benmcmahan.bsky.social
Creative software engineer. At Glean, previously Puppetry, Google (Doodles), Amazon, mcmahan.games. He/him
26 posts
177 followers
428 following
Getting Started
Conversation Starter
comment in response to
post
"Fascism attacks the vulnerable.
We all have vulnerabilities."
Beautifully put, that will stay with me.
comment in response to
post
Just a guess: they have as much infighting but their message of fear and hate is better at unifying against things so they vote as a bloc. When it comes to actually building something they do worse at unifying.
comment in response to
post
My favorite example of this is Guitar Hero. We would still be playing it if they launched one every year. Instead they made 25 in a 5 year period and burned the brand out.
comment in response to
post
What if you had Monday off?
comment in response to
post
I'm jealous, that's a great keepsake.
comment in response to
post
It's kind of like how a well trained dog gets more freedom.
comment in response to
post
I haven't looked but it might be constructive to look at raising money timelines to explain these things.
comment in response to
post
My hope is this relies on the assumption that money is the only motivator in the world and that that isn't actually true. Someone will release open weights for some philosophical belief.
comment in response to
post
One example was the threat of tariffs. I was all ready for a recession this week but it became nothing, but just the threats were enough to make every trading partner start thinking a lot more about contingency plans, which are now also more likely to be implemented.
comment in response to
post
I can't imagine this works. Basically a crawler add urls to a priority queue as it find them, those will be sorted by importance (number of times seen). They don't crawl a whole site and move on, they do url by url. So each of these links just gets buried. You never expect to finish a crawl.
comment in response to
post
There was a time when there was a push to outlaw congress buying stocks. This account was one of the ones pushing it. Now it's no longer an issue that's top of mind but people don't cull their follow lists much.
comment in response to
post
Check for every sub penny remainder to go into some account. Hopefully they got the decimal places right.
comment in response to
post
I don't think they've put out the technical report for r1 but there is a paper and it talks about how many steps, training examples they use: arxiv.org/pdf/2501.12948
comment in response to
post
Note this is for the Deepseek v3 which came out in December. Not the r1 model that just came out last week that's causing this media storm, but it was also fairly cheap.
comment in response to
post
The tech report says 2.788 M h800 hours:
arxiv.org/pdf/2412.19437
At $2 an hour that comes out to around $5.5 million.
comment in response to
post
My favorite technique for this is to pick how many in a row you are okay with (say 2). Put 2 copies of each number in a bag and pull out of the bag randomly. When the bag is empty you replace them all. Feels random but removes big streaks and reversion to the mean happens quickly.
comment in response to
post
Does this mean they'll cancel people's student loans now? Seems like emergency price relief to me.
comment in response to
post
I love that i immediately started reading that in Rorschach's voice.
comment in response to
post
Just a reminder to not trust anything you see anymore.