Eating the Internet with a Fork

The early internet was a buffet. Distinct dishes, labeled and compartmentalized. You went to Slashdot for nerd chatter, Flickr for photos, Wikipedia for facts. Each site had a shape, a tone, a job. You could draw a map.

That internet is gone. Today the internet is soup. Content is not static. It morphs, flows, takes the shape of whatever container is currently rendering it. TikTok eats text. YouTube absorbs podcasts. Substack ate blogs. Twitter ate everything, then changed its name, then ate itself. The serving platform dictates the flavor of the dish.

The shift mirrors the death of the expert and the rise of the generalist-amateur. You can watch a YouTube channel that lectures on quantum physics between mukbangs. Reddit threads can rival journal articles in depth right up until a meme rips through and the thread is cooked. Expertise dissolves into omnivorous curiosity. Learning today isn't linear. It's opportunistic and fragmented. That's terrible for deep competence and great for intellectual breadth, and you usually only get to pick one.

Here's the part that took me a long time to internalize. Every hour you spend scrolling costs you something. Learning is not the accumulation of information. It has opportunity cost. What you choose not to finish is as important as what you do finish. You skim five "best pasta recipe" articles in the time you could have spent learning actual knife skills. Every click has a shadow.

The productivity-tooling industry is the dark reflection of this. Apps and systems that promise to help you "hack" your life often just keep you remixing, organizing, reorganizing. One more folder. One more highlight. One more bookmark. The thing you were actually trying to do remains untouched. It's antiproductivity dressed as productivity. The skill is distinguishing the shiny wrapper from the core action.

So how far into the soup do you go before you stop tasting options and just eat? Here's the rough heuristic I use. Call it the Fork Rule. You sample alternatives until your hand hovers, prong-ready, over something both immediately usable and not obviously suboptimal. You stop when the marginal utility of more searching is less than the marginal benefit of just starting. It's crude. It beats infinite scroll.

The meta-point is that the internet is a fractal of choice. It amplifies both freedom and paralysis. Every platform warps content into its mold. Every click fractures attention. Every new niche invites unending exploration. The challenge isn't lack of information. It's strategic ignorance and focused execution.

There's a second-order version of this for the people building models. The model is also eating the internet with a fork. It's also doing it badly. It's also struggling to tell signal from noise, gold from garbage, current from outdated, opinion from fact. The bias of the corpus becomes the bias of the model. The temporal cliff of the training data becomes a cliff in the model's worldview. The optimization pressure of "what scores well in eval" warps the content the way social media's optimization pressures warped what we read.

If you accept that, the goal of consuming the internet stops being "absorb everything" and starts being something more disciplined. You're trying to build a personal corpus that's good enough to support good output. The same problem the model has. Same constraints. Same trade-offs.

The strategy that works for the model also works for the human. Curate the corpus. Limit the input. Privilege primary sources. Notice when the platform is shaping the content. Stop reading when you've reached the point of diminishing returns. Use the fork.