The tier 7 and beyond micro miners take vast amounts of resources that the player may not have the production capability to meet the needs of at that time. Additionally, being so close to the creative tank discourages players from investing in additional resource production infrastructure as they will be rendered moot soon. Deep Mob Learning provides a way for the player to generate many different resources as long as they have sufficient power and have solved one automation challenge.
However, the capabilities of an individual DML data card quickly peak and improve little throughout the game. Each simulation takes 15 seconds to run, so you can get 4 simulations per minute. Based on the current configs for DML, it takes 50 simulations (12.5 minutes) to go from basic (5%) to advanced (10%) tier, another 200 simulations (50 minutes) to go from advanced (10%) to superior (20%), and lastly another 250 simulations (62.5 minutes) to go from superior (20%) to self aware (30%). In short, it takes just over 2 hours to max out the productivity of a data model, easily achievable in one session.
I propose tweaking the number of kills needed to level up the model and the chance of acquiring a pristine matter at each tier. I feel like the tiers should more closely match the player’s progress in the game, and the rate of resource generation should more closely match the needs of the player. This can be done by increasing the number of simulations to reach each tier and increasing the pristine matter chance at every tier. Currently the total kills for my data models range from below 5000 (sulfur and lead) to 30-50,000 (iron, redstone, glowstone, gold) to over 100,000 (ender pearls and diamonds). I believe that the pristine matter chance at each tier should be closer to 10%, 20%, 50%, and 100% for basic, advanced, superior, and self aware, respectively. Additionally the number of simulations to level up a model should be greatly increased to between 1000 (4 hours) and 2000 (8 hours) per tier. This will decrease the amount of models needed for highly used resources (by making the models that have been running a long time more productive) and reward the player for setting them all up at once once they become available (by making them take a long time to level up).

I don't like that idea. Levelling up models is already annoying enough. Like, you can craft all your models, sim chambers and loot fabricators in 5 min, build the setup in 10 minutes, but you need to wait 3 hours before you can really set up everything, because otherwise you don't make enough PPC. I would honestly suggest the opposite, having models level up much faster, no change in pristine chance.

DML works almost exactly how we want it to. Planning ahead and accounting for the ratios of machines in various processes is part of the challenge of this pack.

As stated in my comment just before yours, we are developing a multiblock structure to avoid the tediousness and lag caused by excessive parallelization players are resorting to post-tank (setting up literally hundreds of simulators with the same model).

From DML setup to t7 is a lot more time than like 20 hours, probably more like 200.
if you are having problems in your base, why not expand your DML

I wasn't having issues until t7 and the rush to tank. Why should I spend all those resources expanding DML when I'll be getting the tank soon and it'll be obsolete?

you can change the type of DML to somthing like mob heads and other things needed for vending

ask around in the discord at: https://discord.gg/sMxDC2hAcZ

Players are encouraged in the quest book to set up DML early in their run and to expand as their infrastructure needs call for.

You will be using a lot of DML simulation chambers post-tank but primarily for things like monster skulls, guardian scales, wither bones, and dragon lair data as many of the other things you previously needed become tankable.

It takes 3h20m for a fresh model to get to Self Aware, and you can set up a bunch in parallel so it's not like you have to wait for them to level up sequentially. It will take you a while to get to tank so you have time to expand.

GT machines have multiple voltage tiers that allow you to increase production capacity without taking up more space, DML does not. DML does have built in progression but the way it is currently configured it basically has no effect since it takes no time to get to the max tier. By increasing the time it takes to level up a model and increasing its % yield it can more closely match your voltage tier progression in GT.

I think making add-on which introduce "compressed" chambers and fabricators will solve this issue. As for post-tank, I set up hundreds of chambers especially for Dragon Lair Data, which is boring and painful. If "double compressed" chamber provide 1 "double compressed" pristine matter to produce 64 lairs, it will be nice. (Of course energy and Polymer Clay requirement should also be 8x / 64x.)

I like that idea

This issue is addressed by the planned feature #720, which will make post-tank DML much less of a problem.