A few months ago, I found myself in an odd position — with some time off. A lot of it. Naturally, I decided to use the time not to disconnect, but instead to double-down on tech news. I set out to achieve “Pocket Zero” and catch up on nearly everything I had saved to read later but never got to over the past year or so.

A few dozen articles in, a pattern started to emerge. By the time I had read a few hundred articles, I had my thesis for what I’ll call the “75 — 20 — 5 Rule”.

That is, on any given day, I’d say 75 percent of what you read in the tech press is somewhat accurate, 20 percent is complete bullshit, and 5 percent is actually true.

This may sound like an exaggeration. It is not. At least not from what I can gather from these hundreds of articles written over the past year-plus. For some of the stories, I had first-hand knowledge of a situation being inaccurately reported on. For others, time simply exposed the inaccuracies. Combined, I feel like I have more than enough data to feel comfortable with my thesis.

And really, this is something I’ve been dwelling on for a year-and-a-half now. Time is proving that things are getting worse, not better.

You probably want some examples, but really, almost everything you read is an example. Here’s a quick and obvious one in recent memory. Back in May, on the topic of Waze:

Especially because multiple credible sources contacted by TechCrunch have told us that the Google acquisition rumour is not at all true.

A couple weeks later? Whoops.

One has to wonder if those “multiple credible sources” are the same ones that told PandoDaily recently all kinds of juicy details about BeachMint falling apart. Whoops.

I’m picking on TechCrunch and PandoDaily here because I know both fairly well. I still write for TechCrunch from time to time and I’m still technically an investor in PandoDaily through CrunchFund. Both, of course, are still capable of doing good work and both often do. Yet both are also key parts of a system where the gears of truth are calcifying with bullshit.

Kevin Roose of New York Magazine thinks the solution is for tech blogs to fully pivot into trade publications that mainly serve to celebrate the industry they cover. I disagree only because we can already see the results of such a maneuver. All the worst posts on these sites are the ones that are little more than glorified re-writes of press releases.

And it seems to me that the opposite of Roose’s idea is actually happening. Rather than celebrate anything, a lot of the tech press is now bogged down with extreme cynicism of everything. It’s depressing.

Meanwhile, the other half of our press is well on their way towards a sort of Idiocracy future, where tragedies with only ridiculously tangential tech angles (“Hey, it was talked about on Twitter!”) are turned into GIFs for pageviews.

Still, I hold out hope. Some sites do seem to pride themselves on focusing on the “5 percent” — you know, the truth. AllThingsD remains particularly good here, based on what I gathered from my summer reading binge. The mainstream tech press also tends to be more accurate — though they also occasionally shit the bed badly at times. And, of course, they’re often pretty slow.

There has long been a “speed versus accuracy” debate within the tech blogosphere. When I was in that world, I was definitely in the “speed” camp. Get something out there and let the truth reveal itself — process journalism, baby. If the readers aren’t comfortable with that, let them read elsewhere.

But my fear now is that we’re veering too far into the world of half-truths and straight-up bullshit. Everything reported on, no matter how inaccurate is often taken as gospel and spread further. Speed and exaggeration have won, accuracy and nuance are nearly dead. It’s not quite another age of yellow journalism yet, but we’re getting there.

This is all made worse by the fact that it’s increasingly rare that any of these sites are held accountable for inaccuracies. The PandoDaily situation is a total outlier. Those with knowledge of the actual truth rarely speak out because, what’s the point? In our age of breakneck news cycles, a story will likely be forgotten about the next day unless you make a stink about it.

And, quite often, there’s an inverse situation where a company/investor actually wants an inaccurate story circulated for various reasons (spur a bidding war, etc). It’s gamers being gamed by gamers. The only losers are the truth and us, the readers.

I wish I had some solution. I don’t. Maybe a natural purging of the forest is coming. Maybe not. I’d just warn everyone who loves reading tech news as much as I do to no longer take it with just a grain of salt, take it with the entire canister of salt. Maybe desalt the ocean and use that.The likelihood that what you’re reading is completely accurate is very, very, very small. There may be some truth there. There may be none. Just know that.

Tech reporters, don’t be lazy. You can talk yourself into any truth, but an absence of facts is not a validation of anything beyond the fact that you have bad sources. The best way to get to the truth is often to triangulate. The liklihood that you’re going to get all the information you need from one source is infinitesimal. Or they’re chatting you up because they’re spinning you. Get bits at a time. Play that information off each other. Get creative.

Don’t write something because you can. Write something because you should. Or don’t write anything at all.