The ultraviolet catastrophe analogy is spot on. That idea of runaway signal intensity due to a lack of constraints maps eerily well to what we are witnessing across our social and technical systems.
I've been giving thought to how many of our institutions rely on the friction of effort, the assumption that time, energy, and authenticity act as natural and very necessary filters for intent and also as a result limit volume needing to be addressed. Job applications, college admissions, grant requests, and financial aid were all built on the idea that effort gates access. If someone applies, we assume they are serious, and the teams receiving these applications are staffed to handle a certain volume of them in a human way.
That assumption isn't just under pressure, I think we could say it's already broken.
AI can now generate resumes, essays, recommendation letters, and grant narratives in seconds. The cost of applying has collapsed, and we are seeing the consequences. In higher education, fake students are applying for aid, submitting auto-generated coursework, staying enrolled just long enough to trigger financial aid disbursements, and then disappearing. Entire programs are beginning to buckle under the weight of this.
In hiring, the shift is dramatic. A few years ago, I would get fifty applicants for a role on my team at MSFT. Now I see over a thousand. Many are nearly indistinguishable at first glance, well-written, tailored, full of the right keywords. Interestingly however, the "right" keywords are becoming more homogenized, when everyone uses them they do not carry the same weight. Some resumes are very clearly synthetic with em dashes and "Ethereal fractured tapestries". Others might just be optimized to the point of fiction.
And this pattern is spreading into other domains that were never designed for zero-effort interactions. Scientific publishing, civic feedback systems, online reviews, even legal filings. Institutions that depend on sincerity and scarcity are being flooded by volume and automation.
This is no longer just noise. It is interference. When signal can no longer be separated from spam, the system loses meaning. We lose trust, not just in messages, but in the structures meant to help us navigate them. Your idea of shifting from addressable identity to encrypted channels, replacing nodes with edges, feels like the most realistic path forward. Mutual, private, consent-based channels where communication is only possible with shared context and permission. That is a powerful mental model, and I am glad to see it gaining traction.
There is real turbulence ahead, but I am with you in thinking there is also a chance to rebuild with better defaults.
AI token generation is not a zero cost activity either. The current price is also subsidized by VC to be lower for end users than it really is. It also takes time. Spam will always be in an arms race with spam filtering. Like virus/anti-virus. Slop/anti-slop will be the latest battle. But I strongly doubt that this results in the end of the internet.
Spam filtering doesn't seem to be winning, for me. I cannot use my email without help. And to my knowledge there is zero protection on SMS and Messaging endpoints. I hope you are right, but I'm not so sure.
Great topic and I dig the naming! Another example to posit: the generated Youtube channels, content and comments have quickly morphed the platform into an uncanny community.
I feel it's only a matter of time before we build inner networks, cities of trust built on humans vouching for one another's humanness.
I suspect all the alternative duct tape like biometrics, i.e. Worldcoin Face Auth, are still digital systems which will continue the same cat mouse race mentioned in comments here like viruses, spam, captcha, and so on.
Also, I just have to point out the light emerging with this shadow. There’s an imminent and explosive wave of co-creativity as people who never saw themselves as “creative” pick up these tools and leap past the technical roadblocks. There’s something deeply therapeutic in this rekindling of our relationship to self-expression. Human-to-human value exchanges are accelerating here. If the noise filtering works, we walk into an age of abundance with more human agency than ever before.
I enjoyed your article. Who knew that 70 some years ago the “ party line on the phone “ would be a precursor to what we face on the internet today.
Although I understand very little of the technical issues you address. I understand and am subject to the practical results, on a daily basis, as I look at my phone and computer. Thanks for the future insights. My next communication to you will probably be by mail.
Couldnt AI agents working on our behalf also toss packets it knows we have no interest in? I for one believe whoever figures out the AI personal advocate programmed to live and learn locally on our personal cloud and guard us from unwanted packets, unreasonable TOSs, and keep our payments and groceries topped up will be the next billionaire du jour.
Fully agree Philip. Add to that the increasing number of Robocalls on our cellphone numbers that are AI generated sales/spam calls making also our phones useless soon...
Another example: https://www.nytimes.com/2025/06/21/business/dealbook/ai-job-applications.html
Going to have to go back to good old face-to-face interviews.
True, and fortunately still easy to recognise as non-human. The title is great, Ultraviolet Catastrophe sounds like a very cool song.
The ultraviolet catastrophe analogy is spot on. That idea of runaway signal intensity due to a lack of constraints maps eerily well to what we are witnessing across our social and technical systems.
I've been giving thought to how many of our institutions rely on the friction of effort, the assumption that time, energy, and authenticity act as natural and very necessary filters for intent and also as a result limit volume needing to be addressed. Job applications, college admissions, grant requests, and financial aid were all built on the idea that effort gates access. If someone applies, we assume they are serious, and the teams receiving these applications are staffed to handle a certain volume of them in a human way.
That assumption isn't just under pressure, I think we could say it's already broken.
AI can now generate resumes, essays, recommendation letters, and grant narratives in seconds. The cost of applying has collapsed, and we are seeing the consequences. In higher education, fake students are applying for aid, submitting auto-generated coursework, staying enrolled just long enough to trigger financial aid disbursements, and then disappearing. Entire programs are beginning to buckle under the weight of this.
In hiring, the shift is dramatic. A few years ago, I would get fifty applicants for a role on my team at MSFT. Now I see over a thousand. Many are nearly indistinguishable at first glance, well-written, tailored, full of the right keywords. Interestingly however, the "right" keywords are becoming more homogenized, when everyone uses them they do not carry the same weight. Some resumes are very clearly synthetic with em dashes and "Ethereal fractured tapestries". Others might just be optimized to the point of fiction.
And this pattern is spreading into other domains that were never designed for zero-effort interactions. Scientific publishing, civic feedback systems, online reviews, even legal filings. Institutions that depend on sincerity and scarcity are being flooded by volume and automation.
This is no longer just noise. It is interference. When signal can no longer be separated from spam, the system loses meaning. We lose trust, not just in messages, but in the structures meant to help us navigate them. Your idea of shifting from addressable identity to encrypted channels, replacing nodes with edges, feels like the most realistic path forward. Mutual, private, consent-based channels where communication is only possible with shared context and permission. That is a powerful mental model, and I am glad to see it gaining traction.
There is real turbulence ahead, but I am with you in thinking there is also a chance to rebuild with better defaults.
David
AI token generation is not a zero cost activity either. The current price is also subsidized by VC to be lower for end users than it really is. It also takes time. Spam will always be in an arms race with spam filtering. Like virus/anti-virus. Slop/anti-slop will be the latest battle. But I strongly doubt that this results in the end of the internet.
Spam filtering doesn't seem to be winning, for me. I cannot use my email without help. And to my knowledge there is zero protection on SMS and Messaging endpoints. I hope you are right, but I'm not so sure.
Great topic and I dig the naming! Another example to posit: the generated Youtube channels, content and comments have quickly morphed the platform into an uncanny community.
I feel it's only a matter of time before we build inner networks, cities of trust built on humans vouching for one another's humanness.
I suspect all the alternative duct tape like biometrics, i.e. Worldcoin Face Auth, are still digital systems which will continue the same cat mouse race mentioned in comments here like viruses, spam, captcha, and so on.
Also, I just have to point out the light emerging with this shadow. There’s an imminent and explosive wave of co-creativity as people who never saw themselves as “creative” pick up these tools and leap past the technical roadblocks. There’s something deeply therapeutic in this rekindling of our relationship to self-expression. Human-to-human value exchanges are accelerating here. If the noise filtering works, we walk into an age of abundance with more human agency than ever before.
Peter
I enjoyed your article. Who knew that 70 some years ago the “ party line on the phone “ would be a precursor to what we face on the internet today.
Although I understand very little of the technical issues you address. I understand and am subject to the practical results, on a daily basis, as I look at my phone and computer. Thanks for the future insights. My next communication to you will probably be by mail.
Peter
Couldnt AI agents working on our behalf also toss packets it knows we have no interest in? I for one believe whoever figures out the AI personal advocate programmed to live and learn locally on our personal cloud and guard us from unwanted packets, unreasonable TOSs, and keep our payments and groceries topped up will be the next billionaire du jour.
Well said Philip. Most of us already don't answer a phone call unless we recognize the number. Thank goodness for caller ID!
Fully agree Philip. Add to that the increasing number of Robocalls on our cellphone numbers that are AI generated sales/spam calls making also our phones useless soon...