Clifford Richardson

Reading a foundational biochem simulation corpus and it included SchNet which made me realize it's basically just GNNs... that's the state of the art up to the period of papers I am reviewing. It makes sense after a brief thought, but I am guessing I will learn how they solve some more extrinsic effects than cannot be learned through message passing.

Agree. I want to engage more with folks, AI comments make it impossible to hold meaningful dialog with larger players that become targets for monetization abuse.

Explaining statistics to people is hard. You have a detector of some event. If probability of a false positive is 1% but incidence of the real event is also 1% your actual chance of a detected event being a false positive is 50%. Sounds like nonsense. Isn't.

I don't see it so black and white. Investments are both a way to grow shorter term capital but also sometimes come with an air of credibility and long term network connections that can really expand what you can do. It's not just about the money which is why who leads and who they can pull in is important. It's true that rounds are a lot of work and ofc it's ideal if you can prop up on revenue, but money isn't the only factor if you really want to grow your reach and marketing is great but having an in to other investors that have stakes in companies is better (i call it the sequoia effect). Dilution risks can be worth that. Remember as a founder your ultimate goal is to make your vision of how your company betters the world come true. Not the money. Naturally i again repost because PGs been captured by bots overwhelming his comment section. Still a big problem still annoying. Shame X can't manage to solve it.

This is not only the engineer's misstep. Sure, letting a laptop get compromised if from ignored security practices tbf is a form of negligence if that's how it went. But still, the larger sin is the default access seemingly pointed at here. No one should have default production execution access or any production access without someone else's signoff. Tragic levels of poor access design. There are cases in banks you can trigger an incident to grant yourself such privileges but I've never seen it enabled on systems that are involved in immediate money movement because that is insane levels of financial risk.

"I'm a 100 MARR Founder... " - and then you look and it's a one-week trajectory of user growth extrapolated linearly. 🙃 What has happened and what drugs are people taking to make such bold claims for clout?

Last post I made was a bit too harsh. Here's a watered down version RE: IATA's rollout of BIX. Moving data serialization to JSON/XML does not make your data format modern. This is a great step for the infra around baggage handling, but directionally bad long-term for maintaining the legacy data format like this. Speaking from massive amounts of industry experience around data management here, this is very well-intentioned, but a mistake that hurts massively and for no real benefit of interop. It's decisions like this that really hold the industry back and keep IATA as a bottleneck and in the stone age. I would rather airlines/agents/distributors actually gain the long term monetary benefits that we all ofc want to pass down to travelers. However this is a very good way to kill all that. Core recommendations here: * You have a unique opportunity to create a layer of pristine entity separation where references that use ambiguous codes no longer needs to exist. Leverage this more than you do today * Move all of said references to a federated model of trust where you can still be the central trust authority but can much more cheaply offer ways for players to interact in a global network of data transfer for flights/bags/etc. Ensure verification of who is who and what is sent can be done offline (note: this is a solved problem). * Keep the old legacy legacy. Ensure there is a strong separation when translating data there, and if you need a legacy style entity reference, track generation and use of that separate to the new system, don't bleed fields like that into said newer system. What is created here has a 100% chance that still every player is creating their own bespoke separate formats and then yet another translation layer to your systems while they secretly plan to go the RyanAir approach and just directly communicate to each other outside of your network.

Hmm i wake up and learn there's a trump phone. It is embarrassing that anyone touts this when they can't even get the flag right. fascinating really because the American flag is really easy to validate to anyone that gives a hoot about America. 13 colonies -> 50 states.

Broke: Don't hate the player, hate the game Woke: Hate both the player AND the game

I'm lost, do people think tracking systems today convert a tracked object only into coordinates sent to the cloud for GPS tracking? This was the standard long before GPS. Seek, acquire target then track. Payload detonation optional. The actually cool thing here is the obstacle avoidance, which had nothing to do with GPS even before this, but is a cool thing to solve

Incredible momentum in betterment we are seeing. github.com/alrojo/fp4Attn

Building a Diffusion Transformer model has proven to be more about scheduling what conditionals you introduce to steer more than just learning structure from the entire set. This has become annoying, but client pays so work continues.