Right now I naturally want to eat ~2600 calories/day. This will be a ~2-4 month experiment with an eye towards learning more about what it would take to get back to 15% bodyfat - if and when I achieve that, maintenance would be much easier to do naturally I think.

4. I bought retatrutide from Peptide Partners, which a friend recommended. I'll microdose it (dunno how much yet) and tweak my dose every week until I find the minimal dose where I feel satisfied from just eating my meal prep food and don't want to snack beyond that.

3. I realized that even though I've been doing my own meal prep every week for the last 5 months, I consistently procrastinate by 1-3 days before actually setting aside 2 hours to cook and clean, so this week I hired a helper from Care .com to come on Mondays for $60-90/week.

The diet: docs.google.com/document/d/...

2. I workshopped a meal prep plan with Claude that's 2150 calories/day (roughly a 300cal/day deficit, so 1lb loss every 10 days), and 220g protein/day (which is close enough to 1g per lb bodyweight/day to gain muscle).

I track workouts with the Hevy app for my Apple Watch. Over the last 6 weeks I've only worked out 53% as much as I wanted to 😬 but I am still getting stronger.

1. This week I reduced all my workout volume by 38% (I went from 16 total working sets to 10) because 16 sets took me an average of 40m and so I kept skipping them. Now they should take 25m, which hopefully means I will consistently do them 4x/week, even when I'm feeling lazy.

(It's not public, sorry, it's just a dumb little script built into a pile of other utilities I keep)

Results are pasted underneath the original line. Here's how they look:

Whoa, cool! Post more about it when you can.

My grandpa asked "Are there bad monks?" and I said yes, of course, awakening doesn't automatically heal your dysfunctional patterns. His reply was "Would you say one of these bad monks was Hitler?" which just goes to show that you can't anticipate every objection.

idle thought... perhaps in the same way that we ought to use AIs to systematically upgrade the security of codebases,

we could use AI to comprehensively upgrade the alignment of the proxy metrics we're optimizing?

aligned AIs as the meta-optimizers?

AI can turbocharge any optimizer. perhaps it'll reveal a bunch of these hidden alignment problems

but it also makes it easier to define better proxies

"optimization disease": when you maximize a proxy, it eventually becomes problematic

it's worse the more powerful your optimizer is

prob every metric is its own little instance of the alignment problem, it's just that most aren't being optimized hard enough to be problematic

rumor has it if you refresh the waiting screen 5 times they let you in. (for real, try it)