June 23 2026 - A24 + DeepMind: AI Is Rewriting FilmmakingA24 + DeepMind: AI Is Rewriting Filmmaking

**Taylor** (0:00)
Hey everyone, welcome back to AI Signal & Noise, your daily dose of everything artificial intelligence. It is Tuesday, and oh my gosh, we have some absolutely wild stories to get into today.

**Morgan** (0:14)
Hey there. Yeah, today is definitely packed. But let us try to separate the actual signal from all the marketing noise.
What is our first big story, Taylor?

**Taylor** (0:25)
Okay. So first up, Hollywood is officially getting a massive AI shakeup. Google DeepMind is partnering with A24, the studio behind Civil War and everything everywhere all at once.

**Morgan** (0:39)
Wait, A24, the indie darling studio? That is a huge deal. But what is the actual nature of this partnership?
Are they just letting AI write the scripts now?

**Taylor** (0:52)
No, not exactly.
According to The Decoder, it is a long-term research partnership. And get this, the Wall Street Journal says Google is investing like $75 million in A24.

**Morgan** (1:07)
$75 million? Wow, that is not pocket change. But I am always skeptical when tech giants pour money into creative spaces.
What are they actually building together?

**Taylor** (1:20)
Well, they are calling it AI filmmaking research. They want to see how DeepMind's video and audio models, like Veo, can actually help filmmakers in the real production pipeline. It is so cool, dude. Right.

**Morgan** (1:34)
So it is probably about VFX, pre-visualization, maybe editing. But honestly, A24 fans love them because of their raw, human-driven storytelling. Won't AI tools alienate their core audience?

**Taylor** (1:48)
I mean, maybe some purists will complain, but honestly, A24 has always been about pushing boundaries. This is just the next logical step for them, right?

**Morgan** (1:59)
I guess so. But there is a fine line between experimental art and just pushing a button to generate a scene. The human element is what makes their film so emotional.

**Taylor** (2:09)
That is a fair point. But think about the budget side. Indie filmmakers could use these tools to create massive sci-fi worlds without needing a $200 million budget. It democratizes the process.

**Morgan** (2:23)
Or it just floods the market with cheap generated content. I guess we'll have to wait and see what kind of films actually come out of this partnership.

**Taylor** (2:33)
Totally. But I am super excited to see what A24 directors do with it. They are always so weird and experimental anyway. It is going to be wild. True.

**Morgan** (2:43)
If anyone can make AI film tools look artsy and cool, it is probably them. Let us move on. What is next on the list?

**Taylor** (2:51)
All right. Next up, we have some massive news for the developers out there. Sakana AI just launched something called Sakana Fugu.

**Morgan** (2:59)
Sakana Fugu? Like the puffer fish? Okay. That is a unique name.
What does this model actually do, Taylor? Is it just another LLM?

**Taylor** (3:10)
Nope. It is actually an orchestration model.
According to Mark Tech Post, Fugu routes tasks across a swappable pool of frontier LLMs. It is like a super smart traffic cop for AI.

**Morgan** (3:25)
Interesting. So instead of using one massive model for everything, it decides which model is best for a specific prompt and routes it there?

**Taylor** (3:34)
Exactly. They have Fugu and Fugu Ultra. They are apparently leading most coding, reasoning and agentic benchmarks because they can swap models out on the fly.
It is so smart.

**Morgan** (3:48)
But wait, isn't routing expensive and slow? If you have to analyze the prompt first, then send it to another model, doesn't that add a lot of latency?

**Taylor** (3:58)
You would think so, but they designed it to be super efficient.
The cool part is the swappable pool. As new, better models come out, you can just plug them right in.

**Morgan** (4:09)
Okay. That part is actually pretty clever.
If future proofs your AI stack, you don't get locked into OpenAI or Anthropic, you just let Fugu route to whoever's winning the LLM race.

**Taylor** (4:21)
Yes. Dude, that is exactly it. It prevents vendor lock-in, which is a massive headache for enterprises right now. Sakana is really cooking with this one.

**Morgan** (4:32)
It makes sense on paper. I just want to see how it handles complex, multi-step, agentic tasks under heavy load.
The routing overhead could still be a bottleneck.

**Taylor** (4:43)
Fair enough, but the benchmarks look super promising. Speaking of multi-step agents though, our next story is about how the AI world is getting, well, loopy.

**Morgan** (4:55)
Loopy? What does that even mean? Are the models losing their minds or is this some new buzzword we have to tolerate?

**Taylor** (5:04)
No, it is actually a really cool concept. TechCrunch wrote about this.