slashtom

It's like you're not aware RAM is a precious commodity.

Qwen 3.8 Flash Next with mtplx on 128GB M5 Max is amazing, full context I'm getting around 72tok/s, it has replaced the frontier models for me.

Fantastic review, this is how it should be done with local AI.

Very cool, hope you share your presentation for all if you can.

This isn't reddit..

It's great to see work being done to highlight an issue but I do wonder what background does the author have? Would recommend gestalt/cleveland as a good grounding, the visualizations is editorial rather than analytical.

Choosing US versus Japan, which Japan has the lowest cost and highest life expectancy in the OECD, it's cherry picking. I'd recommend showing the full distribution of OECD per-capita spending rather than just a single cherry picked comparison.

This also is troubled by McNamara Fallacy, we're looking at metrics that are qunatifiable but ignoring what can't be measured or overlooked, is speed of access being considered, how about innovation incentives, quality and outcomes variation across systems, patient choice and flexibility, in addition to workforce compensation (nurses and physicians in the US earn significantly more). Where are the trade-offs?

Summary Statistics can be dangerous. 254% of medicare is a single ratio summarizing enormous variation across thousands of hospitals and procedures. Median markup of 3.96x inherently hides the distribution, some hopsitals may be higher or lower, why is that?

I think the biggest one to me was the confirmation bias, the $3 trillion gap that represented 'fixable waste' was the conclusion. Every price difference is interpreted as waste rather than investigating the potential cost drivers, was there a null finding framework in place where US spending appears justified or is it all bad?

Overall, glad someone is looking into the data and pulling insights, please don't take this as discouragement just a comment from the peanut gallery.

This does an honest good job of walking through the beginnings, I would still say understanding/decomposing a decision tree and going through the details and choices /trade offs one makes with how they prepare the tree like binary split or discrete/binning for continuous data. What reducing entropy means, etc. Maybe even start with parametric versus nonparametric modeling pros/cons. You really get to see how probability and statistics is applied in the formulas that eventually will be thrown into a dot function in python.

There is a lot of content on pytorch, which is great and makes a ton of sense since it's used so heavily, where the industry needs a ton of help/support in is really the fundamentals. Nonetheless, great contribution!

I purchased this as soon as it was announced, I was surprised they had it ready to ship on the day of the CES announcement.

I do enjoy it, with Fancyzones, I can set up Unreal Engine Editor, Rider, discord/teams and a small corner window for searching and/or youtube watching on the side. At first I thought the pixel count was going to be too low but from my position it 'feels' retina at 125% windows scaling. Yes you can do the same with multiple monitors but I don't get the fatigue of turning my physical head, it's the perfect size to sit in middle and use your eyes to adjust/focus if that makes sense..

120hz and fast motion helps a lot. DCS World looks amazing on this, it feels like it's your full fov when playing games. Granted this isn't an OLED panel, I wouldn't play anything competitive on here but EU V and/or RTS games are very nice at 6k/52.

This replaced my dual 4K 120hz monitors. Recommend if you're not gaming.

I learned on the MIPS processor, computer organization / architecture one of the most challenging CS courses for me. I don't remember much but I definitely remember the mips pipeline...

I think you're double speaking here, the majority of the population who were eligible to vote, voted for Donald Trump in 2024.

People become statisticians and scientists overnight when it comes to COVID-19 Vaccines.

The science is very simple, get the shot and you reduce the likelihood of hospitalization or death. It's more effective for those with co-morbidities or those older in the age brackets.

I am not sure why this was politicized to death the way it has been. The mandates early on were to curb the hospitalizations / deaths that was occurring. Really the mandate should have been for people with co-morbidities or in the upper age bracket but you know what, people are stupid and everyone thinks they're healthy, so instead of being nuanced they applied it to everyone.

I think this is sort of it but I don't think it's the carrot that's the problem here. I believe it's the process and yeah ultimately the culture.

I don't think you want sales concerned about security, their focus should and only be on growth. The problem is if you don't give jurisdiction and power to the other side to actually say no this priority (security fix) goes in before work is done on this new feature, then you have an imbalanced system.

If the project manager who is incentivized toward growth is the decision-maker for deciding what is prioritized, well of course naturally you'll have the PM choosing growth over security.

Process needs fixing, give more agency and jurisdiction to the other side to effect change. It's not like security doesn't see what the issues are, it's just the fixes are not prioritized and the culture and process isn't balanced between both.

Except it's not meant as a drug you take for life, as it will require higher and higher dosages in order to get the body to respond.

The problem isn't the results, the problem is the factors that caused the obesity. I'm for it as long as it comes with healthy lifestyle changes as that's the only true way to have a long-term impact.

How much of this is just the fact that alcohol is typically 'banned' from patients taking this drug?

Sorry didn't read too much into it but are they testing the reduction specifically for alcohol addiction or was alcohol addiction measured from people currently taking it for weight-loss/diabetes?

3.23 is smooth, performance is great, a lot of gameplay loops are in. 4.0 is around the corner. It's disappointing that they split the development between Squadron 42 (single player) and Star Citizen, but now that SQ42 is feature complete, the pace at which Star Citizen is being developed has been very nice. A lot of the features from SQ42 is being released/polished for the MMO and has already seen many features ported/implemented.

4.0 is going to be by end of year which will have server meshing, at which point, all of the difficult tech behind the scenes will be finished.

I think CIG is going to license the engine once this is completed.