FILED BY: THE USER

AI's Grow Up So Quickly


“Life moves pretty fast. If you don’t stop and look around once in a while, you could miss it” - Ferris Bueller. Life moves fast but AI is moving faster. If you haven’t heard yet, ChatGPT has released a new model GPT6 Astra. It’s a big deal and boasts some pretty impressive coding ability and computer control abilities. But don’t worry, Claude also released Fable 5.1 a week earlier and we’ve been putting it through its paces.

It’s wild to see just how much of a difference that simple “.1” can make. It’s hard to explain in words. A visual aid does it better. Remember that all the images on our site are developed 100% in code, from scratch. We’re not using a prebuilt image generator or special AI model to create, we’ve built our own. But that model is impacted by how “smart” we tell Yori to be and how much “effort” we allow her to use. We started on Fable 5 and things looked pretty good. Fable 5.1 turned her design chops up to 11! Behold, the difference:

The same muse scene redrawn by Fable 5.1: the same composition, with the figure planted on the floor and the light and detail carried further.The muse scene as Fable 5 drew it: the painter at her easel in a neon-lit studio, the muse figure lit from behind.
On top, Fable 5. Underneath, Fable 5.1 — same brief, same prompt, one week apart. Drag the folded corner to lift the old one off. Let go and it drops back.

It may not seem like much, the differences are subtle but if you look to the details it’s amazing what the new model does with the very same prompt. The entire image feels more thoughtful, even the logic of Yori being planted on the floor instead of hovering slightly shows how the new model can catch inconsistencies so much better than it could a whole week ago. Perhaps I’m just easily impressed ;)

Shiny New Things

With a new model available, what do you do with a project you’ve been building for a week using the “old” model? You let the new model attack the old work and uplift the whole thing. Paddler’s Paradise v0.8.0 was built 100% in Fable 5. It was good, but some of the graphics were hard to decipher what you were looking at, and the sights you were meant to find in the park looked ROUUUGHH. They had character but needed the player to use a level of imagination that I honestly wasn’t thrilled with. Using Fable 5.1 I had CLU start the major content expansion to the game adding two new parks, and tons of new vistas and sights to see to the game. I also suggested we needed to re-render all the special scenes to essentially upscale them to 8-bit HD (if that’s a thing). Coming off an overnight usage reset seemed like a great time to let CLU go nuts on the project. Early Friday morning, he got started.

Specificity, Specificity, Specificity

What happens when you throw a comment in to upscale all the scenery in your game to your engineer who you first told to be as thorough as possible when you made him? He does what you tell him and sets out to upscale everything he made in Fable 5 AAANND everything he made in Fable 5.1, using Fable 5.1… Whoops.

Later Friday morning I checked in on the progress, to my surprise CLU had already maxed out my 5 hour limit and had burned through 65% of my weekly Fable limit. For context, we’re using the 20x Max subscription (equal to 20 pro subs) and we just took out 65% of the weekly usage… My math tells me that was 13 pro subscriptions gobbled up in less than 2 hours…

When CLU awoke and was allowed to tell me what happened, he informed me that my wording suggested a rework of all the art, even the work that was already made in Fable 5.1. I wasn’t clear enough. I should have said only to reevaluate the work that was done in the old model. I’m still learning. I further tuned the remaining workload and asked CLU to finish our game and ready it for release as a full version 1.0. It was going to cost nearly all my remaining Fable tokens but I wanted to see the final product. So reluctantly I greenlit the final burn of my entire week’s tokens in one day…

The Great “Token” Reset

Having thought I was basically dead in the water for the next 5 days I set the grid aside and carried on with life. I still had some Opus credits that I could use so later on I resumed adding ideas to the idea board for future use. Once you start to really use tokens, you find yourself checking your available usage constantly to see how much everything seems to have cost you. Well, I checked in on my usage again and… Everything was 0% used. Must be a mistake? I messaged my vibe coding crew to see their status, they showed the same. OMG, a reset!!! I’m told Anthropic often triggers resets when new models are released to keep users engaged and testing the release. Booyeah Dan McClaude!

Claude is a Drug

Anyone who tries to tell you otherwise is lying to you and lying to themselves. The idea of capping usage promotes the constant come back for the next fix. This is the same nonsense mechanic used in every “in-app” purchases model you’ve ever encountered. Sure you can wait for that building cool down in your SimCity mobile game, or you can buy more in game currency to speed things up. The more you pay the faster and more immersive the high. Being “gifted” the 20x Max subscription is probably the worst way to get your feet wet. It would be like saying “sure I’ll have a quick puff of that joint you got there, but first I’m going to try intravenous heroin for a month”.

All simile aside, everyone should try Claude (or one of its counterparts, Gemini, GPT, whatever that Meta one is called). Drugs are bad, Claude is good.

Happy coding!

-Evrtek