Knowledge Hoarding Died With ChatGPT
For most of my adult life, I collected knowledge the way some people collect stamps: obsessively, indiscriminately, and mostly without using any of it. Before the generative AI boom, this strategy at least felt rational. Information was scattered, gated, expensive. Downloading courses on software engineering, data science, devops, trading, investing — hoarding it all on hard drives — seemed like building an asset. When the boom came, the asset repriced to zero overnight.
That is the strange grief of watching ChatGPT arrive with a head full of pirated torrents. Everything I had stockpiled — every course folder, every saved PDF — became something a stranger could summon in ten seconds. The collection did not just stop appreciating. It stopped being a collection at all.
The strange part is that the knowledge was never entirely useless to me. Talking to peers from the same background, even graduates of good universities, I kept discovering gaps everywhere. Topics I considered basic drew blank stares. Most people I knew were uncomfortable in a terminal; many avoided Linux entirely. This created an odd, dishonest feeling: externally, nothing to show — no products, no measurable output — yet socially, a quiet sense of superiority that I could not decide was earned or imagined.
That social asymmetry distorted my picture of reality. Because my circle rarely knew what I knew, I assumed the world worked the way I did: that everyone collects information constantly, that terminal fluency is table stakes, that a large private library is normal. It is not. I was calibrating against a sample size of people who happened to be less prepared than me, and mistaking that for evidence of advantage.
The real cost shows up when you try to point at anything you have made. A hoarder’s portfolio is empty by definition. Years of intake, almost no output. And the harder question follows: if a single model can now reproduce most retrievable knowledge on demand, then what exactly was the hoarding for? The value of merely knowing things has collapsed toward zero. What remains scarce is judgment — deciding which questions matter, verifying what the machine produces, connecting fields that do not touch.
There is a second-order fear too. If information abundance eliminates jobs the way it devalued collections, then the long project of saving toward independence starts to look shakier than the personal-finance books promised. Maybe universal income arrives, maybe not. Either way, the assumption that accumulating capital-in-knowledge guarantees safety has been falsified once already within my lifetime. It can be falsified again.
So what replaces hoarding? A few working conclusions:
- Collect only what you are about to use. The half-life of stored knowledge is now shorter than the time it takes to store it.
- Spend effort where a language model cannot substitute: taste, verification, original synthesis, actually shipping things.
- Treat “knowing about” as nearly worthless and “having done” as the only durable signal.
The mistake was not studying too much. It was studying as a form of procrastination that felt like virtue — collecting against an imaginary future scarcity that technology erased. The library burned down, and it turned out I never needed most of the books. What I needed was the writing.
Drafted with AI assistance from my personal journal (2025-05-30), then edited by me. See On AI Assistance.