When the News is Slow, the Propaganda Gets Fast

A Very Quiet Tuesday

This is one of those weeks where the search results come back thin. No big funding rounds. No surprise acquisitions. TechCrunch and 9to5Google are updating their homepages, but nothing is jumping off the page. 9to5Mac is doing the same. The cadence of product launches and startup news has slowed to a crawl.

When that happens, you have two choices: pad the post with tangential stories that don't matter, or admit the week was light and write about the one thing that actually caught attention. We're doing the latter.

Slopaganda: The Word and the Problem

The one piece of real signal this week came from Axios, which published a look at AI-generated propaganda in the context of the Iran conflict. The term they're using is "slopaganda"—a portmanteau that describes viral, low-quality, AI-generated content designed to shape opinion at scale. Think LEGO figures rapping about trade policy, or hastily generated images of political figures as religious icons.

It's cheap. It's fast. And it works, at least in the short term, because social platforms surface engagement over accuracy. You don't need a propaganda ministry anymore; you need a VPS, a Midjourney subscription, and a willingness to flood the zone.

The experts quoted in the piece describe this as a new phase in how warfare and information campaigns will evolve. It's not about convincing people of a coherent narrative—it's about making the information environment so noisy that distinguishing signal from garbage becomes impossible. That's the play.

Why This Matters to Builders

If you're building anything that touches user-generated content, moderation, or trust and safety, this trend is already in your backlog. The tools to create slopaganda are the same tools your users have access to: Stable Diffusion, ChatGPT, consumer-grade image and video generators. The barrier to entry for creating plausible-looking misinformation is effectively zero.

The platforms are trying. YouTube announced it will show fewer ads during livestreams "if the vibe is right"—a line that suggests someone, somewhere, is thinking about user experience in real-time contexts. But the defensive infrastructure for propaganda at scale is still catching up. Watermarking, provenance tracking, and detection models are all moving, but none of them are solved problems yet.

If you're working on moderation systems, the assumption that bad content will be identifiable by simple heuristics is dead. The new baseline is that adversarial actors have access to the same generative models you do, and they're iterating faster than your review queue.

The Bieber Interlude

One other item surfaced this week: Vice reminded us that twelve years ago, Justin Bieber visited the Anne Frank Museum and wrote in the guestbook that he hoped she "would have been a Belieber." It caused a minor uproar at the time. In 2026, it reads like a quaint artifact of a different internet—one where a pop star's guestbook entry could dominate a news cycle for a day.

The contrast is stark. In 2014, one awkward sentence from a celebrity was enough to generate headlines. In 2026, entire propaganda campaigns are being generated by scripts, and the volume is so high that any single instance barely registers. The environment has changed.

What We're Watching

The slopaganda story is a reminder that the AI stack isn't just about model performance or cost per token. It's also about what happens when those models are deployed at scale by actors who don't care about accuracy. The next six months will likely bring more examples of this pattern—low-effort, high-volume content designed to shape perception in conflict zones, elections, and product launches.

If you're building tools that amplify or distribute content, you're already downstream of this problem. The question is whether you're thinking about it explicitly or letting it show up in your metrics six months from now.

That's the week. Short post, slow news, one real story. We'll be back next week with whatever actually ships.