Pedma

pushed a lot of changes on my code today.
but now of 1000's of trades in an historical test, there's 2 trades not matching yesterday's.
meaning, something changed, but so small I am almost ignoring it. but I just can't. I won't leave desk until I know what that is.

old code refactoring is a nice thing opus 4.6 is helping me a lot with.
I had a behemoth of a code I built 1.5 years ago to manage my positions and it just one-shotted the whole thing in like 15 minutes.
still in beta test but all tests mirror exactly what my prior code did.

I am now more times facing the decision of shipping a new model, that I might not fully understand, but that seems reasonably good on an historical test, or to go back, spend 1-3 days going through the math and intuition to understand it.
I am the bottleneck, but I want it to be so. I think that's the feature of what we do and not the bug. I can't push something that will handle my own capital into prod, that I don't fully understand.
I am not talking here of a BS signal that people test and then call it a system... I am talking about full-fledged models, that seem to capture some behavioral

passed my new portolio's order execution history, cumulative P&L, benchmarks, etc, to claude and came up with this detailed 13 page report on what to do to improve it.
a lot of it I already was aware of, but idk man, pretty detailed report if you read the entire thing.
there's a few things I don't agree with, like his obsession with hedging out the beta, which is something I don't care that much for right now, as I am fully aware I am just a beta farmer in higher risk markets.
overall pretty neat and added a few reasonable solutions.

people that are worried about AI replacing their job because now 1 person can do the job of many (not all cases ofc).
but how many more jobs will spun up because of exactly that? the cost of bootstrapping a new business is rapidly going down and there's so many issues in different fields to tackle, and new ones will also arise.
question is what's delta of job loss vs new jobs generated by what AI is offering the market place in terms of efficiency. I think people are not pricing in how much you can ideate now that would take a ton of time and investment in the past.

no matter how good AI gets, you still have to put in the work and consistent effort. there's no way around it, that's the default basis to get anything done and most people still can't do that consistently.
it's just an amplifier of work ethic, not a substitute.

I think a lot of the AI hype requires a lot of creativity that only a human can provide. if speed and iteration is becoming less and less expensive to achieve, if you have domain expertise, high agency and creativity, I think you'll do really well in the years to come.

I literally just saved myself dozens if not hundreds of hours of research a month by coming up with a simple app that took me 2 hours to build. really impressive stuff.

sonnet was definitely nerfed

claude has a weird obsession with defaulting to something even when forbidden from doing it.

use tough markets to build, to get better. attention is not scattered so there\'s time to focus on a single target. I find it hard to concentrate when everything\'s flying because i\'m a bit of a degen and want to be playing everything. calm times are where I find time to build.

at least the decoupling theory was right.just not the decoupling decouplooors wanted.

sometimes you have to go back, spend a few days on something that is already built, but that can be improved, so that you can save time going forward.