R[ai]dio on the Internet

It's a pun!

“My big feeling was I just missed it — I missed the whole thing. It had happened in the ’80s, and I got here too late” – Marc Andreessen on moving to the Bay Area in the early 90s.

Thesis: Platform shifts always feel obvious in hindsight — from putting radio on the internet to moving CRM to the cloud — but at the moment they emerge, they’re messy, non-consensus, and easy to dismiss. Today’s AI moment mirrors those past shifts: the real value won’t come from copy-pasting old formats, but from the unexpected ways AI will reshape consumer behavior and workflows, creating the next generation of ‘radio on the internet’ opportunities.

If you were investing / building a company in 2018, or 2020, or 2023, I bet you had the same feeling I had, which was I know exactly how Andreessen felt. Building a tech company was over – all the interesting things have been built. Of course, I never actually said that, and maybe I never even really believed it (how can you as a VC?), but in my darker moments, it certainly felt like that could be possible.

And in some ways, it was right. It’s not like there were a ton of great companies built from 2017-2018. It was a lot of SaaS, a lot of marketplaces trying to be the Uber for X, or something like that. There were a lot of facsimiles floating around out there, but any truly great platforms built during that period?

No, in fact, that was a period of platform consolidation. Meta / Facebook was gobbling up its app development platform, Apple was putting the screws on its app store, and the Aggregators were Aggregating. The race had been run, and now it was just a matter of divvying up the spoils.

But the cool thing about technology is that it’s hard to really know where great innovation is going to come from. Yes, in 2017, there were a lot of people excited about AI, but not in the way they are now. And they weren’t even really talking about AI. The serious people called it something else altogether – Machine Learning.

Interestingly, Google did release the transformer model paper in 2017 . They used it, but not necessarily in the ways people use transformer technologies to build LLMs. It was an internal framework that helped improve NLP tools like BERT and improve Google Translate. But, it took a couple of years before it started actually having an impact outside of Google.

So, as an investor during this period, I remember having conversations with the grizzled veterans of technology, and I would inevitably hear stuff that would make my jaw drop. Not because it was so revolutionary or radical, but because it was so obvious. Of course, everything is obvious looking backwards.

“You can’t connect the dots looking forward; you can only connect them looking backwards. So you have to trust that the dots will somehow connect in your future.” – Steve Jobs

At the time, it was probably pretty non-consensus, whatever it was that they were building. But the point was always driven home better by Silicon Valley’s Russ Hanneman – the man who “put radio on the internet”. Throughout the course of the show, it is played for laughs that he is a billionaire because he put radio on the internet. Not to be an over-explainer, but the core point is that HOW HARD COULD THAT HAVE BEEN, or rather, MAN IT WAS EASY TO GET RICH BACK THEN.

And that’s the thing about looking back at platform shifts. A lot of stuff that seems extremely obvious now was a great way to generate value and build a business back then. Most things tend to appear obvious looking back, especially when framed around the growth of said platform (such as the INTERNET!!!).

But even the internet was not obvious when the Real Russ Hanneman (Mark Cuban) put radio on the internet. Commercial use wasn’t broadly permitted on the internet until the early ’90s, with full opening by 1995! That’s not that long ago! That changed – and the com in .com stands for commercial. Mark Cuban sold broadcast.com to Yahoo! for $5.7bn in April 1999. Of course, that was near the peak of the bubble, but it’s worth noting that the company was founded in 1995 and by 1998 was making ~$20M in annual revenue. That revenue is NOT WORTH $5 BILLION. But it was a business with impressive growth, all things considered.

What’s funny about the Russ Hanneman joke is the fact that Radio on the Internet actually might not have ever been a great idea. One of the most compelling things about the internet, and still is today, is that it allows for asynchronous communication, with wide swaths of people, in a very short period of time. It has led to the idea from Kevin Kelly of 1000 True Fans – meaning, among many other things, that the internet would allow for niche communities to blossom. This would, in turn, lead to what some would refer to as the death of the monoculture. Meaning that as a society, we were no longer focused collectively on one thing at one time. Game of Thrones is somewhat famously referred to as the last Monoculture Television show and that show ended on May 19, 2019, six years ago!

And what is the Radio good for? Broadcasting to wide swaths of people SYNCHRONOUSLY. It is inherently not a stored-up medium. It is great for live shows, sports, musical acts, etc. It has a purpose that the internet can service (distribution), but it already had great distribution - 80 million people still listen to the radio today, and most are not doing so through the internet. To this day, radio on the internet is not exactly a thing? There is Sirius (although, 90% of their listeners are in a car, so not sure it’s really an internet company), but what other mediums allow for “radio on the internet” today?

There is podcasting, which has, in many ways, replaced talk radio, although there are still a lot of people who listen to radio still. But, to be very clear, Podcasting and Radio are DIFFERENT MEDIA. They seem pretty similar, but they are different formats on different platforms, with largely different audiences. Podcasts cover niche topics, talk radio covers broad bases. Older generations listen to talk radio, younger ones listen to podcasts. Podcast ad spend grew double digits last year (~$2.5 billion), while talk radio stayed flat (~$12 billion) and down from it’s peak. These things are different.

Alternatively there is livestreaming, which actually fills a similar need to radio, but for a younger generation and in a much different format / consumption paradigm. People think of streamers as being strictly gamers, is that true? Most streamers aren’t even the best “gamers” but rather personalities talking about a wide range of topics that the audience can relate to. They use gaming as a lingua franca – everybody plays video games at that generational level. This is just like how sports talk radio hosts used sports to draw in folks to converse about other nonsense. Other analogies exist as well, like the Chat instead of at-home-callers. They even have similar personalities in a way. IShowSpeed might be the modern day Howard Stern.

So the people who REALLY put radio on the internet? Twitch.

But was streaming video games obvious in 1998? Or even 2008? Or EVEN 2018? No. That’s the trick about these platform shifts – they seem SO obvious looking backwards.

What’s more worthwhile is trying to understand how the platform shift will impact the user and end consumer. It’s almost never as simple as saying “let’s take an existing format and plug it into the new platform”. Even the examples that seem like they must be like that tend to have something working against them.

When Salesforce put the CRM on the Cloud, that would seem like a natural fit – let’s take a big software market and put it into the cloud so that it could be more easily accessible and more malleable of a tool to meet different business needs more quickly. Who says no?

Well, a lot of folks did at first. They didn’t want to upload their private, confidential customer data onto someone else’s computer, aka the cloud. “Will my competitors see my data if it’s hosted outside my firewall?” That seemed like a big security risk at the time and a lot of folks resisted the shift to the cloud. In the early aughts, there were a lot of CIO, CTOs, etc. who were citing compliance and security reasons for not getting on the cloud.

And of course, in retrospect, that seems very stupid. The cloud makes access to data 100x simpler and makes most jobs easier. In the end, cloud crushed competition because it was cheaper (fewer tech folks on staff to manage on-prem, less onboarding costs, etc.), faster (deployed upgrades), more reliable (outsourced security to a team that only focuses on that is a better deal), more scalable, and more malleable / innovative.

A lot of the same questions are currently being asked about AI. Is it safe? What are they going to do with my data? Is the cost worth it? I think those questions will be answered over the next five years. What I am most interested in is what Radio on the Internet moments will we have with AI.

There are some early contenders. Like AI tools that help you create marketing content. Or help you write up proposals for business. Or agents that do research for you. But maybe even deeper is how will AI impact consumer and user behavior?

Below is a quick list of ways that behaviors have changed already because of AI (created by AI for effect!):

· Consumers increasingly turn to AI chatbots instead of search engines for product research, travel planning, and “what should I buy” queries.

· Early adopters are letting AI shopping agents auto-generate carts (Instacart’s AI meal planner, Amazon’s AI search, Klarna’s AI shopping assistant). Shoppers spend less time browsing manually.

· Customers now assume 24/7 intelligent chat support that can handle complex queries (airlines, banks, telcos). Waiting for a human agent is seen as outdated.

· Everyday people use AI for image generation (Memes, MidJourney, Canva AI), music creation, and video editing — lowering barriers to creative expression.

· Consumers consult AI fitness and wellness bots (Cure, Whoop coach, Replika, therapy chatbots) for guidance, not just apps. Many interact daily with AI “companions.”

· Employees increasingly start documents, emails, or slides with AI-generated drafts (Microsoft Copilot, Google Workspace Duet). The blank page is disappearing.

· Developers rely on GitHub Copilot or Cursor for ~40–50% of code lines in some orgs, shifting engineers’ time toward review/debugging instead of raw typing.

· Analysts and consultants query LLMs for first-pass synthesis of reports, earnings calls, or market scans instead of starting with Google + manual reading.

· Early adoption of AI scheduling agents and inbox managers (x.ai successors, Superhuman AI triage, Reclaim.ai) shows workers offloading routine coordination.

Beyond that list, it’s notable that search will definitely be altered. That seems to be the thing on everyone’s mind - does Google finally have a real contender? It’s not that AI is better than current search tools. In fact, in some basic cases, I think it can be worse. But I think it can DO MORE than search. So now there will be new ways that people are interacting with the internet – so what does that mean for end-state of businesses?

What is the next Radio on the Internet? That’s the question everyone is racing to try to answer. It’s important to keep in mind that it’s probably not as straightforward as it may seem. And the race is definitely not over already, but it has started.

And remember, this is just the beginning of this adoption curve. There is much more to come. It will seem obvious in five years, and it will be very much non-consensus today.