EP 060 · 104 MIN IMAGES REACHED BASE CAMP 2 ON EVEREST.

Fast Hours 2025 Wrapped: Tools, Shifts, and Wake Up Calls

January 1, 2026104 min Rory Flynn & Drew Brucker

Watch on YouTube Spotify Apple Podcasts Audio file

The short version

Closing out Season 2, Drew Brucker and Rory Flynn rank 2025's biggest AI turning points: GPT-4.5 and 4o's image update, March's Gemini/GPT-image/Reve wave, VO3's all-in-one video-and-dialogue jump, and Nano Banana Pro, which Rory calls "AGI for images." They argue image models reached "Base Camp 2" on realism while video physics and voice still lag, and predict 2026 will reward constraint-based prompting, an eventual Midjourney edit model, and backlash against AI-generated streaming content.

Key takeaways

01
GPT-4.5 and 4o's image update is the moment Drew and Rory mark as 2025's real start, unlocking natural-language image edits in a chat interface.
02
Google had the best year of any lab in 2025 — Gemini 2.5, VO3, and Nano Banana Pro, which Rory calls "AGI for images."
03
Rory tracked roughly 105 major gen-AI releases from top labs in 2025 alone, about one every two and a half days.
04
Images hit "Base Camp 2" on realism and control this year, but video is stuck on consistency, physics, 10-second length limits, and bad AI voice.
05
2026 prompting will be about constraint, not addition: Rory predicts negative prompting becomes more important than describing everything you want.
06
Even simple edits still break: Nano Banana Pro kept reverting a brand's updated logo to the old one, no matter how the prompt was written.
07
Rory's advice for creators heading into 2026: don't try to master every tool, pick an industry, vertical, or outcome-specific niche instead.

Receipts · said on the record

“And I would just say like overall realism, like you pump in Nano Banana Pro. There are photos basically indistinguishable from reality that we've got there with that, right?”

Drew Brucker Drew on why Nano Banana Pro marked such a big jump in realism

“Images this year went from like here to like, like if we're climbing Mount Everest, they went to like Base Camp 2. It's almost close to the peak.”

Rory Flynn Rory's metaphor for how far image models progressed in 2025 versus video

“So a brand updated their logo and trying to retexture or create new scenes with this logo, even by providing it with the new logo, don't change the logo. Lots of rules, mega prompt around it even.”

Drew Brucker Drew on Nano Banana Pro reverting a client's updated logo back to the old version

“2026, it's going to be more about constraint than it is going to be about achieving, if that makes sense. Like when you're writing prompts, instead of like, do all this, it's like, don't do this because the capabilities are only going to grow.”

Rory Flynn Rory's prediction that 2026 prompting will be about constraint, not addition

“This is an opportunity, still a very, very open window. And again, you don't have to focus on everything. Focus on the things that interest you most and get into that niche because there's really not anybody that are in the niche.”

Drew Brucker Drew's closing advice to creators heading into 2026

Chapters

Questions this episode answers

What was the biggest AI turning point of 2025, according to Fast Hours?

Drew and Rory point to two clusters: GPT-4.5 and 4o's image update in early spring, which unlocked natural-language image edits, and Google's run of Gemini 2.5, VO3, and Nano Banana / Nano Banana Pro, which Rory calls "AGI for images."

How does Fast Hours rate 2025's progress on AI images versus AI video?

Rory says images reached "Base Camp 2" on Mount Everest, close to photorealistic and controllable, while video "got to VO3 and then nothing really got that much better," still held back by consistency, physics, 10-second length limits, and weak AI voice and dialogue.

What does Rory Flynn predict for AI prompting in 2026?

Rory predicts prompting shifts from addition to constraint. Instead of listing everything you want, negative prompting, specifying what a model should not do, becomes more important as raw capability keeps growing.

What's Fast Hours' advice for creators heading into 2026?

Drew says it's still early enough to carve out a lane, but recommends picking an industry, vertical, or outcome-specific niche rather than trying to master every tool, since almost nobody has gone deep in any one niche yet.

Does Nano Banana Pro have any weaknesses?

Yes. Drew and Rory both describe frustrating edit failures, such as Nano Banana Pro repeatedly reverting a brand's updated logo back to the old version, or adding visible noise to a simple flat gradient background despite detailed prompting.

Show notes

In this final episode of Fast Hours in 2025, Drew Brucker and Rory Flynn zoom out to dissect what actually mattered this year across Midjourney, Nano Banana Pro, ChatGPT Image 1.5, Weavy, video models, workflows, and the uncomfortable truth about how fast all of this is moving.

They unpack the real inflection points no one labeled at the time. Why March quietly changed everything. Why Nano Banana Pro rewired image editing expectations. Why Veo 3 reset video. Why Midjourney still feels magical Why workflows (not models) are becoming the real creative advantage.

Along the way, they spiral into mood boards, personalization hacks, node-based systems, AI video limitations, why Hollywood feels creatively bankrupt, how Grok quietly became a research weapon, and why Midjourney’s next move might determine whether it stays an artist’s playground or becomes a professional tool.

It’s opinionated. It’s nerdy. It’s honest. It’s occasionally unhinged. And it’s the clearest snapshot of where AI creativity actually stands heading into 2026.

If you’re trying to keep up, slow down, or figure out where to place your bets next year, this episode is your unfair advantage.

Make sure you stay for the whole episode—it'll be worth it.

Every word Read the full transcript