Unleash Possible

The episode surfaces a handful of genuinely practical tips - using both complex and stripped-down prompts in parallel, building a prompt library with terms like 'editorial style illustration', and the Pixlr black-and-white-plus-brand-color technique - but the majority of the runtime is casual enthusiasm, filler, and mutual commiseration about being beginners rather than dense, transferable advice.

“what I'll often do is I'll take a really complex prompt...and I'll pop that in as a prompt. I'll also do a really simple one...and then I'll just add illustration to the end of each one”

“bring the image into Pixlr and turn the image into black and white...add your brand color on top of it. Change the layer settings so that the image is showing through the brand color”

The episode offers no contrarian or first-principles arguments; the main frames (Midjourney is artistic by design, prompting is unpredictable, iterate like you would with a human designer) are widely circulated takes in the AI image community, and the schizophrenia AI anecdote is interesting but tangential rather than structurally generative.

“with artificial intelligence, you know, it is very much like a person in terms of we don't know what a person is thinking”

“somebody managed to create an artificial intelligence that could predict with incredible accuracy if somebody was going to develop paranoid schizophrenia. But the tool turned out to be completely useless because they never were able to reverse engineer how the AI was actually figuring it out”

Frank explicitly positions himself as a non-expert enthusiast who experiments on personal blog posts distributed via LinkedIn Newsletter and Medium; there is no evidence of deploying these tools at organisational scale, managing a team, or producing measurable B2B business outcomes from AI imagery.

“I'm not an expert in AI and actually that's kind of why it's interesting that I'm here right now”

“I've just been experimenting, playing, having fun, and then just trying to figure out, like, how can we actually use these tools”

The episode does name specific tools (Midjourney, DALL-E 2, Pixlr, Photoshop, Canva), a concrete post-processing workflow, and a specific business use-case (one long-form blog post per month with six images), but there are no business metrics, time-savings data, cost comparisons, or before/after evidence of impact to anchor the claims.

“I can do six images for my blog post...in way less than the time it was taking me to do one image”

“Meta Facebook are working on one at the moment. Google are working on one called...Imagen”

The host asks a few genuinely useful tactical follow-ups (notably 'what do you use today to make the tweaks?') and draws out a concrete workflow, but she frequently injects self-deprecating commentary and frames questions as open invitations ('what other advice would you give people?') rather than pressing on claims or introducing productive friction.

“what do you use today to make the tweaks? Are you using a design tool or filters or how are you doing it?”

“What other advice would you give people?”