I’m going to need some warhammer experts to tell me if fable actually pulls this off

The most human thing about Fable 5 is how lazy it is. I asked for a million token burn and it keeps trying to talk it’s way out of it:

I’m not kidding

The part that really fuhks with your brain with regards to durability is that the orthogonal axes for prior exercise are duration and pacing (not work):

I didn’t know he could sing !? hash tag World Cup

what the fugg ??? fairly sure this is how my cousin ends up in prison

Remember when we fought for steel and you turned your back; now look at this place:

I believe I have finally beaten Claude into submission and simply do what I asked:

you want 2 wheel transport, this is the demographic you should be going after:

Unbound is fixin up a shit show

This is me in Arkansas about to bike race

Take a close look at the radar charts. They have both CP model parameters and FPC scores and plotted as percentiles. It was surprising to me that a high fpc1 doesn’t necessarily track with high CP. FPC1 one is your all around ability, which in the first example is driven more by Pmax and x_inter

What are the FPC loadings in the Critical Power parameters:

you my friend are insane:

every single time you get it to good state of context steering:

This was a surgical extraction of good data:

Finishing up the figures, cut on the left, GCclean on the right:

I’m about to go diaganol on a grid

got it knocked out, took a couple diagnostics

the good part is how fast i can now hunt down the problem.

ugh thats not what you want to see after several steps of 12 hour runs:

ECT doin his thing (read the details):

time to calibrate the ride cropping algorithm,

my gahd the stuff that this thing can just do, is just astounding: can you show me what figure 5 from puchowicz and skiba would look like with the puchowicz sampling scheme , it just literally redid a one shot remake of the entire study with a different sampling scheme.

Opus is in full meltdown mode cowering in a multi-dimensional token space devoid of any connected tokens: I AM THE GREATEST! This is the Puchowicz method for PD basis transformation between athlete and reference curves:

redefining your basis was one of the most liberating concepts. Want a nice sampling scheme, don't sample based on a time scheme or a power scheme, sample on cumulative arc length in semi-log space. also don't ride for 10^6 seconds continuously, apparently the model says you will die.

interestingly replicating with python (which does not have the fdapace package) did not work as well. They don't seem to hold as well at the short durations.