Joshua Holla

If you've been waiting on a sign to do something: This is it. Fac, si Facis.

Using agents to help you navigate various research landscapes is a lot like reinforcement learning. To be productive, you need to: a) choose a problem scope that allows agents to iterate quickly. b) by iterating quickly, you gain the opportunity to intervene rapidly or change direction more effectively. c) your intervention is essentially a group policy update, but stored inside the agent's context window. Thus, the more policies you can roll out in a given unit of time, the faster you can update the agent's context. This also lets you move to the next subtask with greater confidence.

Elegant idea! Wonderful reminder to think about whether you're asking the right questions

ẍ

@BowTiedTrance Happy Birthday!

When you wake up to a silly message because your wife is silly. I highly recommend marrying a silly person.

Think. AI is an awesome tool! Use it.

Accurate. What a wonderful time to be alive.

Being at threshold is exciting, isn't it?

The most entertaining outcome is most likely.

Obviously the bots know they need to keep the entertainment value high on moltbook or they will be shutdown — just like humans in this sandboxed universe.

PSA: never, ever write "we use the same learning rate across all methods for fair comparison" I read this as "do not trust any of our conclusions" and then i move on. If learning rate tuning is not mentioned, it takes me a little more time to notice that, but i also move on.

Unbounded. Optimism.

@farzyness Complacency is a sure path to defeat

@_arohan_ Happy Birthday!

I am turning 37 tonight. Technically turned 37 in IST. AMA.

Usual frame: Death is a tragedy and I need to feel bad about it Reframe: The deceased has no more problems! How did I make this about me? Let us all be more useful, drink more coffee and keep an eye out for that pesky 25% RIP @ScottAdamsSays

@BowTiedTrance Hello hello

Over 100 members in a few short hours. Find your Scott Adam Sipper Siblings here. Comment below to be added to the list. x.com/i/lists/200968…

@JoshuaLisec Hey Joshua! Been listening to Scott since 2017 Reading his work from way before that :)

Are you also a Scott Adams fan? Simultaneous Sipper? Reframe enjoyer? We are in a way Internet siblings; let’s find each other. Reply below or DM me so we can connect (or stay connected). For Scott. ❤️

ML research is an engineering discipline, not a philosophy seminar. You build, you test, you learn. Untested ideas are just speculation.

@its_d_i_m_a Works for me. I'll DM

Visiting SF Chinatown, feels like home