherever they were. Twenty years of broadband expansion did what it couldn’t do alone: it became the foundation for a behavioral change that only a crisis could trigger. And when that crisis hit, it also changed the need for speed, in a direction no one saw, or bothered to even think about. The need for better upload speeds has become the new need for speed.

Overnight, our typical internet usage changed. Upstream video is now part of everything from going to school, talking to doctors, and shopping. We don’t even think twice about FaceTiming, nary a thought about what network we are using, Wi-Fi or 5G.

According to the latest OpenVault Broadband Insights report, upstream usage surged 21.7% year-over-year in 2025, the fastest-growing metric. In a side-by-side comparison of fiber and DOCSIS subscribers, fiber subscribers with symmetrical 677 Mbps service consumed 93 GB of upstream bandwidth in Q4 2025. DOCSIS subscribers provisioned at an average of 17 Mbps upstream, used 56 GB. That’s a 66% gap, on identical customers, explained almost entirely by what the network allowed them to do. OpenVault concluded that latent upstream demand already exists on DOCSIS networks. People will use the pipe if you give it to them.

Ookla’s data shows what consumers want. Connexion, the municipal provider in Fort Collins, Colorado, delivered over 300 Mbps upload speeds. Comcast’s Xfinity, its main competition in that market, started at 28 Mbps upload and, to its credit, pushed that to about 99 Mbps by December 2025. Still less than a third of what Connexion delivers. In Georgia, OptiLink, another municipal provider, clocked upload speeds in the 230 to 250 Mbps range against Spectrum’s low-20s.

Why is municipal broadband winning on upload? They built new fiber networks. Fiber is symmetrical by nature. The cable industry spent twenty years architecting its networks around download-heavy TV and streaming. Now it is scrambling to retrofit with DOCSIS 4.0, mid-split upgrades, and high-split architecture to claw back the upload headroom it gave away by design. Comcast and Charter are spending real money on this. They are playing catch up.

Municipal providers aren’t dominant when they face off against private fiber competition. AT&T Fiber is competing with LFT Fiber in Lafayette, Louisiana, pulling 473 Mbps down and 424 Mbps up against LFT’s 113 Mbps and 107 Mbps. GFiber in Utah is outperforming UTOPIA on raw speed metrics. The advantage municipal providers have is specifically over the incumbent cable monopolies.

Cable broadband has ruled the roost for so long, but it is facing all sorts of competition. Not just from muni-broadband, but from other emerging technologies.

Fixed wireless started stealing cable’s lunch, though it seems like it has had its boom moment. Each of these shifts had one thing in common. An infrastructure change forced behavior change. And behavior change exposed infrastructure’s limits. This is the unending yin-yang of any and all technologies.

Communities that bet on municipal fiber, often against fierce lobbying from the cable industry, are sitting on a quiet infrastructure advantage. Most of those cities built symmetrical fiber because it was the technically correct thing to do, not because they predicted an upload-heavy future. They got lucky. Or maybe being technically correct is its own kind of luck.

The cable industry will catch up eventually. DOCSIS 4.0 is real and the upgrades are happening. But “catching up” is not the same as being ahead. And in infrastructure, being first matters more than being right on paper.

It might have started with Zoom in the pandemic, but we are now going to be living with applications and services that will see us transform from being content consumers to data sources. Live commerce, cloud gaming, AI voice and video agents, and spatial computing are already starting to tiptoe into our digital life. It is hard to predict what is the next big thing. After all, video conferencing needed its pandemic moment to become a mainstream behavior.

March 25, 2026. San Francisco

Pop some popcorn. Put some butter. Add some salt. Because opportunists (politicians) are pointing their muskets at villains (tech bros), using children’s welfare as the ammunition. In case you were wondering, I am talking about the battle between New Mexico AG and Silicon Valley’s villain in chief.

The next bout is on May 4. So mark your calendars. Why?

This week two verdicts came in quick succession. First, a New Mexico jury ordered Meta to pay $375 million for knowingly enabling child predators on Instagram and Facebook. Then, a Los Angeles jury found Meta and YouTube negligent for designing platforms that addicted a young woman who first used YouTube at age six and Instagram at nine.

On the surface this is big win for ambulance chasers. However it could be much bigger if politicians actually have the best intentions that go beyond winning the next elections. History tells me, they are what I said they are, opportunists.

Let’s be honest about what New Mexico AG is. He is an elected official who identified the most despised man in American technology, a category that has never been more unpopular, and filed suit. Zuckerberg is the perfect patsy. Ian Fleming couldn’t have created a more perfect caricature of a villain for our post social age. Weird. Awkwardly dressed to appear cool. Black AR glasses.

Rich, arrogant, visibly indifferent to the damage his products caused, and now on record in two courtrooms defending decisions his own safety teams told him were harming children. The internal documents that came out in both trials showed the same pattern: engineers and safety researchers raising alarms for years, recommendations made and ignored, Zuckerberg choosing growth. None of this is new. I wrote about Facebook’s growth-at-any-cost DNA back in 2018. The pattern has not changed. Only the courtroom has.

New Mexico AG smelled blood. A politician at the end of the day, doing what politicians do. Scoring points against the current villain in chief, building a reputation on the wreckage of someone else’s failure, and calling it justice. The “independent monitor” he is asking for is the tell. And that is not how change happens. That is a patronage job dressed as accountability. That said, underneath the political theater, the structural demands are real. And if a judge grants even a portion of them, they could change daily life for two billion people.

Think about what your daily experience on Instagram or Facebook or TikTok or YouTube actually is. You think you made some choices. Pat yourself on the back. You have been fooled into thinking that you were the one making the choices. Now, let’s get real.

You open the app and a feed appears. Like I did. This morning. I went to Instagram to see what my “pen friends” were doing. I didn’t see a damn thing. What I saw, I didn’t choose. Neither do you. You see what they want you to see.

An algorithm shows you what it wants to show you, trained on everything you have ever paused on, liked, or watched past the halfway point. It is the ultimate illusion, where perception is conflated for reality. It has one job: keep you there. The most reliable way to do that, as the internal research showed and as the companies knew, is to surface content that provokes anxiety, outrage, desire, or envy. Those are the emotions that generate engagement. Engagement brings in the moolah. The loop is the product. I called this trap back in 2018. The feeds have only gotten more sophisticated since.

Layer in the design features built around that loop. Infinite scroll, so there is never a natural stopping point. Autoplay, so the next thing starts before you have decided you want it. Notifications calibrated to pull you back at the moment your attention drifts elsewhere. Variable reward, the unpredictable appearance of a like, a comment, a new follower, built like a slot machine. None of this happened by accident. All of it was tested, measured, and optimized.

This is what the courts are now calling a design defect. And to understand why that matters, you need to understand the wall that has protected these companies for thirty years.

Section 230 of the Communications Decency Act, passed in 1996, is the founding legal pillar of the modern internet. It says platforms cannot be held liable for content posted by their users. It is why Facebook is not responsible for a defamatory post, why YouTube is not liable for a radicalization video, why the whole architecture of user-generated content was able to exist at scale. Every time someone tried to sue a social platform for harms caused by what users posted, Section 230 was the wall. Case dismissed.

The Los Angeles plaintiffs found a door in that wall. The jurors were specifically instructed not to consider the content Kaley saw on the platforms. Not because it was irrelevant to her harm, but because content is where Section 230 lives. As NPR reported, the lawyers pursued a case of defective design precisely to get around the high bar set by Section 230.

The entire case was built around the design. Infinite scroll, autoplay, algorithmic feeds, notification systems. Those are not user posts. Those are the company’s own engineering decisions, made in their labs, tested on their servers, optimized by their employees. CBS News put it plainly: this case centered around how the apps are designed, not the content itself. Section 230 does not protect you from your own product choices.

The jury answered yes to seven specific questions per defendant. Were they negligent in the design or operation of their platforms? Was that negligence a substantial factor in causing harm? Did they fail to adequately warn users? And did they act with malice, oppression, or fraud? That last finding triggered punitive damages.

The pipe (aka the Social Media platform), not what flows through it. That argument won and opened new legal ground that can change everything that comes next.

If the design is the defect, the remedy is redesign. Not a warning label. Not a terms of service update. The product itself has to change.

Not surprisingly, I have some suggestions. Actually pretty simple ones that even politicians can grok.

Start with defaults. Right now every addictive feature ships on by default. Flip that. Ship the product in its least addictive form. Chronological feed. No autoplay. Scroll that ends. Notifications off until you choose otherwise. Every addictive feature requires an explicit, informed choice to activate.

That is consent architecture. And it is exactly what these courts have been asking. Did you warn users? Did you get meaningful consent? The answer, established now by two juries, is no.

Real age verification is the other demand with teeth. But it is also a hairball that will unwind a lot of the internet, security and privacy. That is for experts with more gravitas and knowledge than me.

Instagram’s Adam Mosseri knows what is coming. His year-end memo was full of language about authenticity, credibility signals, and platform responsibility. The same language is now showing up in court filings. At trial, he testified that there is always a tradeoff between “safety and speech.” That is a telling formulation. Safety as a constraint on the product, not a foundation of it. When I pushed on what Instagram is really positioning itself to become, he pushed back. But the architecture he described, who posts rather than what is posted, is exactly the consent architecture these courts are now demanding.

The practical problem for the platforms is simple. You cannot run two different algorithmic systems in one product without everyone noticing. Once you build the less addictive version for minors, the question becomes impossible to avoid. Why is that version not the default for everyone? Why is the more addictive version what adults get, when they never consented to it either?

The legal fight is about children. The logic does not stop at eighteen.

The tobacco parallel is everywhere in the coverage this week. It is the right one, but people are using it loosely. The important part of the tobacco story was not the $206 billion settlement. What changed behavior was what came alongside the money: mandatory disclosure of internal research, advertising restrictions, forced redesign. Joe Camel did not die because of a fine. Joe Camel died because a court said you cannot market this product to children, and you cannot keep making it more addictive than it already is.

The companies held out for decades. Denied the science. Attacked the researchers. Then they did not.

But here is where the parallel gets uncomfortable. Tobacco did not go away. It became Juul. I wrote about Juul back in 2018, when Silicon Valley was celebrating a $15 billion e-cigarette company as a tech startup. I called it Camel 2.0 then. Same addiction, new delivery mechanism, new investors, new PR. As a former smoker who nearly lost his life to cigarettes, I was livid. Tobacco companies figured out how to sell the same addiction in a cuter package, and Silicon Valley helped them do it.

Zuckerberg will do the same. He is too good at survival not to. The feeds may change form. The addiction machine will find a new name. Addiction is a lucrative business. It always finds a way.

Meta and YouTube have already announced appeals. They have the money and the lawyers to fight this for years. And they will. But discovery in each new case keeps producing documents. The internal research keeps surfacing. The bill keeps growing. Eventually it is cheaper to change than to fight.

Zuckerberg’s doctrine is simple: two steps forward, one step back, and make it appear as if he is changing. His superintelligence memo last July was the latest example. Talk about the future, avoid the present, and let the lawyers handle the rest.

That is the logic of the crowbar.

All eyes on Santa Fe on May 4. There just might be enough structural change. And if consent architecture is put in place, then we can all feel the glow, for a moment. Billions of people deserve a different default. Till the battle starts again. Like Juul.

March 25, 2026, San Francisco

PS: I want to apologize for sending two long emails in a single day. I had no intention of doing so. I was definitely not going to write about Meta or social media. I was thrilled to just nerd out about broadband today. However, I can’t help myself when it comes to Facebook and Zuckerberg and their shenanigans. I have been covering them since they were a tiny startup.

I have been a writer long enough to know this in my bones: jargon sucks. It is a crime against writing. It abuses your readers and your listeners. And it is a sure sign you do not know what the heck you are talking about. If you cannot explain something in plain terms to your mom, you do not have a clue. You are reading from a script, pretending to be master of the universe.

And yet.

Listen to any podcast. Scroll any LinkedIn post. Skim any Twitter thread. You are drowning in it. The calling card of the clueless and the pretentious. Jargon!

And it is the lingua franca of corporate America, as new research shows.

Shane Littrell, a cognitive psychologist at Cornell, has created the Corporate Bullshit Receptivity Scale (CBSR) to measure how susceptible people are to what he defines as “a semantically empty and often confusing style of communication that leverages abstruse corporate buzzwords and jargon in a functionally misleading way.” Now that sentence itself is a perfect example of jargon. Littrell should follow his own advice.

Jokes aside, his research shows that many people cannot tell algorithmically generated nonsense (sentences assembled from real corporate vocabulary but stripped of meaning) from actual quotes from Fortune 500 executives. Littrell surveyed more than 1,000 office workers.

Ouch!

Like I said, jargon is the clutch of the clueless. The research shows that individuals who score higher on “corporate bullshit receptivity” perform worse on work-related decision-making tests. They see their boss as a visionary transformational leader, inspired by empty corporate mission statements. People who are receptive to this stuff tend to produce more of it. They are not cynics playing the game. They actually believe the words work.

Now look around Silicon Valley. We have our own dialect and it is getting worse. Tech bro slang these days ranges from “vibe coding” to “agentic workflows” to “founder mode” to “10x” to “unlock.” With every year, I hear more and more MBA speak, with extra reverence for whoever invented the latest term.

Littrell’s research draws a line between jargon and bullshit. Jargon can be useful when it carries real meaning for people who share context. Bullshit mimics the structure and rhythm of jargon but is empty on purpose, or by accident, or because no one bothered to check. Sadly, even well-meaning jargon meanders into bullshit. “Product-market fit” once described something real.

Illustration made with ChatGPT

I have been a WebPass customer for years. Fast, reliable, founder-run. When there was a problem, you could reach someone who gave a damn. It was the kind of internet service that made you forget you were dealing with a utility.

Then Google bought it.

That should have been the warning sign. Google buys things it should have no business touching. It launches products without a plan. Its decisions to get into new markets are a map of some mediocre executive’s ambition.

I watched the service slowly get worse. More outages, stagnant speeds, the kind of indifference that sets in when a product becomes a line item rather than a mission. Like it was for Charles Barr, founder of WebPass. When Barr was running the show, the speeds went from 50 Mbps to 100 Mbps to 200 Mbps. All for less than $50 a month. Eventually the speed went to a gig per second. It was one of the fastest in the country.

Since then, as competitors like Sonic were pushing 10 Gbps across San Francisco, my WebPass connection sat frozen at 1 Gbps. The founder used to pick up the phone. Now I was just another entry in a spreadsheet at Alphabet, buried inside “Other Bets,” a division whose name tells you everything about how seriously Google took it.

Google is a company where nobody has real product conviction anymore, just bonus targets and quarterly metrics. It wears a veneer of innovation, but in reality it is a financially optimized extraction machine run the McKinsey way.

Anyway, now comes the news I didn’t want to hear, but knew was going to happen anyway.

Google has sold the whole thing, Google Fiber (including WebPass), to Astound Broadband, owned by Stonepeak, a private equity group. They are calling it a “merger.” It is not a merger. A merger implies two equals coming together to build something. This is a flip. Google wanted off the hook for a money-losing division, and they found a buyer. Customers are just assets in the transaction, transferred like furniture.

Here is the most telling detail. So far, Alphabet has not filed an 8-K with the SEC as part of the announcement. An 8-K is what public companies file when something material happens. Google’s lawyers apparently concluded that losing your internet provider did not rise to that level. GFiber sat inside “Other Bets,” a segment that in 2025 generated $1.54 billion in revenue across everything Alphabet deemed non-core, while losing $16.8 billion. GFiber itself was less than half a percent of Alphabet’s total sales. To their investors, this was a rounding error. To you, it’s your internet connection. That gap tells you everything about where you stood in the relationship.

Then there is Dinni Jain, the CEO of GFiber who will now lead the combined company. His story is presented as reassurance, a steady hand, a cable industry veteran, a man with a Vermont farm who talks about patience and humility. A banker turned cable guy turned Google, and humility. Lolz.

But look at the actual career. He started at NTL, a UK cable company that filed Chapter 11 in New York in 2002. At the time, its $10.6 billion bond default was the largest in US corporate history. He left a year before the collapse. Moved to Insight Communications, a cable operator largely owned by the Carlyle Group, which Time Warner Cable bought for $3 billion in 2012. He became COO of Time Warner Cable, which in turn was acquired by Charter. Every company he has run has been consolidated, absorbed, or flipped. He made enough money to retire to a farm in Vermont in search of humility. And pigs have wings too.

He was hired as Google’s third CEO “out of retirement” to run the fiber unit in 2018.

To stabilize a business in freefall.

He did real work. Under Jain, GFiber cut support calls in half and moved to ten-minute technician appointment windows. In 2024, he hired GFiber’s first-ever CFO. That tomTomed the real plan. You don’t hire a CFO to run an internet service. You hire a CFO when you are preparing to sell. He stabilized GFiber well enough to make it sellable. And now it is being sold.

Let me tell you what Astound actually is, because the name was invented to obscure it. And while we are on the subject of names, notice that Google rebranded “Google Fiber” to “GFiber” before this deal closed. Scrub the Google brand, hide the Astound connection, give customers nothing to search for. That is a classic private equity move. Polish the turd before you sell it.

The only reason I know about Astound is because when I lived in New York I signed up for an upstart called RCN. It wanted to take on Time Warner Cable. And Verizon. And AT&T. I like upstarts. My whole life is betting on the underdog. Monopolies suck, destroy innovation, and throttle progress. RCN convinced me that my apartment needed them just by existing. Sucker, that was me, because in a few months I found that what seemed like startup gold was spray tan on a tin can.

I moved to San Francisco. And because I am like a jilted lover, I kept tabs on that shit show. Being a broadband nerd made it simple. So now you know why I am not “astounded” by Astound. I know the genesis.

In February 2017, private equity firm TPG bought RCN, a cable operator in New York, Boston, Chicago, and DC, for $1.6 billion, and simultaneously bought Grande Communications in Texas for $650 million. A year later, TPG added Wave Broadband on the West Coast for $2.36 billion. Then came the bolt-ons. enTouch in Texas, Digital West in California, WOW!’s Chicago and Maryland markets for another $661 million, Harris Broadband in central Texas. In January 2022, they stapled all of it together under a new name, Astound. The name came from a small Bay Area provider Wave had previously acquired.

In August 2021, before the rebrand, Stonepeak bought the entire platform from TPG for $8.1 billion, $3.6 billion in equity plus $4.5 billion in assumed debt. That is a leveraged buyout. The debt sits on the balance sheet and has to be serviced from operating cash flow, which means it has to be extracted from customers.

Last July, Astound had to refinance. Stonepeak injected another $400 million, pushing debt maturities out to 2029 and 2030. They called it good news. What it actually was: a company under pressure, buying itself more time.

Now add GFiber and WebPass to that debt stack.

The playbook is not complicated. Low introductory pricing, step-function rate increases after twelve months, retention discounts for customers who call to cancel, and underinvestment in everything that doesn’t show up on a cash flow statement. The BBB and Trustpilot reviews are a graveyard of billing complaints and broken promises.

Astound makes the old Comcast look almost saintly. At least with Comcast you knew what you were getting. Astound is as dodgy as there are angles in a tapeworm. Google Fiber’s reviews on Yelp are already failing grades on billing, customer service, support, and performance. Under a PE regime, it is only going to get worse.

Game is the game.

And debt is the game PE plays. Stonepeak has billions in debt to service. That debt requires predictable and increasing cash flow. The easiest way to get more cash flow from a captive broadband customer is to raise prices. There is no mystery here.

What angers me most is the language. The GFiber email I received, and I suspect you got one too if you’re a customer, promised that “nothing is changing about your service.” Not the speed. Not the price. Not the “extraordinary customer experience.” This is the kind of sentence that only gets written when everything is about to change. Companies actually keeping their promises don’t need to make them this explicitly.

“Nothing is changing” is corporate for: we need you to stay through the deal close.

The regulatory picture offers little comfort. The FCC under Chairman Brendan Carr views consolidation favorably. The current commission is not going to save you. California’s Public Utilities Commission is the one real lever. The CPUC has actual teeth. It forced Verizon to make concrete commitments before approving the Frontier acquisition.

Citizens can engage that process. Advocacy groups can push for price locks, service standards, build-out requirements. It won’t block the deal, but it might constrain the worst impulses of a PE-owned operator trying to wring cash from a captive customer base. Given the demographic profile of WebPass customers, none of the consumers are going to complain. They will just pay-up.

Here is the worst part. There is a better option in San Francisco. Sonic.net. And no, it’s not available in my building. A lot of tech people are soon going to find out that Google has sold their internet experience to a terrible company, with a patchwork infrastructure and a poor customer track record.

Let me be specific about that infrastructure. Astound is not one network. It is five or six different networks, built at different times, by different operators, using different equipment, stitched together under a single brand with a management layer on top. As of today, 86% of Astound’s footprint is still HFC, the old hybrid fiber/coax cable architecture from the 1990s. Only 14% is genuine fiber-to-the-home.

Their broadband upgrade path runs through a Harmonic virtualized platform that can theoretically be migrated to fiber someday. Notice the word “theoretically.” Their mobile service rides T-Mobile’s network. Their TV runs on TiVo, distributed through a DirecTV streaming partnership. Their WiFi hardware is Amazon eero. None of these are bad choices individually. Together they describe a company that has outsourced every layer of its stack to someone else.

Google Fiber changes the mix somewhat. GFiber brings about 2.8 million locations passed across 15 states, approaching 1 million subscribers, all on genuine fiber. Markets include Kansas City, Austin, Atlanta, Nashville, Salt Lake City, and a recent expansion into Las Vegas. WebPass adds fixed wireless and ethernet connections in dense urban markets. That is a real fiber asset being dropped into a predominantly HFC platform. Whether the combined company upgrades the HFC side to match, or lets the GFiber advantage erode to fund debt service, is the question that determines what kind of company this becomes. History suggests the latter.

The founding promise of WebPass was simple. Fast internet, honest pricing, people who answer the phone. Google broke that promise through neglect. Astound will break it by design. These are two different kinds of failure, but the customer ends up in the same place.

In the shit.

March 27, 2026, San Francisco

A few days ago, the Day One journal app gave me a prompt: what is the one word that would describe you. That made me think hard. I was thrown for a loop.

I have always struggled to describe myself. Not sure how others see me versus how I see myself. This is not the first time I have had to confront this question. As a child and as a young man, I dealt with this same challenge.

I have been thinking about this for a few days. It is hard to use one word to describe a whole person. It is a strange way to think of yourself. I came up with many descriptors, but they are not the whole thing. I knew that already. Still, I wondered why they were the fragments that I chose to put down on paper.

When I offered my fragments to Claude, it pointed out that the underlying theme to them all, the one that ties it all together, makes me a “seeker.” And almost instantly I realized that’s the word that describes me best, more than anything else.

I have always believed that you need others to see you better than you see yourself. Just as I am able to see, learn, and appreciate others better. In this specific case, Claude found an underlying correlation.

Over the years I have amassed many fragments of self. The phrases I ended up using to describe myself.

Curious. Interesting. Sarcastic. Optimistic. Cool. Forever young. Worry wart. Uptight. Indecisive waffler. Taste maker. Curator. Never finish. Early adopter. Careless.

Curator and Taste are real. But they are outputs. Descriptive of what I produce. Seeker is the reason I am. The AI pointed out that most seekers are better at outward motion than the inward one. Or maybe the AI was just doing what it is trained to do: be a sycophant, aimed to please, saying what you think you need to hear.

But, I do trust my own view of things. And of me. More than AI, or more than any other person. I just lack the vocabulary to describe myself for myself. Words are very important, but when it comes to the self and labeling myself, they have failed me.

Because words are your salvation, your reason to be, as a writer you feel your verbal shortcomings more acutely.

“We know more than we can tell.”

Michael Polanyi.

And while Polanyi was writing about science, it applies here too. Charles Taylor, in Sources of the Self, points out that articulation isn’t just description. It changes how you see what you know. The gap between awareness and articulation is something we don’t think about enough.

But we should.

As I came closer to my own fragments, I realized that the glue that holds them together is a fundamental quality I don’t even think about. Caring.

Curiosity means I care enough to dig deeper. My sarcasm is a mask for caring enough to be disappointed. I worry because I care too much to let go. An indecisive waffler? Maybe I don’t want to get it wrong. You get the point. The fragments point to just one thing. And I didn’t even realize it. It took a long few days of introspection to even come to this realization.

Henri Bergson, the French philosopher, had it right: language chops up something that was never meant to be fragmented. I suppose that’s where it all started. Where my fragments of self managed to hide the one word that describes it all.

Maybe because you are too close to yourself. Too clouded to see clearly. And that’s why you have to go outside to get a better perspective. Or maybe that’s the journalist in me. A larger perspective, a bigger context, a lens that’s not so close.

Weirdly, this translates in my photography as well. I find beauty in a landscape through its contours and its outlines, not in its details. And even when I get close, I always seek the essential.

So maybe “Seeker” is the best descriptor. What fuels the seeking is that you care. I care about a lot of things. Not sure why. But I do. Maybe I will never know. Maybe that’s the point.

Ancient philosophers across all traditions, Western, Zen, or Vedic, point out that seeking is noble. But the seeker has to make peace with the idea of never arriving.

I think I am okay with that.

March 28, 2026. San Francisco

The news cycle has been moving at hyperspeed and sometimes it is really hard to keep up with so much information, most of it not much more than a drop in the bucket. So, I did what I usually do. Take a step back from it all, and decided to just slow down the cycle around my own interests. As a result, most of my writing this week is a bit more personal.

Whether it was celebrating my 23 years in San Francisco or simply using AI to identify the right phrase to describe myself, it was quite fun, and an exercise in some philosophical meandering. I also took the time to write about broadband, and yes, I got all worked up about Google flipping my favorite internet service provider to a PE-owned company. Now I am going to be looking for a new service.

Nonetheless, I took the time to enjoy reading what others are writing, especially the longer, less covered articles.

Here are seven stories I recommend.

• American Diner Gothic. Robert Mariani writes that economic stagnation, the death of regional culture, and the internet have converged to produce a new American type he calls the “dinergoth.” The pierced-up, anime-watching, gender-fluid Amazon warehouse worker who streams on Twitch and dreams of voice acting. I don’t normally even think about this. Now I do. [The New Atlantis]
• When ChatGPT Broke an Entire Field: An Oral History. Quanta Magazine interviewed 19 current and former natural language processing researchers to tell the story of what happened to their field when the transformer arrived. This is a fantastic read. Read it, and then read it again. [Quanta Magazine]
• How Oregon’s Data Center Boom Is Supercharging a Water Crisis. Rolling Stone and FERN investigate what happens when Amazon builds data centers on top of an aquifer already poisoned by decades of agricultural runoff. The water used to cool the servers evaporates, but the nitrates don’t. The wastewater gets sprayed back onto farmland. It seeps into the same wells people drink from. It leads to miscarriages and others losing a kidney. This is ugly. [Rolling Stone]
• Despite Doubts, Feds Approved Microsoft Cloud Service. Oops. Microsoft has a history of overpromising and underdelivering. Microsoft cloud is no different, as federal reviewers found out after five years and 480 hours. They wanted to verify how Microsoft protects sensitive government data in its cloud. They couldn’t. One reviewer called the documentation “a pile of shit.” They approved it anyway. [ProPublica]
• The $165 Billion Annoyance Economy. The “annoyance economy,” the cost of being on hold, fighting cancellation flows, disputing junk fees, navigating health insurance bureaucracy, costs American families $165 billion a year. This is capitalism gone wrong, and it rankles me no end. [Business Insider]
• How AI-Assisted ABS Will Impact MLB. It was Opening Day 2026. And it was also the introduction of robot umpires. Cornell researchers studied what happens when you replace human umpires with an automated ball-strike system. A nice reminder that “more accurate” and “better” are not always the same thing. [Cornell]
• Operation Red Wings: The True Story Behind Lone Survivor. Reality is not a movie. Politico Magazine revisits one of the most mythologized events in recent military history and finds a story far more complicated than the book or the movie. Read it slowly. [Politico]

What I wrote this week, ICYMI:

• Jargon Sucks — A short rant on why we hide behind words that mean nothing.
• Seeker — I used AI to find the right word to describe myself. The answer surprised me.
• 23 Years — A short photo-filled love letter to San Francisco on the anniversary of making it home.
• We Are Now an Upload (Broadband) Nation — Upstream usage is surging 21.7% year over year. It is the biggest challenge of infrastructure.
• Meta’s May Day — Meta has lost two cases in court. The problem is the way it lost them. Will it force them to rethink their entire operation?
• More Magic Math from OpenAI — The numbers still don’t add up.
• ‘Astound’ed. Google Flips Its Fiber to PE — Google sold GFiber to a PE-owned company called Astound. I am taking it personally.

March 29, 2026. San Francisco

“One should not only photograph things for what they are, but for what else they are.” — Minor White

I have not been out in the wild making landscapes for a while now. Almost a year. However, I have found new subjects to explore. San Francisco, obviously, remains my favorite muse. California is my darling. But at home, I have been exploring how to push myself creatively. And one arena I have been exploring is fountain pens — my other lifelong hobby.

I write with them. I cherish them. I collect them. What I wasn’t doing was photograph them.

I also found that most fountain pen photography is quite boring. The sameness of it. Product shots, meant more for a sale than for enhancing the beauty of objects so lovingly created. I wanted to do something different. More abstract. More me.

As the philosopher Dōgen said: “Not to seek reality apart from appearances, nor to cling to appearances as reality.”

I am trying. In the process, learning. To find beauty in the things I use to create words, which have a beauty of their own.

Two new tools have made this possible.

My friend Kiran Karnani of Harlowe sent me a couple of small portable lights — the Sol — that attach to the iPhone via MagSafe. LEDs with the ability to change color and temperature. With two of them, you can muck about with color, with shadow, with the way light wraps around a hard object and makes it feel alive. I don’t intend to become a macro expert. But these lights give me enough control to mean what I’m doing.

The second tool is a ShortStache Shift Diopter made by PolarPro. They sent me review units in 49mm and 82mm sizes. I have been using the 49mm version in combination with my Leica Q3 43, which has a built-in macro mode. A diopter, to put it simply, is a close-up lens — a secondary optical element you screw in front of your lens. It reduces the minimum focus distance and increases magnification, letting you get closer than your lens could otherwise focus. Particularly useful for macro work. Shift diopters come in three strengths — +2, +4, +8 — each pulling you closer to the subject. (I also have a Leica ElPro e52, and will do a comparison separately.)

Shift diopters opened creative possibilities I didn’t expect. The selective focus, the way sharpness falls away at the edges — it isn’t a technical compromise. It’s an aesthetic choice. I decide what matters, and the rest becomes part of the abyss.

It is an edit. Just like writing itself.

What I am really trying to do and in the process understand is how much of my approach to landscape photography translates to pens. Can I get to the very essence of their existence — not the whole object, but the fragments that tell the story? The light catching a section of celluloid until it stops being “pen” and becomes color, curve, depth.

The most important thing I keep reminding myself: leave something unexpressed. Give the viewer room to imagine the landscape. Rather than show it all.

I want to make pen photographs that feel that way. Nothing resolved. Nothing explained.

“In all things, irregularity is intentionally preferred.” — Yoshida Kenkō, Essays in Idleness

March 30, 2026, San Francisco

An update to More Magic Math from OpenAI

The final mad dash to IPO is on for the big AI companies. SpaceX, OpenAI, and Anthropic have all made their intentions clear. And nothing could be more obvious about OpenAI’s intent than today’s new funding announcement. A few things have changed since I wrote that piece. Some confirm my thesis, and one surprise, though not really.

So, the company says the round closed. $122 billion in committed capital, up from the $110 billion announced in February. Does committed mean money has passed the transom? We won’t know. What we do know is that at a post-money valuation of $852 billion, the anchors are Amazon, Nvidia, and SoftBank. Microsoft participated again, though not clear for how much. The additional $12 billion came from a who’s who of institutional money, including a16z, Sequoia, BlackRock, Blackstone, Fidelity, Temasek, D1, and Dragoneer. The FOMO gang!

So many brand-name investors show up at the last minute because none of them want to miss out on the sweet IPO pop. It surely will win them points with their own (limited partner) investors. I guess FOMO is also an affliction for the super rich.

By the way, nothing puts more in “t “less is more,” than more itself. In 2024, OpenAI raised $6.6 billion and sold about 4 percent of the company. In 2026, they raised $122 billion, twenty times more, by selling roughly 14 percent. Existing insiders and early employees must be in heaven.

To be clear, the circular financing problem hasn’t gone away. Amazon’s $50 billion is tied to an eight-year AWS contract. Nvidia’s contribution is compute, not cash. When I wrote about this in March, it was just an observation. Now it has a name. Bloomberg, Reuters, and others are now calling it “circular financing.”

OpenAI says it now has a $4.7 billion credit line from JPMorgan, Goldman Sachs, Citi, Morgan Stanley, and Wells Fargo. That’s not a lending syndicate. That’s an IPO underwriter roster auditioning for the job. It reminds me of those lining up outside Don Corleone’s room on his daughter’s wedding day. The credit facility is the gift they bring to get in the room.

Now let’s talk about the fix. The $3 billion in investments from individual investors. Axios reports that they are customers of three of the largest banks. Wait, the same three large banks that extended the credit facility and want to be part of the IPO underwriting syndicate. More circular economy at work.

OpenAI CFO Sarah Friar told Axios, “We are really trying to take to heart our mission, which is AGI for the benefit of humanity and thinking about access. Not just access to the technology, but also access to the economic upside that it’s driving.” That’s a nice line. It’s also an IPO marketing strategy.

OpenAI said it will be included in several ARK Invest ETFs. Cathie Wood, who previously invested in OpenAI through her venture arm, now gets to channel her retail base into pre-IPO OpenAI shares. Think about what that is. A private company, not yet public, getting into retail ETFs. That’s a new thing. It’s smart, too. You create demand before the IPO. You distribute the story. You make millions of people feel like they have skin in the game before you even file. That’s the fix.

ARK is overrated, to put it mildly. ARK’s flagship fund peaked in February 2021, then fell 75 percent. Five years later, a thousand dollars invested then is worth about $573 today. And ARK is the vehicle OpenAI chose to democratize the upside. Make of that what you will.

One last thing. OpenAI is quietly pivoting, shutting down Sora, its much-hyped video app, and concentrating resources on a “superapp” for developers and business users, with coding assistants at the center. Why? Because enterprise is exactly where Anthropic is eating their lunch. The $122 billion has bought more time to beat the competition. And did you notice that CFO Friar is doing all the press versus Sam?

You focus this hard because you want to go public. Fast. After all, you don’t want to be the one without a chair when the music stops. Your move, Dario!

Previously on this topic:

• More Magic Math from OpenAI?
• OpenAI Has New Focus (on the IPO)
• $110 Billion in Name Only
• The Debt Beneath the Dream

“A newspaper is not only a collective propagandist and a collective agitator, it is also a collective organiser.” — Vladimir Lenin

In 1902, Lenin argued that his revolution needed a newspaper of its own, and that newspaper was (unironically) named Pravda, which means truth in Russian.

“The standard communications playbook just doesn’t apply to us. We’re not a typical company. We’re driving a really big technological shift.” — Fidji Simo, 2026

Simo, OpenAI’s CEO of Applications, explained this to OpenAI staff as to why OpenAI had just bought TBPN. Different century. Same logic to explain an emerging new socioeconomic order, a new post-revolution reality.

To recap, OpenAI, weeks if not months before its public offering, has acquired the Technology Business Programming Network (TBPN). It is a daily, three-hour live tech talk show hosted by founders Jordi Hays and John Coogan. The show had 58,000 YouTube subscribers, $5 million in revenue last year, and a cult following in Silicon Valley. It was profitable, had no outside investors, and was growing fast. As a former media person, any media exit is a good exit. Congratulations to the founders.

Now let’s dig into what OpenAI really bought and why.

Think of TBPN as a room. A room where people in tech come and talk openly, without having to worry about the antagonistic queries that big media is often posing them. They go because it feels like peers talking, not press interrogating.

If you lived through Internet 1.0, CNBC’s morning show had the same vibe. ESPN’s SportsCenter did the same for the cable sports revolution. They don’t speak to power. They amplify power. TBPN made that playbook its own.

The best part of the announcement is where TBPN sits inside the OpenAI org and who it reports to. The show sits under “strategy” (and not communications) and will report to Chris Lehane. He is OpenAI’s chief political operative, the man associated with coining “vast right-wing conspiracy” as a deflection tactic during the Clinton White House years, who built Fairshake, the crypto super PAC that spent hundreds of millions targeting anti-crypto candidates in 2024.

You don’t put an editorially independent media property under your political operative. That’s not a media strategy. That is part of overall strategy. But they still have to sell it as keeping it editorially independent. Pravda was technically editorially independent too.

Anyway, maybe I have lost my rose-tinted glasses, but when a company says “do no evil,” you know what they are going to do. And if a press release mentions editorial independence four times, you know where that is going. And if you have to invent a new phrase, “Editorial Independence Covenant,” then I don’t really need to spell it out for you.

“The conscious and intelligent manipulation of the organized habits and opinions of the masses is an important element in democratic society.” — Edward Bernays, Propaganda, 1928

Bernays, who was Sigmund Freud’s nephew, invented the PR industry. He thought of it as a key part of corporate design. He was hired by corporations such as United Fruit and American Tobacco to do exactly what OpenAI is doing now. Build the perception layer around a product by presenting it as a massive societal shift.

The Soviet term was agitprop. The Silicon Valley term is earned media strategy. A century apart, history still rhymes.

Related Reading

• OpenAI: The Fix Is In — How OpenAI is engineering its IPO narrative through circular financing and retail ETFs.
• OpenAI Has New Focus (on the IPO) — On Simo, Facebook-style engagement loops, and the real game being played.
• More Magic Math from OpenAI — The PE deals, the JV structure, and why the real product is the IPO prospectus.
• The New Announcement Economy — How OpenAI mastered the press release as a capital-raising instrument.