David Harmeyer

[David Harmeyer @SecondThread]

@QPHutu DPPO and GRPO both in blue is confusing

[David Harmeyer @SecondThread]

@junkliao @FakePsyho It's an AtCoder contest that starts at a reasonable time in Japan, but 3 AM Pacific time

[David Harmeyer @SecondThread]

@cdbrw15 @GuillaumeLample They cut the learning rate probably. You use a larger learning rate at first to learn faster but its less stable, and then you decrease it to get to the very top of the hill

[David Harmeyer @SecondThread]

Just spent so much time working with loss curves that when I opened up my 401k today, I panicked because I saw it going up and to the right

[David Harmeyer @SecondThread]

Yes, looks like it's a combination of that, and not enough variation in samples that killed it.

tanh() is a million times better

[David Harmeyer @SecondThread]

World's ugliest loss curve.

Is this just what you get for using sigmoid in hidden layers?

[David Harmeyer @SecondThread]

The AI times have begun

[David Harmeyer @SecondThread]

@steipete @openclaw Legendary

[David Harmeyer @SecondThread]

@xubinnrencs RAS syndrome

[David Harmeyer @SecondThread]

@tsoding At that point the only reason that'd be useful is as precompute-expensive, query-expensive dataset compression.

But in the real world, data is cheap and compute is expensive, so nobody would do this unless there's some immensely tight bottleneck on transmittable data.

[David Harmeyer @SecondThread]

Sleep is just "compacting conversation..." for humans

[David Harmeyer @SecondThread]

@sciencegirl I guarantee if you give this to actual people, they will misjudge the number of steps on the way down and trip

[David Harmeyer @SecondThread]

@zebriez Competence

[David Harmeyer @SecondThread]

@hyhieu226 If your reasoning isn't autoregressive, it's probably rationalization, not reasoning.

Maybe rationalization makes sense in some cases, but it's a lot more artistic than it is logical.

[David Harmeyer @SecondThread]

https://t.co/wfWuC54hAW is amazing, but the fact that it's not available in the extensions search is confusing.

It strikes me as obviously the best way to use Kotlin, but it's so hard to find

[David Harmeyer @SecondThread]

Many people fail because they go off to build a generic platform, and then find usecases afterwards.

It's so much more compelling to find a problem, build a solution, and then after that's proven, make the solution generic.

[David Harmeyer @SecondThread]

Register here: https://t.co/E77ljn7sib

[David Harmeyer @SecondThread]

Looking forward to Hacker Cup Round 1, starting tomorrow morning!

[David Harmeyer @SecondThread]

Every time I remember that "m" comes before "n" in the alphabet I do a double-take

[Charlie Kirk @charliekirk11]

Good men must die, but death can't kill their names.