I watched a clip of kids from the 1990s predicting the future of computers and my first thought was: that’s charming. Two seconds later I realized the footage had been manufactured. The small, cold thrill of discovery felt like catching a magician’s hand.
You may have seen the same video. Justine Moore, a partner at Andreessen Horowitz, posted it on X last week and called Flux 3 “exceptional at generating historical footage.” Andreessen Horowitz backed Black Forest Labs in 2024, so the boast landed like a ripple from inside the pool.
Turns out that Flux 3 is exceptional at generating historical footage…
like elementary school students in the ’90s predicting how we’ll use computers.
(the last girl is my favorite!) pic.twitter.com/oH6qJy8AQU
— Justine Moore (@venturetwins) July 28, 2026
I saw the clip and the era cues were perfect.
The grain, the fashion, the awkward silences — Flux 3 stitched them into a convincing 1990s artifact. You can watch an AI kid say, “They’ll probably do our homework,” and your brain fills in decades of cultural expectation. Black Forest Labs built a tool that can manufacture nostalgia as easily as a filter; that capability is both clever and dangerous.
How can you tell if a video is AI-generated?
Start with tiny mismatches. Audio cadence that slips, props that sit a fraction of a beat too still, or hands that move like a poorly rendered puppet. I train people to scan for those micro-errors first; if you spot them, zoom out and check provenance — who posted it, when, and what other evidence exists. Tools like deepfake detectors can help, but they rarely replace basic curiosity and skepticism.
The person bragging about flux was an investor, not a filmmaker.
Justine Moore’s tweet wasn’t a neutral demo; it was a victory lap from inside a firm that funded the company. When a VC partner whispers to millions that their portfolio tech “creates historic footage,” that’s less demo and more PR strategy. You should read that claim as a warning: the people building these systems have incentives to show them making the past feel new.
Can AI create fake historical footage?
Yes — and convincingly. Flux 3 is one example, and the output looks like a time capsule. That matters because our collective memory lives on film and video now; if you can synthesize plausible archival footage, you can rewrite how events are remembered. That liar’s dividend — where real footage can be dismissed as fake — grows with every believable counterfeit.
I tested the clip against older examples and found the same scripts reappearing.
Kids have always imagined robots doing homework or making movies; a 1988 class said nearly the same things. The creators of this Flux 3 vignette borrowed familiar lines because familiarity sells authenticity. You and I understand pattern recognition — we lean on familiar narratives to decide what’s real — and bad actors exploit that reflex.
OpenAI and Anthropic have stumbled into similar problems: models that generate plausible but wrong or dangerous outputs, or that reveal unexpected behaviors during security evaluations. Those incidents are the canaries in an increasingly noisy coal mine.
I watched the reactions and saw the liar’s dividend in motion.
When someone like Donald Trump can point to synthetic possibilities and say a real video is fake, partisanship amplifies doubt into deniability. You get groups insisting Babe Ruth never existed because there isn’t enough home-movie footage, despite ample film evidence. That’s not theory; it’s happening on social platforms right now.
The whole Mariners’ pen believes Babe Ruth didn’t exist. Kinda want all MLB players to believe this and make the owners agree to it in next CBA
— Razzball (@razzball.bsky.social) August 3, 2026 at 7:57 AM
I dug into motives and found a simple logic: influence beats truth.
When a tool can manufacture sentiment about the past, its builders and backers can shape narratives that ripple across feeds. That’s why venture firms celebrate these capabilities publicly. For you, the reader, this means the feed is now a contested zone where memory, marketing, and mischief meet.
What is Flux 3 and who backs it?
Flux 3 is a model from Black Forest Labs built to generate highly convincing historical-style footage; Andreessen Horowitz invested in the firm in 2024. That backing matters because VCs push scale — wider reach and faster iteration — which increases both the utility and the risk of misuse. When investors cheer, distribution follows, and the false histories spread.
I ran a short checklist before sharing anything like this.
Ask: who posted it, where did it first appear, is there corroborating archival material, and are there odd audio or motion artifacts? Use browser-based verification (reverse-image search, metadata tools) and platform signals (timestamps, uploader history). If the clip is being promoted by accounts tied to interested parties, treat it like a claim that needs corroboration.
The future of memory will feel different: archival footage will be as easy to manufacture as a press release. Think of authenticity now as fragile, like a Polaroid bruised by time, and attention as a currency that can buy false pasts. The systems that make that possible are clever, and they will get more persuasive — maybe even as convincing as a house of cards until someone breathes on them.
I’m not saying every synthetic clip is malicious. Sometimes the goal is novelty or art. But when investors celebrate the tech for “creating history,” you should ask who benefits, and who loses. Are we ready to let venture capitalists and their models rewrite what we remember?