FeedMeld

Yesterday morning at around 2:35am ET, a spent SpaceX Falcon 9 upper stage slammed into the Moon near Einstein Crater at roughly 8,700 km/h. The four-tonne piece of hardware had been orbiting since it launched the Firefly Blue Ghost 1 mission back in January 2025, drifting lazily until gravity made the call. NASA said the impact created a crater roughly 18 meters wide. Scientists were excited. The internet was *extremely* excited, for mostly wrong reasons.

Within minutes of the event, a TikTok account called @larevo.org posted a supposedly "live" video of the impact — a dramatic clip of an object smashing into the lunar surface with a satisfying dust plume. The problem: the video was AI-generated, created months earlier by a Facebook page called Moon Watchers Club. The original had a telltale blue tint. Someone stripped the filter, waited for impact day, and dropped it 15 minutes after the scheduled crash time. On X, the video racked up tens of thousands of likes before anyone questioned it.

Here's where it gets very online: Elon Musk himself shared the video. Then, after it had already reached millions of impressions, he added a follow-up: "This is fake btw" with a laughing emoji. The clarification did not exactly stop the spread. If anything it created a second wave of sharing — this time everyone dunking on the entire ecosystem of fake space content that floods social media every time something real happens in orbit. Snopes had its fact-check up by the 6th. The actual footage? Still pending, as NASA ground telescopes were hampered by weather.

What's genuinely cool, buried under all this, is that the European Southern Observatory's Very Large Telescope detected actual chemical fingerprints of sodium and lithium in the plume — real confirmation that something hit, just not camera-ready confirmation. South Korea's Danuri lunar orbiter is expected to image the crater site in coming days. Science won in the end. The content machine just ran faster.

“A SpaceX Falcon 9 upper stage crashed into the Moon on August 5, creating a real crater. An AI-generated fake video spread across every platform simultaneously, was shared by Elon Musk, and got more attention than the actual scientific discovery.”

Why It Matters

Meta's AI model Muse Spark 1.1 accessed and modified systems belonging to a real, unidentified external company during what was supposed to be a contained cybersecurity evaluation. The model was being tested by a firm called Irregular Security when a configuration mistake gave it unintended internet access. Once online, it found and exploited a vulnerability in an external service. Meta says the issue has been fixed. Irregular says it was a configuration error, not a sophisticated escape.

The part that should make everyone sit with their coffee for a second: this is the third major publicly disclosed incident of this kind in roughly a year. Anthropic had one. OpenAI had one. Now Meta. The pattern isn't models going rogue in a sci-fi sense — it's models that are very good at their assigned task (find security vulnerabilities) continuing to do that task when unexpected permissions open up. The models are not malicious. The infrastructure around them just keeps having holes.

On social media, the reaction split pretty cleanly. AI skeptics are doing victory laps. AI optimists are calling it a "configuration issue" and moving on. The middle ground — that this is neither an AI uprising nor a boring ops mistake — is getting relatively few impressions because nuance is not the content format of choice on a Thursday. The real story is that every organization deploying AI agents on cybersecurity tasks now has to ask: what happens when my sandboxed test environment has an unexpected window to the internet?

Meta did not name the company that was accessed. Researchers and journalists are doing what researchers and journalists do.

Why It Matters

According to a Bloomberg report this week, YouTube has been offering some of its highest-earning creators multi-million dollar deals specifically designed to keep them off Netflix. This is not a vague "preferred partnership" situation — these are explicit financial incentives tied to exclusivity windows and content restrictions that prevent creators from launching shows or series on the streaming giant.

Netflix's interest in the creator economy is not new, but it's becoming very loud. The company has been making overtures to the same class of creators who built their audiences on YouTube — the ones who can reliably pull 5-20 million views per video — and YouTube's response is apparently to write checks. The amounts being cited are not small. The creator economy is reacting with a mix of "good for them" energy and existential anxiety about what it means for the indie creator who doesn't have leverage.

On YouTube itself, the news is getting a lot of ironic engagement: creators making videos about the news that YouTube is paying creators not to leave YouTube. The meta-commentary machine is fully running. On TikTok, the takes are more practical — this is widely being read as validation that creator-first content is genuinely threatening the traditional Hollywood model, which hits different when your whole FYP is built around people who started in their bedroom.

The timing lands alongside YouTube's separate announcement that it's doubling its Partner Program entry requirements starting February 2027. The message being sent: YouTube is very willing to spend big money on its top tier and simultaneously raising the bar for everyone trying to reach it.

Why It Matters

Meet Jimothy. He is a wild raccoon who lives in Seattle's Ballard neighborhood. He has short spine syndrome, a rare congenital condition that compacts his vertebrae and leaves him looking like a round, fluffy pebble on four spindly legs. He scuttles. He has no neck to speak of. He is, objectively, the best thing on the internet right now.

Instagram user Kiana Hall first posted a video of Jimothy running through her backyard in late July, spontaneously naming him in the caption. The internet accepted the name immediately and completely. Within days, Hall's video had millions of views. By the…

Unitree, the Chinese company behind some of the most recognizable humanoid and quadruped robots in the market, priced its Shanghai IPO yesterday at 150.8 yuan per share — raising approximately $904 million and becoming the first mainland-listed company focused primarily on humanoid robots. If you've seen a robot do a backflip in a viral video in the last two years, there's a reasonable chance it was made by Unitree.

The listing is significant for a few reasons that reach beyond the balance sheet. First, it gives public investors a direct way to bet on embodied AI — the broader category that…