Shouldn't the end goal be just to train an ai on all the pdfs and give the doctors an interface to plug in all the details and get a treatment plan generated by that ai?
Working on the data structure feels like an intermediate solution on the way to that ai which is not really necessary. Or am I missing something?
AI/ML techniques in medicine have been applied clinically since at least the 90s. Part of the reason you don't see them used ubiquitously is a combination of a) it hasn't worked all that well in many scenarios so far and b) medicine is by nature quite conservative, for a mix of good and not so good reasons.
Pi-hole is a bloated mess compared to this IMO. At the end of the day pi-hole is still just a fork of dnsmasq with a load of scripts and a bootstrap gui whacked on top. You need to add on extra bits and pieces to get anything like modern tech whereas AGH has https gui, multi-user support, DoH/DoT/dnscrypt/etc, toggles for quick blocks, access to a 'realtime' blocklist for emergent threats all baked in. It's also a single self-updating binary with a single config file instead of spraying bits all over your OS. Runs on pretty much anything you can think of, too.
pi-hole was great back in the day but unless you're just keeping on keeping on with an existing install there's better options available now.. AdGuard Home, Blocky, Technitium DNS etc.
I often compare pi-hole to DD-WRT inasmuch as it was awesome back in the day but times have changed and you probably wouldn't use it as first choice these days given what else is now available to you.
You can be fully protected just using vanilla dnsmasq and downloading fresh blocklists from time to time. It seems all the more ‘marketed’ flavors of adblockers are just web bling on top of dnsmasq. What else do they really offer?
Well 'the alternatives' are many so there's no quick answer to this, but restricting to just AGH as per this post then...
Encrypted upstream lookups.
Responding to encrypted lookups made to themselves.
Realtime threat protection via API.
Quick toggle of blocks instead of rebuilding lists.
Ability to quickly change blocking of individual devices.
Decent Metrics.
Probably more.
But if you just want something with no web bling then there's other alternatives to dnsmasq which would be worth looking at which give some of the above features whilst keeping it commandline and manual blocklist building.
dnscrypt-proxy is wonderful, for example, and can do most of the stuff you can do in dnsmasq.
Anecdotally, I’ve been a sysadmin for 20 years, been around computers since I was a toddler (apparently slept on top of a Data General something-or-other as a baby..). I have the skills to learn how to do dnsmasq-based blocking from scratch, write the scripts to fetch blocklists, init scripts etcetera. However, I run AdGuardHome on my OpenWRT router because I want to spend my time elsewhere. It was a case of install the package, fiddle the DNS routing slightly, pick my blocklists, and pick my up streams.
If I want metrics, I just open a browser and see what clients have been the noisiest, what’s being blocked a lot and so on. Generally I don’t even think about it.
I can easily see what domains are blocked in the web ui and see that Adobe products are trying to phone home so often and which clients are trying to resolve what domains.
I keep both on my network running on two different raspberry pis.
AdGuard Home is a lot cleaner to use. In particular it makes it much easier to control routing for queries by domain and supports forwarding over DNS over TLS, DoH, and DoQ natively. SSL support is a breeze. This means that my ISP can see the IP addresses of hosts but not their domain names unless they get aggressive with snooping. The single binary and clean configuration is nice.
PiHole seems to have a better landing page for analytics out of the box. It also works a little better for configuration for some devices.
I’ll likely retire PiHole in favor of AdGuard Home the next time the SD card dies on that Pi.
My preferred configuration is using some fairly invasive scripts to redirect all outbound DNS except to NextDNS. I’ve got blocklists for DoH hosts because I can’t just block port 443. AdGuard then routes to one of two different backends: for local domains it routes to CoreDNS that gets the hosts from my UDM-Pro to give everything nice hostnames. Everything else goes out via DNS over TLS to NextDNS. On PiHole it’s a little more complicated as it can’t directly forward with DNS over TLS.
It’s amazing how many semi-hostile devices this found on my network (looking at you Samsung TV and devices that hard code in Google’s DNS). It also reminds me of how terrible the internet is when I don’t have these protections.
“Trillionsof dollars were spent on wind and solar projects over the last 20 years, yet the world’s dependence on fossil fuels declined only 3 percentage points, from 87% to 84%.”
Core issue is he’s at best justifying his points with sloppy thinking, we didn’t just transition to solar for fun we got actual energy from those investments. The world also spent around 100 trillion on just oil over 20 years excluding natural gas and coal yet tomorrow we are going to still need to spend more. So, on that basis it looks like a great investment and huge progress.
Except look at where exactly that 87% number is from, it includes fuel oil to heat homes while excluding passive solar gain. So it’s a number that does represent something just not “dependence on fossil fuels.” After all the primary use of energy by humanity is either plants converting sunlight to chemical energy or even more holistically the energy needed to keep the earth from freezing. In the end he picked those numbers because saying saying humanity’s dependence on fossil fuels dropped from something like 0.00087% to 0.00084% just doesn’t have the same impact.
1.5 extra hours would be my expectation as well.
If we consider a long trip for vacation 3 times a year, the 4.5hrs invested into charging is offset by less day-to-day charging stops with an EV (if you can charge at home), the reduced costs of electricity vs gas and the reduced costs of maintenance.
I would optimize the car choice for the 90% use of daily commute instead of optimizing for the 3-times-a-year vacation trips.
Did the exact same trip with a Model 3 2021, 1200km ski trip from Germany to Austria. Came back yesterday. I experienced it way differently.
- We were able to make 300km with 90% to 10% battery (to not hurt the battery longevity too much)
- Outside temp was -4 to +2 °C
- Inside temp set to 20°C, seat heating 2/3 for two passengers
- So a we made a charging break every 300km, so approx every 2-3 hours
- Recharging those 80% at a supercharger takes about 30-50min depending on the Supercharger-version.
- We had 0 traffic/wait times at the super chargers (we drove both directions on a sunday)
- We would do a 10-15min break anyway every 2-3 hours to grab a coffee or do magic pee, so the extension of the charging breaks over our normal breaks aren't event that long
- All superchargers had a <5min detour from the Autobahn
Overall we spent approx. 2hrs more on breaks as we would have with a conventional car. I think thats a fair trade-off for 2-3 vacation trips a year, figuring in the time saved for normal refilling stops with a non-EV cars during commutes (when you are able to charge your EV at home).
To me, the future of "driving into holiday fully electric" is already possible with a Tesla LR model. With other EVs without Supercharger-Access/smaller battery/slower charging speeds probably not so much.
You can even save more time by using tools like ABRP[0].
This even gives you better charge-planning with shorter, time-optimized stops also figuring in detour times.
The thing that gets me with the success stories is the issue with risks and planning for success only.
When you’re driving in the colder parts of Europe it’s generally advisable to keep your tank at least 50% full all the time. If the shit hits the fan, like it did for me in Switzerland once, and you’re stranded for 4 hours due to a crash out of your control, your car becomes a fairly important life support system until the road is cleared. There is no recovery option when there are a few hundred cars in the same shit.
So you’re 3 miles from a supercharger with 15% battery left and your car is a frozen brick in under an hour. You can’t deliver more fuel to it and your efficient route plan is a liability and there’s a queue of bricked EVs waiting for flatbed recovery.
I’m not criticising the concept but the current execution and the perception of it.
> So you’re 3 miles from a supercharger with 15% battery left and your car is a frozen brick in under an hour. You can’t deliver more fuel to it and your efficient route plan is a liability and there’s a queue of bricked EVs waiting for flatbed recovery.
There are a few things to consider.
First up, if you want to optimize for energy efficiency then the best option in an EV is to carry a blanket and rely on the seat heaters as much as possible.
Consider this scenario with a long range model 3. 80kwh, 15% means you have 12kWh available (Let's drop that to 8 due to cold weather). The seat heater consumes 500W at low power. That gives you 16/people hours of heat.
But let's say you just run the HVAC straight. You've still got 1 hour of heat (assuming it's using the 6kw restive heater. More if you are using the heat pump).
In any event, the approach to "I'm in an EV and stuck in traffic" is exactly the same as if you were in an ICE with low fuel. Shut things off. Wait until you are freezing, turn it on again. Ration your fuel/energy until you are unstuck.
To get to your charging destination in this scenario, you need roughly .9kwh of energy (300wh / mile, which is on the high end) or about 2% of your battery.
This is good practical advice for how to maximize EV energy use in an emergency.
That said, these vehicles are clearly not made for a long haul cold weather existence. Can they be made to work for occasional use with some planning and prayer? Sure. But these stories and the counter-claims leave no doubt that people who park cars in -30F and do 300 mile one-way trips through mountain passes and deserted highways should stick to ICE vehicles. If you're doing some variant of this with regularity (say 0F and occasional 150mile trips), you should probably do the same for safety reasons, even though EV will probably be fine for all but the rarest disaster.
Regardless of ICE or EV, people doing this kind of driving are well-advised to prepare for having a non-functional vehicle. It's always best to stay with the vehicle if it's habitable and if rescue is what you can count on, but having cold-weather gear, food, shelter, means to create/use external heat sources, and ability to "hike out" are some basic rules of the cold road that even 'tourists' should abide by. If you're doing remote winter driving, basically also pack for winter backpacking. This in addition to road flares, small shovel, tow strap, etc. to support the vehicle. It's less necessary now than it was in, say, the eighties when vehicles were far less reliable, but it's still practical advice: be prepared.
> That said, these vehicles are clearly not made for a long haul cold weather existence. Can they be made to work for occasional use with some planning and prayer? Sure. But these stories and the counter-claims leave no doubt that people who park cars in -30F and do 300 mile one-way trips through mountain passes and deserted highways should stick to ICE vehicles.
I'm guessing this is hyperbole, but really, this is a scenario that does not exist pretty much anywhere outside of Alaska and Russia. And even still, you'd probably be shocked (heh) at how many fast EV chargers present in these locations.
Every year, the situation with EV chargers has gotten better, by a lot. Consider the fact that John Day OR has a fast charger [1].
There aren't many places in the continental US more than 100 miles from a fast charger. Very little prayer is needed, though a bit of planning is nice. It's hard to find a location that you can't comfortably reach with an EV that has 300+ miles of range.
Canada has high latitudes as well. Only part hyperbole, which is why I make the statement that even if you're doing only 0F and 150m, the same prep is a good idea. Really any drive where, in the event of your vehicle breaking down exposure would be life threatening (Alps, Rockies, Cascades, you name it), you should prepare accordingly whether EV or ICE. Having climate control in your vehicle should not be your primary contingency in the event that forward movement ceases.
Considering EV's lose range based on temps, that charging stations are still sparse in hostile environments, that a passers-by can't siphon out fuel for you (I know an ICE can jump start a Tesla, but not sure how practical it is to charge a Tesla from an ICE), ICE engines are still a better bet if you're in such conditions.
Car and Driver did a test using an older Model 3 with a resistive heater (the newer models use a much more efficient heat pump), and found that it used around 2.2% an hour to keep the cabin warm.
In your example with 15% left, you'll use ~9% battery while in traffic for 4 hours keeping the heat and car on, leaving 6% to get to the next charger. At ~300wh/mile you'll arrive with ~4-5% left. There's also buffer under 0%, but it's not guaranteed.
4-5% is not a comfortable number to be at, but I think it's acceptable in a worst case scenario like this. That being said I would definitely turn down the heat, and drive slower for the next few miles (and check for alternate chargers) to minimize power usage.
I really wonder how a heat pump can achieve so much efficiency gain. Heat of electric resistance is about 99% energy efficient: almost all energy is converted into heat, nothing else.
A resistive heater directly converts electricity to heat, whereas a heat pump instead moves heat, from the outside to inside of your car. The heat your car gains is reflected by the heat the outside loses. In that way, they can be 300-400% “efficient”, because we do not care about the outside air around the car getting a little bit colder.
It just goes back to the battery to complete the circuit. Heat is created by resistance and if the energy going through the heater isn’t being resisted enough, there is no heat, just wasted energy. Most of these heaters regulate temperature by turning off/on/off and it’s the coming up to full resistance that is when the energy is “wasted” and doing it at human temperatures is a lot of off/on cycles. Also, it’s worth pointing out that not 100% of the energy is converted to heat, or you’d have a gigaton bomb instead of a heater.
I do ~900 km trips with 2019 Model 3 SR+ multiple times per year in Scandinavia. -20°C winter temperatures are not unusual. Travelling with my family, including young kids. No problems. 2-3 hours driving, then 30-50 min charging. I actually like how EV forces you to have more breaks.
Not parent, but calculation from given data: 1200km, charging every 300km for about 30-50min -> 4x 30-50min -> 120-200min -> 2h-3.3h in one direction -> 4h-6.6h in both directions
I did a similar trip earlier in January. 1100 km from Germany to the alps (550km per way).
I had to stop once for the full 1100km trip to refuel (before leaving Austria on the way back cause it’s cheap to tank there, not because we actually needed to refuel).
It was ~4:30 per way. Nobody in the car had to pee, etc.
I think doing 15 min break per way would have been ok. But doing 1h per way or more if breaks is really not great. Particularly with kids in the car (we leave really early put them as sleep, they woke up for the last 2:30 h or so and that was borderline).
> - So a we made a charging break every 300km, so approx every 2-3 hours
> - Recharging those 80% at a supercharger takes about 30-50min depending on the Supercharger-version.
How is this so totally different to the above commenter? I also need brakes during my rides and after a three hour ride a break sounds totally reasonable. But nowhere between 30 to 50 minutes?
I like to start with code.
If you use typeorm and type-graphql you can write a Type once and generate the database as well as the resolvers out of it. You can then even import the same type in frontend (as part of a mono repo). This stack does not need any generation routine for types. You just have to generate a migration if the postgres target schema changes, but that's normal and I don't see a way around this for the future.
Typeorm looks nice, but I need polymorphism for some of our "business" logic and have pretty specific requirements about how I want correctness enforced that typeorm doesn't appear to support (but raw SQL does).
Typeorm does not scale to large datasets. Real nice for MVPs. Just a warning. Hydration times explode for lists and joins that postgres barely registers on the query side, and hydration is blocking.
I am from germany, and the avg. salary for a decent senior dev here would be between 60 and 80k€ depending on the town you are living in.
Babel is a very complicated piece of software. Even as a senior dev I wouln't feel comfortable to lead-maintain it.
You wouldn't find more than a couple of ppl. in Germany who have the skillset and mental strength to maintain a project like that. And those few people would easily be making 150-250k€ as freelance consultants.
I think that 11k$ for maintaining one of the most important OSS projects in the webdev world is more than fair.
I'd expand on one of your best skills you already have.
The "stack" gets more and more complicated so "full stack" will stop to be a thing in the future.
If you are already good in react, learn more libraries / frameworks in that ecosystem and increase your day rate. You are a frontend architect now.
I'd start with things like nextjs, Gatsby or Redux saga.
You could also go into the devops side of Frontend. Learn the different approaches of deploying / scaling a Frontend with server-side rendering (vercel, digital Ocean, docker...)
