As will become clear part of this post is mostly authored via AI. When I set out to write I resolved to not use AI - beyond maybe to bounce ideas off of, and graphic design (which I’m very bad at). But this post is different.
.
.
I’ve always wanted to write a script for the Star Wars universe postmortem idea I shared earlier. But did not feel confident in producing a script. I wrote up the Star Wars post, which I felt reasonably good about. I played around a lot with AI generating “art” for the postmortem article, and fun graphics for a few other things I’m writing in parallel. But then I went back to bemoaning I still didn’t have a script.
On a lark I decided to take the constant advice from my friend Chenyang I just decided to ask ChatGPT write the script. My logic was that I wasn’t going to get to it otherwise, and it would be interesting to see what came out if I seeded the prompt with half of the article I wrote about an after action deep dive on the Death Star being destroyed. I have mixed feelings about the results. On one hand it’s really pretty good. On the other hand it’s arguably somewhat disturbing for humanity’s future. But, ChatGPT made the Designer the hero of the story - so it clearly the machines still have a ways to go in terms of credibility. ;-)
Not content to stopping where AI nailed it I realized maybe AI could produce the short film directly from the script. So in a hurry I checked out a number of tools that are out there, bought some credits on one and did a script→video GenAI experiment. It wasn’t a super exhaustive experiment - and the results are at the top of this post. The tools I used had some pretty basic flaws - most notably they didn’t maintain visual context from one cut/scene to the next and therefore the character change around a lot. Yet there’s a gonzo style to the result that had me cracking up. I was impressed that things turned out coherent at all given a script and a minimal prompt. The traditional Star Wars scroll was created with another tool and just connected together in a video editor. Also, yes - you caught me - I know the difference between Episode IV and Episode VI. But I didn’t catch the mistake until just now and I’m too cheap to kick in another $5 to fix it. Actually, I’m a little too lazy to retype and re-edit. I just wanted to be clear that’s a human and not an AI mistake. :-)
It’s possible I missed some critical setting in the generation that would have resulted in visual continuity. I tried a few post processing prompts but at times that made things worse. So I just stopped for now. Maybe using a prompt to do a single tracking shot with no cuts might have worked better. But a subsequent video had people standing in the middle of the table - so I sort of doubt it. Also - even a casual Star Wars fan will notice that everyone looks like a human - not a Wookie in sight. Some of the more sophisticated tools felt like they might be able to handle that - but I gave myself a pretty limited time budget for now to do my “vibe animation.”
Below I present Death Star Post-Mortem: Lessons Not Learned. Inspired by Star Wars and RandomWalkThroughTech.com. I hope you enjoy.
Coming Next: I’ve also created a script and video for Death Star Post-Mortem: Lessons Finally Learned by asking the model to rewrite the story in a world where leadership was still evil, but had adopted ideas from my engineering culture article. Subscribe to be notified when that comes out. That video is truly hopelessly bonkers.
.
.
.
Title: Death Star Post-Mortem: Lessons Not Learned
Setting: A sterile Imperial conference room aboard an Empire Star Destroyer. Stark lighting, minimalist decor, a flickering holographic projector. Imperial propaganda posters dot the walls, slogans like: “Fear ensures quality” and “The Force is with our deadlines!”
Characters:
Jira, Lead Engineer (skeptical, pragmatic)
Caden, Product Manager (overly optimistic, status-driven)
Vera, UX Designer (empathetic, ignored, sensible)
Commander Tarkin II, Project Sponsor (authoritative, fear-based leader)
Milo, Junior Engineer (nervous, inquisitive, newly hired)
[SCENE OPENS]
Commander Tarkin II: (slamming fist on table) How exactly did a single X-Wing destroy our trillion-credit battle station?! I need explanations!
Caden: (standing, forced optimism) Commander, let's focus positively. Our KPIs show remarkable progress until… well, the catastrophic failure.
Jira: (deadpan) The "catastrophic failure" was an exhaust port that anyone with basic knowledge of thermal dynamics would know needed shielding.
Commander Tarkin II: So whose failure was this? I demand names.
Vera: (hesitantly raising hand) Actually, our initial design research indicated that a single point of failure like this was problematic, but our feedback wasn’t addressed in planning.
Commander Tarkin II: Silence! No one wants to hear about your "research"!
Milo: (raising timidly) Um, I’m new here, but shouldn’t there have been a backup defense in depth?
Commander Tarkin II: (cold glare) Did you say "shouldn't," junior?
Milo: (gulp) Uh, I meant… could we possibly… learn from this?
Jira: (supportively nodding) Milo is correct. We didn't consider contingencies adequately because any discussion of failure was shut down.
Commander Tarkin II: The Empire doesn't discuss failure. We destroy failure!
Caden: Commander, perhaps a new slogan would help motivate the team. Something like: "Failure is not an option...unless it already happened."
Vera: Or we could stop making the same mistakes and listen to feedback from diverse perspectives?
Commander Tarkin II: (annoyed) Your feedback is noted and ignored, Designer.
Jira: Look, if we don't address these issues—lack of open communication, blaming instead of learning—we'll probably repeat these failures.
Caden: That’s a great point, Jira. Let’s assign blame efficiently and move forward quickly.
Milo: (innocently) Doesn't blaming prevent honest dialogue?
Commander Tarkin II: (exasperated) Enough! Clearly, this was sabotage by rebel terrorists exploiting… thermal physics. Jira, your team should have anticipated terrorist physics!
Jira: Sir, physics isn’t a terrorist conspiracy.
Commander Tarkin II: Silence! Action items: Build another Death Star, immediately, using exactly the same blueprints but bigger.
Vera: (sarcastically) Maybe we should put the single point of failure somewhere harder to exploit?
Commander Tarkin II: Fine, put the vulnerability behind some teddy-bear-infested moon or something. That should deter anyone.
Milo: (hesitant) That doesn’t sound secure…
Caden: (cheerful) It sounds innovative!
Commander Tarkin II: Exactly. Innovatively secure.
Jira: (facepalms) Right.
Commander Tarkin II: Also, add another layer of bureaucracy. More management oversight! Create a committee to oversee the committees!
Vera: (sarcastically) Because more management always solves engineering problems.
Commander Tarkin II: Exactly! I want daily progress reports. And double the stormtrooper presence to ensure absolute loyalty.
Milo: Does loyalty help prevent engineering mistakes?
Commander Tarkin II: (ignoring Milo) Furthermore, we'll hire external consultants from the Galactic Consulting Group. They'll tell us exactly what we want to hear!
Jira: So more echo chambers and fewer solutions?
Caden: But their slide decks are incredibly persuasive!
Commander Tarkin II: Indeed! Now, finally, add more intimidating weaponry. Let's put cannons everywhere. You can never have too many cannons.
Vera: Except when the actual vulnerability has nothing to do with firepower...
Commander Tarkin II: (interrupting) Enough from you, Designer! Meeting adjourned! Fear and excessive management oversight will ensure victory!
[Commander exits dramatically again]
Milo: I think I preferred the first meeting.
Jira: Yes, less paperwork at least.
Vera: (sighs) Looks like we’ll be back here after the next one blows up.
Caden: (enthusiastically) Can’t wait for the next retrospective! Great team spirit everyone! The catering was good today!
[Scene fades as team sighs even deeper, Caden happily munches on a cookie.]
[END SCENE]
Share this post