John Grange

@jmgrange.bsky.social

Co-founder + CTO @ OpsCompass | Co-founder + GP @ MOVE Venture Capital | Cybersecurity, AI, software | Tinkerer | Dad

Videogames actually don't need a lot to be *cool*. Like, here's the intro from DOOM 64. Tech-wise it wasn't particularly impressive even when it was brand new. But it was cool. And even now, more than a quarter of a century later, it's still kind of cool.

On the one hand, military history has a more even political balance than most history (that is, there are more conservatives there), but on the other hand, every military historian I know, even the conservatives (*especially* the conservatives), pretty much despises Pete Hegseth.

Glonzo In Exile 🇨🇦🇬🇱🇨🇦🇬🇱@glonzo-in-exile.bsky.social · 2y ago

I was repulsed by military history for a longtime because I (wrongly) associated it with people like Hegseth. I now realize that the Hegseth types are basically anti-military historians.

Quantum mechanics is described by a consistent & VERY well tested mathematical framework, & nothing we've seen in particle physics experiments is inconsistent with its basic tenets. It's incredibly precise -- we can predict properties & behaviors of particles to truly amazing levels of precision. 🧵

This thing where folks are convinced it's cheaper to order doordash than make a burrito because they're comparing the cost of one dashed burrito to the cost of ingredients for like 8 burritos feels like the food version of folks who grew up with ipads and smartphones not knowing how to use computers

jamelle@jamellebouie.net · 2y ago

i feel like no one taught a lot of these kids about per unit cost? like, even if i buy the most expensive ingredients, the per unit cost of making chocolate chip cookies at home is still cheaper than buying them from a bakery.

I spent the day with o1-preview-2024-09-12. Ignoring costs, because that wasn't really my concern today, my impression falls pretty much in line with its results on LiveBench Language benchmark, a benchmark I feel is fairly reflective of model capability.

Bild

Chain of Thought helped LLMs reason forward through problems. This paper found teaching them to also work backwards (like humans checking their work) makes them even better “reasoners” Applying basic human thinking strategies to LLMs continues to yield interesting new ideas arxiv.org/pdf/2411.19865

Bild

This is exactly right. The 90s/2000s idea of 'mainstream' media simply doesn't exist anymore. Republicans own the mainstream media, as it exists today. The Democratic party definitely needs to recalibrate.

Post nicht verfügbar.

Wow, this is such a useful resource of industry LLM applications! And filtering via search/tags is so responsive. I was thinking of compiling something like this over the holidays (ala applied-ml) but thanks to @strickvl.bsky.social I can spend the time reading instead ♥️ zenml.io/llmops-datab...

UI for the LLMOps Database with the search query of "eval"
Alex Strick van Linschoten@strickvl.bsky.social · 2y ago

🤔 Do you ever wonder how companies are putting LLMs and GenAI apps into production? What stacks do they use? What architecture did they go with? I put together a database of known public technical writeups with summaries of the key technical features.

GIF of me scrolling through the LLMOps database website

Somehow what I'm working on tonight ended up requiring me to build a Windows 11 VM in Azure, with a very particular WSL distro setup. Basically, witchcraft. Pray for mercy upon my soul.