mikeyla85

A while ago I built a harness to control After Effects and Cinema 4D with Claude, and Fable has been able to improve it until it works really well.

We used it to solve a very peculiar challenge: a video game level that needs to convey narrative beats by placing videos at certain points, that also had to look like a video game with realistic physics.

The challenge is that making a video game is a radically different process from creating motion graphics. In Cinema 4D, you create and tweak keyframes, and the first step to making something look good is locking the motion. That’s what lets you choose where things will appear onscreen and customize the frame. Whereas in a video game, you throw rules on top of a physics engine and then the player plays.

So Claude pulled real telemetry from games, built a physics model, and created a framework that tracked all of my notes and solved the camera, adjusting weights for different constraints to create different versions of the videos. It's got a player model that responds to threats with movements, and I can put my thumb on the scale with different beats that all get integrated into the solver.

For each round of notes, it puts one folder for each lane we're tracking, with deliverables and A/B's and a markdown file. I give notes, it tracks them, and we rinse and repeat.

Two weeks in and we are almost at a finished level, which we'll import back into Cinema 4D for a last style pass.

I've been pretty mind-blown by Fable's ability to track so many moving pieces and build a cohesive system around them.

This is the kind of work that LLMs have been truly bad at: controlling software that doesn't want to be controlled, tweaking hundreds of values that relate to vision and spatial reasoning. So we did a lot of rounds of adversarial testing. A recent report was "this looks like a handsome screensaver with a health bar" which came with 40 recommendations.

I asked Claude to make a 12-minute video going into detail about how it executed this and go into all the nerdy details. Video is in the Substack post above...Claude really likes byte determinism.

If you want to hear me talking three years ago about most of these same things, before AI could write well: https://open.spotify.com/episode/2KFFxugfB2wXQB3VKYhBLp

I wrote the post and have been working in and researching in the space for a decade and truly did spend a decade making a movie about this because I think it's an existential threat to humanity, but sure.

BLIT was awesome. Reread it recently after watching the Black Mirror Playthings episode on it most recently.

Well, I thought it was cute.

I don't really use a ChatGPT subscription so I have Claude Code create a few versions of an image in different styles. I mentioned it especially because evolving from prompting an image to automated prompting loops matches some of the recent changes that we've been talking about.

I wrote the article, I really have been working and researching in the space for a decade and can share similar articles published online long before LLMs.

Be careful with messing under the hood on this one. I’ve used this feature in the finder for years, but recently used fclones to replace all my duplicate files with clones on my Mac, and completely messed up my free space. Months later, some programs see 1TB free and others see almost nothing.

Nooooo the 90s are truly back

What is trust? What outward manifestations signify or prove it? What is the difference between trust in ourselves that lets us take risks, and trust in others that can send us into the same unfamiliar territory?

It was a fascinating way to query and explore perspectives.

Checkboxes are just as usable as toggles on touchscreen. It doesn't have to do with the tech, it's just that toggles are newer.

My 13 Pro is so good, and it sounds like there are still issues with 48 megapixel photos (a lag in processing the files, compatibility.) I'll wait till the next model.

How can I buy please

Facebook could stop trying to grow as quickly as possible. They are clearly trying, and they are clearly failing. That doesn't exonerate them, it highlights a problem that must be solved for global safety.

Hard disagree. Facebook does not need to optimize profit and growth in every product decision. They make choices about what to amplify, knowing the consequences of those choices.

I remember the 90s

In this context, security by obscurity is literally what protects private keys.

You...don't believe in new coins, innovation on features, or marketplace competition?

Lightning doesn't have privacy and Signal needs privacy.

There's an entire discipline of UX best practices. Some of it is subjective but some of it more generally agreed upon. I don't think you'd find many UX designers who say AWS follows these, and it's up for debate the degree to which it's intentional.

The notice is poorly worded and tells you that the dialog box will not do what its name suggests.

Are you suggesting we pay for a $30/mo support plan for AWS when we planned to spend $40/mo total on the service?

This.

If it doesn't operate like Dropbox under the hood, don't make it look like Dropbox in your interface.

If an action creates a copy of objects, call it "Copy Objects" or whatever.

I think that presenting an S3 bucket as a set of files that you can select and perform operations on sets up false expectations here. And not showing incomplete multipart uploads in the file listing or having any way to see they are there isn't great.

We're a film company with contractors, and I was raising money as we went. I also said that under slightly different circumstances (some last-minute investment) we could have been bankrupted by this.

Sorry you didn't like the tone of the article.

I'm far from the only person who - has needs that can be met by cloud providers at costs I can afford - can't afford to pay a sysadmin - has enough technical skills to hack my way through it

But this is also like saying "You should have hired an expert to tell you that when the sailboat builder kit manual says drill into the back it actually means the front." Regardless of whether someone more experienced at this would know better, this UX is awful, its title is inaccurate, and the whole process around changing storage classes far more confusing than the UX makes it seem.

There are petabytes and petabytes of data from independent filmmakers that are rotting away on drives, and part of the issue is lack of other affordable options. So I get that "shoulda hired a sysadmin" is an easy response but that's not always an option, especially when there's dozens of hard drives or other complexities that make these tasks last weeks or months.

Thanks for the information!

Re your first note -- because I assumed that this operation was being treated similarly to a move (which is a copy and delete in S3), I assumed that I would be dinged for the cost of copying (minimal when on the same location) but the original would be deleted.

Reference: https://forums.aws.amazon.com/thread.jspa?messageID=455101#j...

Second -- this makes sense to me that this is how it works. But I still don't understand why anyone would have a use case for this "Edit Storage Class" dialog box. Is there ever a moment in the administering of an object where you would want to duplicate that object but with a different storage class?

Unfortunately hiring a sysadmin is way out of our budget. We had a friend helping with some AWS stuff and they may have done some of this for us if I'd asked, but I do remember it being a fairly mind-numbing task and he wasn't co-located with the files.

Thanks for the information! This is helpful.

Given that I was pretty knowledgeable about most aspects of backing my data up to S3, it would have been pretty hard to predict where the possibility for error would have been. I can't imagine hiring someone for 2hr to come and check my work, since they really would have had to be there for the entire process.

It sounds like about 2-5 days of work to get someone to set all this up, but perhaps it's more like 1 day at a minimum. So somewhere between $800-4k for that. Later in our project we were running 5 S3 buckets with about 20 stakeholders uploading and downloading, so that could have easily been another 4 days of work.

$800 would have been two years of what we were planning to pay Amazon. $4-8k would have been a major big-ticket item for us (even though, of course, that's the bill we ended up stuck with.)

If I were to do it over again I don't think I'd be able to justify spending on a pro. I would definitely spend on someone to audit our expenses more carefully though.

Out of curiosity, what would market rate for a sysadmin run to order a snowball and sync 24TB of data to it, ensuring that it ends up properly in Deep Glacier?

I "ended up in the AWS console" because S3 was a cheap way to store data so I decided to use their product. I've been a casual web developer, hacked together complex Javascript webapps, built Chrome extensions know my way around the terminal. Enough knowledge to be dangerous but not a professional.

On AWS, I know how to manage IAM roles, set up EC2 instances, etc. etc.

The idea that only a dedicated IT professional should be using this product really cuts out a lot of people who can't afford a dedicated IT professional.