I used to think the danger of crypto Twitter was the obvious stuff. The pump-and-dump callers, the paid shills, the guy with the anime avatar telling you a coin does a 40x from here. Those are easy to spot and easy to ignore. The part that actually got me was quieter. I would open the app with no strong opinion on some token, scroll for twenty minutes, and close it feeling like I had done research. I hadn't. I had absorbed a mood.
The mechanism is worth understanding because it does not feel like manipulation while it is happening. It feels like learning.
Why the same idea from ten accounts feels like proof
Here is the trap. You see a thesis. Then you see it again, phrased slightly differently, from someone else. Then a third account quote-tweets the second one agreeing. Your brain does a thing it does automatically, which is treat repeated exposure as independent confirmation. Ten accounts saying the same thing feels like ten people who each did the work and arrived at the same place.
Most of the time it is one idea passed around a room. The second account read the first. The third is farming engagement off the second. The feed is optimized to show you things that get reactions, and agreement inside a cluster gets reactions, so you are served the cluster's consensus over and over while the disagreements get buried. What looks like the market's view is usually a few hundred people who follow each other retweeting a handful of original thoughts.
You cannot easily tell independent signal from an echo just by reading, because a good echo is indistinguishable from a chorus of experts. Someone confident, using the right vocabulary, getting likes. The format strips out the thing you actually need, which is whether these people reached the conclusion separately or copied it from each other.
Write first, then read
The single technique that fixed most of this for me is embarrassingly simple. Before I open the feed to think about a position, I write down what I already believe and why. Two or three sentences. What is my current view on this asset, what would change my mind, what price or event am I actually waiting on.
Then I read. And the magic is not that the feed suddenly becomes honest. It is that now I have a fixed reference point that the feed cannot quietly move. When I finish scrolling I compare. Did I learn a new fact, or did I just get more comfortable with a view I already had because a lot of people share it? Comfort is not information. If the only thing that changed is my confidence, and no new fact showed up, I made no progress and I should treat my conviction as exactly where it was before I opened the app.
This sounds trivial until you try it and notice how often you were about to size up a position on the strength of nothing but repetition. Logging the view first turns the feed from a place that forms your opinion into a place that tests it. Those are very different jobs and only one of them is useful.
A rough workflow that has held up for me:
- Write your thesis and your invalidation before opening any feed. If you cannot state what would make you wrong, you do not have a thesis yet, you have a hope.
- Read with one question in mind. Not "do people agree" but "is there a fact here I did not have."
- When a claim excites you, find the primary source. On-chain data, the filing, the actual announcement. If nobody in the thread links it, that is a signal about the thread.
- After reading, re-read your original note. Update it only for facts, not for vibes. Note the date and the reason for any change.
Curate like you are hiring, because you are
Following someone is a standing decision to let them into your head repeatedly. That deserves more thought than a tap. The accounts that earn a slot in my diet share a few traits, and the ones that get cut share the opposite.
I keep people who post their losses. Someone who only surfaces when they were right is running a highlight reel, and a highlight reel tells you nothing about their process or their hit rate. The trader who says "I got this wrong, here is what I missed" is showing you the part that matters. I also keep people who are specific. A precise, falsifiable claim is worth ten confident vibes, partly because you can check it later and see if they were right.
I cut the accounts that are always certain, because certainty is cheap and the market humbles everyone eventually, so permanent confidence is a tell that someone is performing rather than trading. I cut the ones whose entire feed is other people's takes with a one-word reaction on top, because that is the echo layer and I do not need a louder echo. And I cut anyone whose posting volume clearly exceeds the rate at which real new information exists, because past a certain point they are just filling airtime and I will absorb the filler as if it were signal.
One more rule that saved me a lot of noise. Separate the people you follow to learn from the people you follow to feel entertained or plugged in. Both are fine. Mixing them is not, because the entertaining ones are entertaining precisely because they are loud and confident and constant, and those are the exact qualities that manufacture unearned conviction. Put the sources you actually trade off of in their own list and read that list deliberately, not while doom-scrolling the main feed at midnight.
The failure mode to watch for
The classic way this ends badly is not a bad tip. It is a slow drift. You never made a decision to believe the popular thesis. You just marinated in it until disagreeing felt contrarian and vaguely stupid, and by then you were positioned for it without ever having stress-tested it. When the thesis breaks, and consensus theses break more or less on schedule, you find you have no idea why you were in the trade, which means you have no idea when to get out.
The fix is the same discipline as the entry. Your reason for holding should be written down and checkable, not a feeling that the smart accounts are still bullish. If your only remaining justification for a position is that people you follow have not capitulated yet, you are no longer trading your own book. You are holding theirs.
None of this requires quitting the feed. I still read it every day and I get real value from it. The point is to keep the feed downstream of your own thinking instead of upstream. Form the view, log it, then let the timeline poke holes in it. When I want the actual mechanism behind a claim, I go check on-chain flows and disclosure data directly, which is a lot of what we built Blockcircle to make quick, so I can confirm or kill a thesis with something more durable than a chorus of agreeing accounts. The habit is cheaper than the tools, though. Write first, then read, and most of the manufactured conviction never gets a chance to form.