bitexploder

You can also ablate it.

It is a nice pelican, though. At least it has that going for it.

It gets better in the cooler months when heating is running in a home :)

I wonder if that is an adaptation time problem? It takes about a month for the body to fully adjust. I live at 1500m elevation in the Denver area of Colorado, but I do recall feeling more restless during sleep at elevation before I lived here.

Magnesium is a great supplement in general. You can definitely have too much in your system and that is undesirable, but a bit before bed time along with 1-3mcg of melatonin work well for me. It is nice after workouts. I use magnesium glycinate in powder form, which is more bioavailable than some forms as well.

I also find sauna before bed is good. I have a bed chiller so I can crank up the sauna before bed and not sweat a lot. Generally if I sauna and take the aforementioned supplements I sleep well. Exercise also seems to help me out a lot. If I exercise during the day, and a 4-5 times that week in general, I tend to sleep well.

Yes. Mythos is almost exactly that. Willing to do in depth vulnerability and POC work.

When a hyperscaler is viewed as a software company their stock and value multiplier is much higher than if they are viewed as a commodity with expensive infrastructure costs. There is now not enough compute resources to serve demand. It requires sustained capital to grow compute resources. The costs are uncollapsing due to the overall demand plus the pressure of LLMs. Capital costs matter.

I just work here. The label is for everyone else :)

That is pretty much the best way to improve cardio as well. There is some benefit to long slow cardio in that it induces the same mitochondrial growth and fitness improvement, but it requires more time. The injury risk is a little lower for low intensity work. Interval training is a proven quantity for cardio fitness as well, though. I don't think you can min-max cardio fitness much more than that. Mainly, just ensure you aren't doing anaerobic work (e.g. sprinting) if your goal is cardio improvement. But that is easy enough, if you really want to baseline it, you just run a mile all out and there are a bunch of calculators that are very good which give you good paces for VO2 Max / intervals. The goal is you want to sort of keep your HR elevated once warmed and stay in that high VO2 max zone sort of ending right at max HR on your final interval or two pushing yourself up to the limit. It ends up looking like the strength training wisdom at the end of the day. Do cardio. Make sure it is challenging and you aren't phoning it in if your min-maxing. Be consistent.

For strength training, I think it comes down to what I captured in my blog post. For a given muscle, if you are challenging it and truly reaching 1 or 2 reps in reserve there isn't much else to do per set. Hit your desired training volumes and you are the happy place for time+gains. If the intensity is there and you are targeting your entire body at the right volume level every week it will come down to nutrition and other factors vs. what you are doing in the gym.

I think over 60 you can, it’s just slow, hard, and requires high diligence.

If I can make a recommendation, compound movements are amazing, but be careful with deadlift. There are many more practical exercises that don’t have so much injury risk. Volume and consistency win over heroic or ego lifts :)

"If you do aerobic exercise, almost all the energy comes from burning fat. " This is directionally incorrect. Your body will burn both concurrently. For low intensity aerobic exercise, fat is used as the dominant energy source. However even at moderate intensity levels like jogging and "zone 2" aerobic you are 50/50. At higher intensity you have crossed the inflection point and are using more glycogen than not. All strictly aerobic exercise. And it all works on a balance anyway. You use glycogen, it gets replaced until everything is topped off. Doing that means it isn't getting converted to fat.

Both forms of exercise are shown to have an "anti-hunger" effect.

And unless you are walking, your body is also shunting blood away from your gut which also has a secondary hunger dampening effect as it doesn't resume blood flow too it immediately.

So for anything we would call aerobic exercise, that is zone 2 "cardio" or greater, I would have to disagree with your main claims about it.

However, if you are a male and age 50 you can definitely expect to still build muscle up to 60 if you are diligent with your strength training. You can maintain mass 60-70. You do need a little more protein. I collected as much proven data and studies on this as I could: https://stealthgoat.com/building-muscle.html

This study is completely unsurprising to me having read a lot of fitness studies over the years. Work muscles harder, muscles get stronger. That is how hormesis goes. The fat loss is simple energy expenditure. You are still producing roughly the same work as someone doing more steady state work. Only effect that might come up is post exercise metabolism elevation but that effect is relatively small and probably present for both groups.

Nope. MLX in LMStudio. The simplest config with zero tuning effort.

It also comes down to inference speed, not "can I run this". 8-bit quant is quite a bit slower on an M5 Pro.

For a MBP I have 48 GB of RAM M5 Pro. It runs at about 12-14 t/s at Q4, you could probably optimize it further. RAM is not a limitation but overall memory bandwidth. Q8 is slower. 35B A3B Qwen is quite speedy, but a little less accurate. With Qwen 3.6 27B dense I can squeeze a 9B parameter model and use that for fast analysis or code scanning while 27B is churning on a task in the background. It is tight, but totally reasonable.

The real sweet spot for Qwen 27B is getting it on something like a Dual 3090 system or some other config where it can blaze at 50-80 t/s and that costs well under 6K currently. It is a surprisingly capable model. Using something like GLM for orchestration, specs, task farming and then letting Qwen churn is relatively inexpensive.

Overall I recommend people try models of this class out using OpenCode and some for pay service to experiment with them and understand how they work. I find they are very useful.

Long term, I am convinced enough that if I wanted to use local models for any number of reasons I would be okay investing in a dual GPU box. The Mac is not fast enough for me and M5 Max is just too expensive relative to GPU linux box. Still, it is nice to have the models local ON the laptop and it is useful for what I care about locally.

They can only slow down open models and open weight models from getting as good as Mythos/Fable and GPT 5.6. Then what?

They will all get distilled, down trained, and the smaller models will get better too.

You've got the direction right, but the numbers are inflated in the Slate's disfavor. The 1.12m is an unmeasured estimate of the hood's high point; the figure that actually drives risk is the leading edge, realistically ~40–42" here — the very floor of the "tall" category. The "50%+ solely from the hood" is really IIHS's 45%, and that's a bin average for everything over 40", a group dominated by full-size trucks whose leading edges sit at 46–50"+. A vehicle sitting right at the 40" line doesn't carry that average, and the 45% bundles height with front-end shape, so it was never "solely" the hood. The F-150 and Silverado people are actually buying are a clear tier worse than the truck you're complaining about.

Kids, women, and the elderly do take a disproportionate share, but nobody measured anything "exponential". "Dozens of studies" is fair. "Exclusively for aesthetics" isn't, though: the front houses a 7 cu ft frunk, and more importantly the Slate's short hood and modest height give it a smaller forward blind zone and better sight-lines than a full-size truck, which is the axis that decides whether the driver sees the kid at all, and the one the 45% fatality stat doesn't even cover. The real villains here are taller, longer-nosed, and parked in way more driveways than this thing will ever be.

This vehicle is a step in the right direction, especially for the urban environment. If we really cared, everyone would drive a Honda Acty in urban zones.

This vehicle is likely to be highly customizable for almost any interior or cosmetic element. People love customizing vehicles. The number of people who get super into the most inane customization details in the Porsche world cracks me up. "Oh yeah, I got mine PTS (Paint to Sample) color Pacific Northwest Green Pine in Late Fall, but with a hint of sunshine.... and check this out I got custom deviated stitching in this place you literally cannot see once the door closes, no one else really has that!" Same thing here. This level of customization gives people a connection to the vehicle at a very affordable price. It is an EV. It is practical. It can be upgraded after purchase. I dunno, it has a lot going for it.

They designed the vehicle to be "easy" to wrap. Meaning you don't have to remove trim or anything like that. It is still non-trivial and you would want to do it inside of a garage IMO.

The hood is very short and you sit way closer to the ground. I have driven one of their comparison vehicles (the 1985 toyota) with a very similar profile. There is no way you are missing someone in front of you compare to other vehicles. This thing sits lower and gives better visibility than many SUVs people are driving. This truck is small.

If these become popular it should be reasonable to expect the aftermarket to provide less expensive options. These are clearly where Slate's margin will live right now.

Slate does not wrap the vehicle at manufacturing time. You do it or a service center does it for you.

The body panels are some sort of composite polycarbonate plastic. They offered a number of colors in the UI for the pre-order system. It basically allows you to configure options that are available and see what it looks like. The wraps are (offered) DIY and they claim it will take two people 14-16 hours. They offer a number of colors that cost $500 right now.

Pretty interesting as far as cars go. I think all of the customization options are really smart, but we will see if that is enough for the average consumer. The number of relatively inexpensive options and ability to customize various details rivals some high end car configuration systems, Porsche is famous for letting you customize every detail.

The way Slate has designed their vehicles, they are making it very easy to change just about anything with the vehicle. I can see this being very popular with young people who cannot afford a lot of car, but still want a highly customized and personalized vehicle that they identify with. We will see if all of the other factors work out in Slate's favor.

I worked in Manhattan often in the 00s and early 10s. Have people forgotten what big city life was like before? Commuters did not randomly strike up conversations. It was an unspoken code you left each other alone. Especially in rush hour commutes. Everyone is waking up or tired after a day of work. It is more about having some personal space in a crowded environment for many. Not everyone processes or experiences that the same way either.

I call it "Manhattan Projecting" them. The amusing thing is I had Fable review my harness (which I have been building for some time) and it helped improve it. It is just kind of funny that it enthusiastically helped build a harness whose sole purpose was to divide agents up and compartmentalize security sensitive vulnerability research.

I also have a 100% success rate jail breaking them by breaking the work down into small pieces and stripping all security related language. Smaller tasks, test engineering and normal programming language. Fable found a few bugs in my harness for me before they pulled it. I was testing it vs ChatGPT, Gemini, and Opus. It was doing well at bug hunting.

What if I told you there are no safety guardrails. I used GLM 5.1 and had fable literally build a harness to avoid triggering guard rails. I built skills carefully and had Fable doing vuln research and exploit repro in a few hours. I called the project manhattan. The GLM models are down for almost anything so I named it Oppenheimer. It orchestrated the fable CLI agents via tmux. This whole Fable/Mythos thing is such a fucking joke. It is all PR and theatre and they know it.

But a good harness lowers the model floor and accessibility and makes stronger models that much better.

I don't even care. It is the same problem advent of code had as a public challenge with a leader board. I now mostly just think either embrace the LLM or keep it to a more in person or vetted audience. But, again, if you create a competition in the spirit of humans without LLMs and that is in the rules and someone uses an LLM that is on them IMO. I am sad advent of code decided to end their competition. LLMs are here to stay, let's embrace that and see what the new universe of competitions with LLMs can be. There will always be a place for human only competition, but for public facing ones LLM accepted is the only tenable position.

This does bring "Pay to compete" concerns and create incentive structures that encourage more LLM use. I don't know what to do about it.