Doctor Paradox Mini Histories
By Barb Dybwad
Mini History videos published
~100
Publishing cadence
1/day
Per video
40s
Can you tell us a bit about yourself?
I can thank my Dad for an early love of technology. We had Pong and an Atari 2600 in the house; he got me my first Commodore 64, my first 286 PC back in the mists of time. It had a 300 baud modem and I met local (much older) kids from the BBSes, MOOs, and MUDs — something I've never seen successfully replicated in the modern social media landscape. Computers were fascinating portals to the outside world and a constant companion while I was growing up, both at home and increasingly at school.
Later I happened to take an elective math course in Artificial Intelligence in college — way back in 1994, where I learned about neural networks and the promising (if then obscure) field at the intersection of probability theory, psychology, and information science. So I was already primed to become an early adopter of AI. My career started out in IT, then web development, and then as one of the first professional bloggers — writing about tech for Engadget, Mashable, Entrepreneur, and other outlets for many years.
After moving over to the business side of digital media, I still kept tabs on the Silicon Valley ecosystem and technology trends, including developments in machine learning, well before generative AI burst onto the scene. I knew right away in November 2023 that this thing was going to be bigger than potentially the advent of the internet itself, so I dove right in. Plus I've also always loved playing with digital tools of all kinds as a writer, content creator, tinkerer, and incorrigible generalist with way too many hobbies. It's pretty awesome to have this Cambrian explosion of AI tools going on, and being able to experiment wildly in both my personal and professional life, as I'm leading the charge around AI at work as well.
What is the project you are working on with Glif?
I've been making 40 second Mini History videos for YouTube and other platforms, with a goal of posting one per day. Over the past few months I've amassed a collection of nearly 100 shorts, which are also great for weaving into my blog that covers political history and the digital information age.
What are your incentives or intentions for making the project?
I have been trying to grapple with, personally and collectively, the absurdist (un)reality of this geopolitical moment, and the situation we find ourselves in here in the United States in particular. I have been glued to politics since (at least) 2015, and have been applying the skills of digital journalism to the problem of understanding how we got here, and maybe how to get out again.
One comforting thing I have found is studying history — which makes it clear that we have faced challenging times before, and that there are many clues that are available to us from the past about how to endure, transcend, and be a positive force for change. In starting to look for answers I realized there was a ton of history I had forgotten, only partially understood, or never learned at all. And the more I learned, the more the puzzle pieces seemed to fit together better. It seemed like there were a lot of truths hiding in plain sight.
Early on I became a fan of and have been heavily influenced by historian Heather Cox Richardson, who has been writing a popular daily newsletter called "Letters from an American" since 2019 (Jon Stewart recently called her a 'salve'). She really has a knack for connecting the dots from the past to the present, and helps bring a crisp clarity to the issues of the day in a way that is both empowering and stress-relieving. I would cite her work as direct inspiration for this project — and I would love to be able to take what has felt like a kind of lifeline thrown, and pay it forward to others.
Your project has a very distinct style — a rich look and feel — what is it about the style that you think works in this context?
I think the tilt-shift miniature format is so perfect for these historical narratives — the visual tone is cute and charming, and inherently nostalgic by default, and I think it creates a little bit of emotional space to take in sometimes difficult truths about our past (and present). It also helps deliver on the editorial goal of making often complex concepts and large systems over time seem more easy to grasp, by literally casting the viewer in the role of an omniscient giant looking down on a coherent context. Museums often use the miniature diorama format for similar reasons: to grab attention, communicate complicated material in a way that feels accessible, and help the experience become more memorable through an emotional engagement that adds more depth than simply delivering dates and facts.
The whimsy and playfulness of the visual style can sometimes draw an almost absurdist type of contrast with the seriousness of the topic, which I think adds novelty to the way history is often covered — and certainly how it was presented to me in school, as a sort of dusty old tome one must slog through in order to graduate. History is a lot more thrilling and fascinating and full of ambiguity than I ever knew as a kid, and the colorfulness and curious wonder of the tilt-shift miniature helps to imbue it with the kind of energy and vibrancy it deserves.
The format is also more forgiving by nature, giving it an advantage in this rapid publishing model I'm trying to adhere to. There's no expectation of photorealism, so generative video models can handle it more reliably — but you can still punch up the likeness effects with real reference photos by just dragging them in (or asking the agent to search for some). Tilt shift miniature embraces being a little bit imperfect and rough around the edges, bringing in a craft-like aesthetic that gives it a more warm, organic, and human feel vs. the harder digital edges we tend to associate with AI — another advantage for a series about the human story.
You often use real photographic imagery to then transform into your style — what's the significance of that for you? How is the fidelity of the image that the agent produces important for your project?
As much as I enjoy the wabi sabi of the tilt-shift mini style, it's still important to me to adhere as much to historical and factual accuracy as possible — and I think it helps when viewers can recognize familiar faces both in the videos, and again afterwards if they see other references to that person or event. Repetition helps with memory and learning, and getting the likeness right aids in my goal of helping people feel more grounded in the facts — and feel more grounded in general.
Sometimes there is really a singular iconic photograph that entirely encapsulates that particular historical moment, and in those cases I will ask Glif explicitly to use it as a reference — other times I will let the agent find its own reference imagery as part of its usual routine that it already knows to do from our past sessions. In the new Glif platform the visuals have been noticeably dialed up as well — the images and videos seem to have an almost hyper-real quality to them while somehow retaining the toy aesthetic of the signature tilt-shift miniature style, which makes for a really unique combination.
Can you tell us a bit about your process for creating your videos?
A video can start from a lot of places, whether it's my backlog of dozens of ideas on deck, something I read about recently, something in the news, some AI research into historical events tied to the current day of the year, or I might just ask the Glif agent to pitch me some ideas. It knows the whole format and what I'm trying to do, and is quite good at coming up with concepts with the right angle.
With concept in hand we'll move into generating a 4-scene script with image and video prompts. The visuals are almost always spot on if the script is honed well, so I usually focus mostly on getting the script right and the rest will generally work itself out with some minor tweaks. If there are any obvious edits I want to make, I'll keep iterating and co-writing it with the Glif agent until it feels like it's in good shape.
At that point I'll deploy my favorite new device I'm calling the Editorial Council, where the Glif agent will create a set of multiple personas (usually ~4) to independently analyze, score, and give feedback on the script and visual prompts (for example The Fact-Checker, the Contrarian, the Visual Director, the Viral Strategist, etc.). It's a great way to get a set of fresh perspectives from a variety of "roles" without even having to leave the workspace, and the agent just spins it up instantly and helps push the quality level up by another significant margin. I really love how effective it is at making the entire piece stronger, and sometimes if there are significant reworkings of the script I'll run it again afterwards to see if we successfully met the mark that the sounding board recommended.
Next we'll do the images and some voiceover auditions in parallel — Seedream really nails the tilt-shift miniature style which the agent remembers along with all my other model preferences. It also knows that I want it to generate a voice picker with several narrator options reading the script from scene 1, so I can choose based on how my go-to favorites sound against the video's theme and try to match the tone and gravitas.
Then we'll generate all 4 audio clips for me to listen to, which is often the place I'll spend the most time tweaking outside of the script itself. This is where I can catch and polish up anything that ends up sounding unclear, incomplete, or not quite punchy enough when listening to the whole thing end to end.
Once those are locked, the Glif agent will generate the video clips, create a background music track, weave in the voiceovers, stitch the whole thing together, create text overlay captions, and present the final 40s video for review. Sometimes we'll do a rev or two on the background track or one of the video clips, but often it will be good to go and once I approve it the agent knows to draft me a YouTube description and pitch me on 5 potential titles.
Sometimes the whole process can fly by and be done in a 15 or 20 minute sitting, and other times I'm chipping away at it throughout the day and just picking it up and nudging it along in spare minutes of time here and there. It's super handy to be able to seamlessly drop in on a thread from another computer in another room of the house, or on my phone when running errands, and feel like I can make real progress towards creating a final product.
How does the Glif agent play a role in shaping your project?
This project would not exist without the Glif agent. It has been the keystone that unlocked the ability to consistently make a video a day amidst a busy life with a demanding job and a 4th grader at home — and actually stick with it. It was also the original inspiration for the tilt shift miniature style — it was the first template I tried and I just fell in love with it right away, and the project seemed to reveal itself almost immediately.
I consider myself fundamentally a writer at heart, and someone who struggles with visualization — so AI and Glif in particular feel like a godsend in terms of being able to self-produce visually interesting work that is actually publish-worthy. It's brought new creative inspiration to the intrinsic joy of simply making things and seeing them come to life, and the agent workflow offers a highly satisfying feedback loop that learns from our adventures together and brings new ideas to fruition in a way that feels like having superpowers.
I think it's really important to keep finding new ways to tell the stories of our past, and to keep renewing our curiosity about where we came from and how we got here. The Glif agent has been a truly exceptional partner in exploring this idea — keeping the process fresh and fun while serving a serious goal, and helping viewers face the challenges of the present with a renewed understanding of the past.
