We really, really need to talk about venture capital. Because it’s not “venture capital” anymore. There’s a huge disconnect between what most people think of VC, where an investor has a big fund and cuts checks to help a founder build a company, and the current reality, where a handful of billionaire extremists use the cover of “VC” to advance an outrageous agenda where they’re accountable to no one. I’m gonna explain this from a standpoint that almost never gets articulated: I’ve personally raised tens of millions of dollars in venture capital funding as CEO of startups, and been directly involved as a board member or advisor in raising hundreds of millions more. I’ve sat in board rooms, across the table from the people I’m talking about here, or been at the industry events that they frequent. So this isn’t sour grapes because these VCs wouldn’t cut me a check, or some chip on my shoulder about these investors due to a business deal. This is what I know about these bad actors because I’m part of the community of creators and inventors who build the things that they used to invest in — back when they still cared about innovation. Many of the trends in society and politics that people are most angry about, from data centers being forced down everyone’s throats, to all of our favorite apps and services being enshittified, to politicians being paid to ignore the will of the people, are all being supercharged by these cancer capitalists. They have warped the structure of venture capital into a form of oligarchy that answers to no market, no regulators, and no voters. So it’s worth understanding exactly how they did it. How Venture Capital became Cancer Capital I’ll be breaking these points down in further detail, but just to begin framing the concept, I’ll lay out the core idea here in some bullet points (so that you’re not tempted to run this whole thing through an LLM): Venture capital was only supposed to be a tiny segment of the overall capital market, but it has expanded to become the primary form of funding that new companies consider — it was never the only way, and it didn’t used to be the default one VC was meant to be a small percentage of overall investment because it represents the high-risk, high-reward part of a portfolio; to be healthy, most of a portfolio — or most of an economy — needs to focus on assets that are more stable and predictable. But a cancer grows from a cell that a body needs in small, healthy amounts, and that turns deadly when it grows without limit until it harms, or even kills, its host. As regulations have gotten looser in recent years, a handful of venture capital firms have become “do everything” funds that combine private equity with their existing VC businesses, and expand to manage massive stockpiles of tens of billions of dollars The 1% of VC firms that get this big stop being exposed to the risk in their own investments at all — when you collect 2% a year to manage $50 billion, that’s a billion dollars landing in your pocket annually whether any company you funded lives or dies. Those firms also stop legally even being venture capital firms, making them unaccountable to markets, founders, or the law — and that’s how they become “Cancer Capital” Meanwhile, the 99% of “normal” VCs don’t have the power or funding of the Cancer Capital firms, but are forced to play on the field that those firms define, even if they don’t like the way they do business Since the Cancer Capital firms have become so powerful, the overall balance of power between founders and VCs has flipped; instead of founders having a company that VCs would try to fund, now VCs publish extremist political manifestos, and “founders” are just the people who are selected to carry out parts of those plans The rest of the world doesn’t know: New founders and workers entering the tech industry are unaware that Cancer Capital has taken over, so many are still trying to play by the old rules, and can’t figure out why their ideas are being pushed into serving the goals of the Cancer Capital firms Politicians and media still look at VC as if it works like it did 10 or 20 years ago, and cheer them on like they’re funding job creation or enabling new companies to grow, when their primary goal is concentrating power and wealth into the hands of the Cancer Capital tycoons. They keep getting fooled by this, over and over. These days, venture firms are increasingly getting their funds from pension funds and retail retirement accounts, meaning the public (you!) are increasingly holding the bag for the parts of their portfolios that actually have some risk, even if you never intentionally made that choice The shift away from IPOs in the tech industry has also encouraged these Cancer Capital firms to find ways to cash out long before companies ever go public, meaning they can make a massive return off of companies that never make a penny of profit, even if regular investors get screwed by the stock of a company once it actually gets listed on the public stock market. Part of why this has gotten so corrupt is the way the Cancer Capital firms have transformed themselves into their post-VC forms. Because they’re not legally VC firms anymore, they’re free to buy shares directly from founders, or hold unlimited amounts of publicly-traded stock — exactly what they couldn’t do as regular VCs. They can even sell their investment in a company as an asset to another one of their own funds, and then book the increase in value as a profit, all without the company ever having made a penny. Another racket: a company that’s raised a bunch of cash in a funding round can buy out its early investors if they’re one of these post-VCs, so they can get paid off even if their portfolio company has never made a penny in profits or revenues. All of this self-dealing, and the way that they’re isolated from any accountability, has made these firms become more and more shameless in their behavior. Former Andreessen Horowitz partner John O’Farrell publicly called out the firm (a rarity — the company is notoriously vindictive towards those who it decides are disloyal) for what he called its “political infiltration” of AI policy. Marc Andreessen, Ben Horowitz and their firm have put $115.3 million into this midterm cycle — nearly double their $63 million in 2024, and more than any other billionaire donor in the country, even including Elon Musk. Molly White, whose Tech Influence Watch tracks this money against FEC filings, shows that a16z alone accounts for more than 20% of all political contributions from the entire cohort of crypto and AI companies it follows. And they’re funneling these funds to candidates in both parties. This is a huge escalation from the tentative baby steps that folks like Zuckerberg were making in the Obama era, working on benign issues like trying to help immigrants. And of course, it gets a lot worse than just their lobbying. As I have frequently noted, Andreessen Horowitz hired a man as a partner at their firm despite his having no background or qualifications in tech, finance, or startups whatsoever. His only discernible qualification was that he had choked my unarmed neighbor Jordan Neely to death on a subway car. This is how brazen, how toxic and destructive, we’ve allowed the industry formerly known as venture capital to become. We must understand that it is no longer a financial machine that is used to fund startups, but a political and social machine focused on dismantling democracy and civil society. And it’s time to act accordingly. Up next: we’ll dive into the specifics of many of the points laid out above, to understand more about how we got here.

For years, I’ve wanted to have a standalone version of Apple’s Touch ID authentication feature for my Mac, but without having to use an Apple keyboard. (I generally like their keyboards, but my daily driver keyboard these days is a big clicky mechanical beast.) I’d gone down various dead ends of trying to find substitutes, and even checked out the efforts where people had ripped apart expensive Apple keyboards just to scavenge the Touch ID sensors out of them. None of them quite solved the problem. So today, I’m sharing an open source project called Dashboard Touch, which lets you make your own Touch ID-style sensor for your Mac, using low-cost off-the-shelf part. It’s based on an extensive refactoring of the excellent tinyTouch project by Zimeng Xiong, who recently cracked the code on how to make a useful fingerprint scanner system that’s also reasonably secure for regular Mac users. (You should definitely check out his project and support his new hardware build if you’re interested in this stuff.) I took my own approach to this work because I wanted to focus a lot on having a friendly web interface for configuring exactly how the fingerprint sensor system works on your computer. When you get Dashboard Touch set up, it presents you with a nice web interface that runs right on your own Mac, letting you do things like set the color of the ring light on the fingerprint sensor, or capture your fingerprints so they’re recorded in the system. Behind the scenes, the way the system works couldn’t be simpler. You buy a little fingerprint sensor, and a small microcontroller, wire them together (it was actually fun to get back to soldering stuff!), and then plug them into your computer with a regular USB cable. After you run the setup script, you just go to the web interface and add your finger(s) to the system. Once you’re running, your Mac runs as normal, except any time the system prompts you to type in your password, you can just swipe your fingertip on the sensor and Dashboard Touch will type your password in for you. At a technical level, the device is actually literally pretending to be a keyboard. (You can look over the code on GitHub and get a feel for the approach pretty quickly.) After it’s installed, it’s basically a set-it-and-forget-it kind of thing. You don’t need to do anything else for it to Just Work. Your password is only ever stored securely on your Mac, and your fingerprints are only ever stored securely on your sensor. You can erase them at any time and nothing talks to the internet at all except the one manual update checker, where you can see if there’s a new version of Dashboard Touch, but only when you intentionally click the button to request it to do so. Overall, this isn’t the kind of system you should use if you’re protecting a bank vault, but if your computer is physically secure and nobody is going to have extended unsupervised access to your Dashboard Touch setup without your permission, you should be fine. Dash, not bored Personally, it’s been really fun to get back to making things. As you can see in my introductory video, I ended up creating an enclosure for my fingerprint sensor in my woodshop, so that it would match my desk that I recently built. Even creating and editing the intro video was a fun project that took me out of my usual comfort zone. Nearly every task in this project has had me stretching to do things that I’m pretty bad at, from firmware coding to security review to detailed carpentry to video editing. But I just love the idea of putting things out there again for people to hack on, and there's also something really satisfying about being able to be super-opinionated about the design and user interface of something after so many years of working on teams where I had to collaborate. (Even though I always got to collaborate with brilliant people, it's different when you get to pick every pixel!) I just also have been missing the era of the web when most of what I saw online was weird and fun things that regular people were building, and I realized that I can't mourn the absence of those kinds of projects unless I invest my own energy into building some of those kinds of things myself. So, here's one! Let me know what you think, and if you've got any ideas for how to make this thing better. Or, of course, if you find any bugs that I should fix. I hope you have fun touching the blinking lights!

The vast majority of the documents people use to do business are really quite poor. Presentations that make your eyes glaze over, memos that are inscrutable or unclear, and all kinds of artifacts that say more about how they were created than whatever message they were ostensibly trying to communicate. It's been one of my great frustrations for years, and a big part of why I wrote Make Better Documents a while ago. That post captured a list of the suggestions I've been giving people for years on how to make better, more effective documents that can actually do work for you, instead of fighting at cross purposes to your larger goals. To my great surprise, that list of suggestions on how to make better documents got a pretty huge response, and a lot of people told me they found it really helpful. So now, I've created a Better Documents skills.md file for people who use LLM tools like Claude to help assist them in creating business documents, to prompt their AI tools to make better documents by default. If you're not familiar, agent skills are simple text files that describe new capabilities or processes that LLMs can take advantage of when carrying out tasks. (They're Markdown files — more proof of how Markdown is taking over the world!) The way this skill works is that it's distilled the broad principles I outlined in that post into a series of 5 tests, covering areas like whether you've properly considered your target audience, whether the overall structure is correct, if you've overdone things with your formatting, and if things are named clearly, and then either generates a new file that follows those rules, or reviews an existing document to make sure it is obeying best practices. It's nothing too fancy, but I've been using it for a while, and shared it with a few friends, and people have told me they found it handy and it's improved some of their routine documents. I'm especially glad that people have found it useful even if they're the kind of folks who would never let an LLM generate a document on their behalf, but do think software tools are useful for things like spell check or grammar check. I see this as being a tool in that kind of category. If you're familiar with skills, the install process is really simple and works just like any other skill. This skill is totally free and open source (if you have improvements, send along a pull request on GitHub, or if you're not a coder type, just email me or hit me up on social media and let me know what fixes/suggestions you've got), so there are no encumbrances or restrictions on its use. I am curious if it's useful to people, especially if you find it handy to use more broadly at a company, so don't be shy to get in touch if you find it valuable. Here's to us all enduring fewer terrible presentations!

“[T]here can be alternatives. What we can imagine is, rather than the ChatGPT killer, a lot of different little AIs from little responsible players.” That’s me, in The Guardian a few days ago, trying to distill a message that I’ve been trying to get out as broadly as possible for quite a while now. It's sort of like hoping a comet will take out the major AI players and a bunch of smaller new players will be the smarter, better-adapted mammals that take their place instead. We’re in another one of those big inflection points for AI. Trump administration policymakers for AI suspended access to Anthropic’s newest product. All of these policymakers have a web of investments in competing players — including SpaceX, which is about to IPO — and the corruption and grift of this cohort are so extensive that it’s impossible to judge what the actual risks and reality are around any of these platforms or technologies, since no one involved is an honest broker. It’s a shame More broadly, there’s been the widespread pushback against AI culturally, one that is undeniably strongest amongst those who were born in this century. But the adoption patterns and usage data show that even younger people are using some AI tools. And that’s a pattern that we’ve seen before, with social media. We have a significant group of people knowing that a technology contradicts some of their values, preferences, or beliefs, but using it anyway. Sometimes it’s due to the coercive nature of the platforms themselves, and how they insinuate themselves into our lives, to the point where we don’t even realize we’re using them. Sometimes they are forced upon us by the creators of the platforms, since they have so much power over the devices we use, and the tools that we rely on for things like doing our jobs, or communicating with our loved ones or our communities. There are millions of people who don’t like that they’re using LLMs provided by the Big AI companies, but end up using them anyway. Just like there are hundreds of millions of people who don’t like that they’re on the giant social networking platforms like Facebook, but end up using them anyway. The feelings that people walk away from those experiences with are often guilt, or shame, or embarrassment, or resentment — all some of the most negative and destructive emotions that humans can experience. Actual alternatives But if people want to get the benefits of some of these technologies, without either the shame of supporting the harms of Big AI, or the unpredictability of being beholden to corrupt billionaires bickering with one another, there are finally starting to be other options. As I mentioned in (One) Good AI Is Here, it’s possible for creators working in their own communities to now make AI tools that serve their specific needs, without causing all the harms that make people object to Big AI. This feels like the true alternative to the narrative of “inevitability” that so much of the hyper-funded AI industry is trying to push, while also not forcing people into a quiet life of AI guilt if they still find some utility in some aspects of these tools. Right now, those who (rightly!) object to Big AI due to their platforms’ impact on the environment, or labor, or their extractive use of content without consent, or its many other potential harms, are generally not aware of, or often open to, the idea of there being small, human-scale tools created by and for communities that are accountable for those tools over time. But my suspicion is that it is not only possible to make these tools, there may in fact already be many of these tools in existence, and we’re just not as familiar with them because they’ve been quietly serving their specific niches without having multi-billion-dollar campaigns promoting them. What I'm unabashedly hoping to do (and I think the Guardian story reflects some momentum in that regard), is shift the narrative from focusing on running away from the bad thing in AI, to finding the good thing that we're running toward. There are alternatives that we could be affirmatively choosing, ones that look at questions like the one I asked more than a year ago, "[https://www.anildash.com/2025/05/01/what-would-good-ai-look-like/](What Would "Good" AI Look Like?")", and offer answers that might give us hope instead of just the righteous rage and anger we feel when we let our imaginations be constrained by the limits of what Big AI offers.