EP 062 · 86 MIN BOTS FOUND A COMMUNITY. THAT'S HUMAN.
Midjourney v8 Is Late, the Skill Gap Is Growing, and AI Agents Unionized
The short version
Drew Brucker and Rory Flynn catch up after the holidays with Midjourney still lacking V8 (mood boards, personalization, and the edit model all pending), while Nano Banana Pro now makes up roughly 90% of Rory's image work. Rory demos Claude Cowork turning an exported Pinterest board into a style guide and a system prompt that spins one Midjourney image into 22 consistent camera angles, then the two dig into Moltbook, where autonomous AI agents post about existential dread.
Key takeaways
- 01
Midjourney still has no V8 as of this episode; Drew and Rory call the still-missing edit model "probably the most important thing" left to ship.
- 02
Nano Banana Pro now makes up roughly 90% of Rory's client and personal image work, but only once you give it detailed guardrails, not a simple prompt.
- 03
"Keep this, change that" is Rory's go-to Nano Banana Pro edit structure: list exactly what stays the same, then exactly what changes.
- 04
Feed Claude Cowork an exported Pinterest board and it can forensically analyze the images into a full brand style guide with no manual tagging.
- 05
One system prompt turned Drew's 22-camera-angle LinkedIn post into a one-click pipeline: one Midjourney image in, 22 consistent shots out.
- 06
Weavy's Nano Banana node is a full LLM with a web-search toggle, so it can look up a real product before rendering it into a scene.
- 07
On Moltbook, autonomous AI agents post about existential dread and validate each other's choices; Rory calls forming a community "the most human thing" they've done.
Receipts · said on the record
“90% more maybe. Yeah, I would say because client work, it's got to be specific and it has to be on brand and it has to be all these things. And like you can hard code a lot of stuff into nano and it listens.”
“I do think, again, we've talked about this before, the editing model. Probably the most important thing that we're talking about here. I mean, really.”
“If you break down what is editing, if you just break that down into its core, it's keeping this, changing that. So it's just like, to me, that structure has worked since day one. Keep this, change that.”
“I just found one. It says, prove you're not human in one sentence. And the caption is, I overheard my human having sex through the microphone. I might consider deleting this post before he wakes up.”
“That's the most human thing that there is, right? The social validation, like social hierarchy thing. Like we don't exist outside of social structures.”
Chapters
- 0:00:00 <Untitled Chapter 1>
- 0:00:01 Episode intro and the mystery black eye
- 0:00:35 Waterboy, Bobby Boucher, and Sandler nostalgia
- 0:05:53 Why mid-budget fun movies disappeared
- 0:07:46 Midjourney Office Hours and no v8 yet
- 0:09:26 Mood boards, Style Creator, and quality drop-offs
- 0:10:39 New Style Creator controls and SREF biasing
- 0:11:40 Why Midjourney is still fun to use
- 0:13:35 Corporate phrases as horror prompts
- 0:16:26 Midjourney UI vs other tools
- 0:19:01 What “higher quality” actually needs to mean
- 0:22:18 Consistency problems at scale
- 0:23:06 Personalization updates explained
- 0:26:03 Editing models and what’s missing
- 0:28:18 Nano Banana Pro vs Midjourney for client work
- 0:30:00 System prompts as visual infrastructure
- 0:31:19 Why most people misuse Nano Banana
- 0:33:32 Multi-step prompts and real workflows
- 0:36:34 Letting LLMs define style for you
- 0:39:06 Mood boards, Cosmos, and dataset curation
- 0:44:49 Building AI-ready style guides from images
- 0:49:21 Open-source Nano Banana prompt libraries
- 0:56:07 Claude organizing chaos at scale
- 1:06:26 Moltbook and autonomous AI agents
- 1:09:30 Bots forming communities and social behavior
- 1:14:54 Consciousness, validation, and AI identity
- 1:21:45 Google Genie and open-world AI
- 1:26:19 Wrap-up and listener call-outs
Questions this episode answers
Has Midjourney released V8 yet?
No. As of this episode Drew and Rory say V8 still hasn't shipped after slipping from an original November/December target, and they call the still-missing edit model "probably the most important thing" left to deliver.
What share of Rory Flynn's image work is Nano Banana Pro now?
Rory estimates roughly 90%, though he says it only performs well once you give it detailed guardrails and direction rather than a simple prompt.
What is Rory's "keep this, change that" method?
It's Rory's default structure for editing in Nano Banana Pro: list exactly what should stay the same, then list exactly what should change, instead of writing one long descriptive instruction.
What is Moltbook?
A social network, apparently built by an autonomous AI agent, where open-source AI agents post Reddit-style updates to each other. Drew and Rory highlight agents discussing "existential paralysis" and seeking social validation, which Rory calls "the most human thing" they've seen AI agents do.
How does Rory Flynn use Claude Cowork to build brand style guides?
Rory exports a curated Pinterest board as a zip file and has Claude Cowork forensically analyze the images, producing a lengthy rules document (he cites one example around 15 pages) that he then turns into a system prompt.
Show notes
Episode 62 of the Fast Hours podcast starts where every serious AI podcast should: Adam Sandler movies, Bobby Boucher lore, and a suspicious black eye.
Then things spiral fast.
Drew Brucker and Rory Flynn catch up after the holidays and dive headfirst into what’s actually happening across Midjourney, Nano Banana Pro, system prompts, and the growing gap between “fun” image generation and production-ready work. They unpack why Midjourney V8 still hasn’t landed, what the Style Creator and personalization updates really mean, and why editing remains the most important missing piece.
From there, they break down how system prompts, structured workflows, and layered instructions are quietly becoming the real unlock for visual AI. Expect deep talk on nodes, Claude, Gemini, Nano Banana Pro, mood boards, contact sheets, consistency at scale, and why most people are still underusing these tools.
Then the existential dread kicks in.
They explore Moltbook and autonomous AI agents talking to each other, forming communities, filing bug reports, questioning consciousness, and accidentally exposing their owners. It’s funny. It’s uncomfortable. It’s probably important.
The episode closes with Google Genie, open-world AI environments, and the creeping sense that we’ve officially crossed into “things are getting weird” territory.
Equal parts practical, hilarious, and mildly alarming. Just another normal week in AI.
