AI for Creatives — Week of July 17, 2026
A rather slow week for AI (creatively speaking)
Better thinking. Better work.
Every week I take a look at some of the tools and ideas emerging in AI. And I focus on the things that are important and useful for creatives and entrepreneurs. No insider valuations or who is doing what… just stuff you can use today. I use Claude CoWork to help compile this list.
Not a heavy week for actual tools. Most of what landed in the inbox was policy talk, hardware rumors, and executives sniping at each other. Here’s what’s actually usable if you make things for a living.
Peter Hurley Says the Middle of Portrait Photography Is Going Away
Peter Hurley, the celebrity headshot photographer who coined “the squinch,” sat down with The Rundown this week and said something worth listening to: “if you can’t beat the AIs in portrait photography, you won’t get hired.”
That’s not defeat talk; it’s math. AI portrait apps already produce results good enough for most people who just need a decent headshot. What survives that shift is the top tier, photographers who bring taste, timing, and a read on real expression an app can’t fake.
Hurley’s read is that demand concentrates hard at the top of the field while the middle gets automated out.
Ten times more work for the best 1 to 5 percent, a lot less for everyone coasting on “I own a camera and a light.”
There’s nothing to click here, no tool to try. It’s a warning about where to put your effort: not into competing with AI on speed or price, but into the things AI still can’t touch.
(via The Rundown AI)
(BTW, same thing in product photography. Gotta do stuff better than the mid-level AIs to get that low-hanging, client doesn’t really care, work. I think blending great photography with some AI backgrounds will keep a lot of photographers busy, though.)
Reve’s New Image Model Gives You Real Editing Control, Not Just Regeneration
Reve pushed an update (Reve 2.1, following the earlier Reve API) that does native 4K generation with element-level editing. You can change one part of an image, a background, a prop, a hand, without re-rolling the whole frame and hoping the rest survives.
That’s the complaint every photographer and designer has had with AI image tools since day one: you get 90 percent of what you want and the model wrecks the other 10 percent trying to fix it. Reve’s edits stay contained to the region you’re actually touching.
Worth a test if you’re doing comps, mockups, or concept work where you need to iterate on one element without redoing the whole frame. Try it at preview.reve.art.
(via The Neuron and The Rundown AI)
Runway Opened a Single API for Its Whole Video Stack
Runway launched Runway Dev, one API giving developers access to Runway’s frontier video models plus third-party image and audio generators and real-time avatar tools, all through one integration point.
If you’ve been waiting for a cleaner pipeline between “AI generates a shot” and “AI generates the whole sequence with sound and a talking presenter,” this is the plumbing for it. Separately, Decart shipped Lucy 2.5, a model that edits video mid-stream instead of after the fact, useful for prototyping motion concepts without waiting on a full render to see if an idea works.
(via The Rundown AI)
A Better Way to Write AI Prompts: Put the Instructions Last
Lea’s Excellent AI Prompts newsletter laid out what she calls the “2026 Master Delegation Contract,” a prompt structure built for getting finished work out of a model instead of a rough draft you have to fix yourself.
The structure runs:
Outcome (the actual result you need),
Context (why, and what time period this reflects),
Inputs (which sources are authorized, and which material should be treated as raw data rather than fact),
Constraints (measurable limits, plus a request for actual prose instead of bullet-pointed mush),
Permissions (what the model can decide alone versus what needs your sign-off),
Validation (a defined “completion bar,” meaning the model checks its own work against disconfirming evidence before calling it done), and…
Output (the exact format you want back).
Put your instructions at the end of a long prompt, not the beginning; models weight the end more heavily on long-context requests. And label your source material “Confirmed,” “Inferred,” or “Unknown” so the model doesn’t quietly treat a guess as a fact. Use this on your next client brief or project proposal and you’ll get fewer rounds of “no, not like that.”
(via Excellent AI Prompts)
You really should subscribe)
The “AI Art Is Effortless” Myth Keeps Getting Debunked
The Neuron flagged a video this week called “ComfyUI Proves AI Art Is Not Zero Effort,” worth watching if you’ve ever had to defend AI-assisted work to a client or a skeptical peer.
ComfyUI’s node-based interface is the opposite of typing a sentence into a box and hoping. It’s a visible pipeline: technical choices, iteration, and the kind of creative intent that a prompt box hides.
If you’re getting the “it’s just one click” line from someone who’s never used these tools, this is the counterargument on video.
One more worth a line: Anthropic is giving verified U.S. K-12 teachers a free year of Claude with lesson-planning tools built in. If you run workshops or know teachers who could use it, pass it along.
(via The Neuron)
Prompt of the Week
Find the Automatable Friction in Your Day
From Excellent AI Prompts: record one ordinary work day start to finish, then feed the log to Claude or ChatGPT with this:
“Review this record of my day. Identify repetitive tasks, unnecessary decisions, information I repeatedly search for, things I forget, systems that do not communicate, and moments when I express frustration. Rank the opportunities by time saved, frequency, ease of implementation, and impact on my quality of life.”
Run it on a shoot day, an edit day, or a client-onboarding day. You get a ranked list of what’s actually worth automating instead of guessing.
(via Excellent AI Prompts)
A couple of extra things today:
Noted:
When George Eastman set out ‘to make photography an everyday affair, to make the camera as convenient as the pencil,’ professionals heard something very different
‘If any schoolboy or girl could operate a one‑dollar Brownie, what would happen to the dignity and the day rates of the professional studio?”
A century later, autofocus triggered the same reflex: ‘Real photographers focus their own lenses; they don’t let the camera think for them,’ went the chorus, as if precision mechanics would erase the desire to see and make photographs.
Yeah.
We thought that.
Now think about what we say today.
Scary how much history rhymes.
An observation:
Every few years, someone announces that photography is dead.
And they say it with remarkable confidence.
But what’s remarkable isn’t that they’re flat-out wrong.
It’s that people stopped making photographs because they believed it.
The Repurpose Tool
Available now for $14 with my Substack Readers Coupon: pca3bfb
Most of us don’t have a content problem. We have a distribution problem. We spend hours writing one thoughtful article, publish it once, and immediately move on to the next idea. Meanwhile, that same piece could have reached entirely different audiences with only a little adaptation. The work is already done. The value is already there. It simply needs to be presented in a way that fits where people are reading.
Repurposing isn’t copying and pasting. Every platform has its own rhythm. A Substack article has room to breathe. A Medium story often benefits from a tighter opening and a headline written for readers who have never heard of you. LinkedIn rewards concise insights and clean formatting. Notes thrive on single observations that spark conversation. The core idea stays the same, but the presentation changes to match the audience.
That’s exactly why I built Repurpose. Instead of rewriting the same article four different ways every time I publish, I let the tool do the heavy lifting. Paste in your original piece, tell it which platform it came from, and it creates versions tailored for the others. It also pulls out standalone Notes and builds a LinkedIn post that actually feels native to the platform instead of a recycled blog post.
The goal isn’t to flood the internet with more content. It’s to get more mileage from the ideas you’ve already worked hard to create. One strong article shouldn’t live one life. It should become a newsletter, a Medium story, a LinkedIn post, several Notes, and whatever comes next. Good ideas deserve more than one audience.
For the people who still believe great work matters.
Creativity isn’t getting easier.
The tools we use are. They definitely are.
And that changes the equation.
So every week, I write about better thinking, stronger creative skills, AI without the hype, and building a creative life that can survive whatever comes next.
No fluff. No hustle porn. No BS.
Just practical ideas for independent creatives who refuse to become average.
My Personal Website
Me on Medium.
My premium members have full access to over 26 productivity and creativity tools, my books, and weekly updates to help them build a better creative business.
Premium Members keep scrolling for your special content.
Short series classes, a sprint this time, with a price that makes it a no-brainer. If you have never taken one of my classes, these courses are an opportunity to get a good taste of what we do.
Luxury Products: There is a theme to creating the incredibly perfect, engaging product work for the clients who demand perfection. 5 live classes, 8 portfolio pieces.
Mondays, 4 PM Pacific, Starts Monday, July 27, 2026 | Fee: | $125
(Limited to six students.)
Editorial Food: This is a wild and wacky class to explore totally off-the wall ways to shoot food. This isn’t your mom’s food class.
Mondays, 6 PM Pacific. Starts Monday, July 27, 2026 | Fee: $125
(Limited to six students.)
Claude CoWork Magic: Learn how to make personal apps, landing pages, and motion graphics. This class is a weekend sprint and is limited to six students. Students will need a Claude Pro license ($20 per month), an Anthropic API key ($20), and the desktop Claude model, Claude CoWork.
Class One: Claude Cowork Magic
Saturday, August 1, 2026
Claude Cowork and Productivity
6 AM Pacific (1.5 hours)
10 AM Pacific (1.5 hours)
1 PM Pacific (1.5 hours)
The amount of creativity you can get from mixing more than one AI tool is, without a doubt, a superpower. We will look at ChatGPT, Claude AI, Claude CoWork/Code, and NotbookLM as a suite of tools that will put a big ol’ S on your chest. We will focus on actionable items you can do immediately.
Class Two: AI Productivity
Saturday, August 22, 2026
Building with a Trusted AI Suite
6 AM Pacific (1.5 hours)
10 AM Pacific (1.5 hours)
1 PM Pacific (1.5 hours)
More information and registration here.
Here’s a fun infographic for you to print and hang on the wall. Use it every time someone tells you the business is dying.
And a lovely little podcast that is really fun to listen to.
