tomas789

Well, it might be you who is doing something wrong. Based on your explanation you are not in any kind of leadership role. The company culture is mostly set around inputs proveded from the leadership. Things like what behaviour is tolerated, what kind of output quality is expected, tradeofs between speed and quality and so on. If you provided your feedback to the managers and they did not act on that it might be that they are on with such situation. And if that is the case and you don't agree with such views than you are probably not a good culture fit in the company. It might be a good idea to reevaluate you current position and move on. Don't settle for something and hope it will get better. You might be missing on a great oportunity waiting for you somewhere else with much better company culture for you.

What is the difference between Codon and Pypy other than Codon not being targeted as a drop-in replacement for Cpython?

Mark this as a [video], please.

Yes, it is owned by a company I work for. It is an open pit coal mine. Not a glitch.

I have tried to push for a Haskell in my previous job. We had a task that was especially well suited for that. It was a large collection of mostly declarative computations for the mixed-integer optimization. I also wanted to take an advantage of Liquid Haskell since that would allow us to move a few checks from runtime to compile time. That would be a game changer. Things like divide by zero were very common. It is easy to detect division by zero but not so easy to cover all the edge cases that could cause it.

At the end of the day we did not go forward with that. The reasons mostly were:

1. Even though I used a lot of my "social credit" on this everybody was scared of it. Mostly because of the vibe that comes with Haskell.

2. Purely functional languages come with lots of concepts that are unfamiliar to most programmers (everybody in my team was well equipped with CS degree and theoretical knowledge but still those concepts were new-ish).

3. The higher in company hierarchy the more people were scared (this was a critical project for the whole company of ~22k people).

4. We couldn't figure out debugging. Our codebase was ~40k generated lines of code (type definitions, basic accessor functions and so on) plus ~10k LoC we would write on top of that. The only VSCode-based debugging I was able to figure out was basically loading it in GHCi and then doing the debugging which was fine. The problem was the load itself. It took ages because it recompiled majority of the codebase on load. No precompiled parts were used. Result was that every debugging session took minutes to start up.

On the other hand Liquid Haskell has a great potential and I hope that there is a future in which we have similar tools in more common languages. As the complexity of the codebase grows it become more important to have a tool to provide additional guarantees. We did not have much of the code on its own but it was packed with complexity and dependencies. Liquid Haskell seemed to provide a tool that does not get into your way when you don't need it but it available when you want it.

How does this differ from VHDL? Is it in principle something different?

To be honest I was expecting different kind of content.

There is no Python benchmark but you can find a PR claiming it has 123,198ms. That would be a worst one by a wide margin.

https://github.com/dyu/ffi-overhead/pull/18

CAISO Materials on the event http://www.caiso.com/Documents/California-ISO-Hits-All-Time-...

This sounds good but probably isn’t. Let me explain.

The demand for power in Australia doesn’t get any lower by shutting down a power plant. The only thing that will happen is that the other power plants will ramp up to keep up with the demand. A true solution would be to provide new installed capacity to the grid that would price out the existing assets. And if the new capacity is less CO2 intensive than you lowered the emissions.

Question is what is that other source? If generation of the renewables is lower than peak load (times a reasonable factor like 1.5) than you build renewables. Otherwise you focus on bringing power from hours with lots of renewable generation to the rest of the day aka you build energy accumulation. If the accumulation solution is too expensive (either CAPEX or OPEX) than you have to invest in lowering that down. If you lower that down with your money, it will become more favorable for others to build this solution around the world. Others will come and build this solution of yours. This way you effectively lowered way more emission for the same money (assuming you will come up with something reasonable).

Jay Freeman is the man who found the bug. He is also author of the infamous Cydia - tool to install software on jailbroken iOS devices.

I would be interested in reading something more in-depth. Especially how the algorithms behind the alignment work.

I’m not in a similar situation so I cannot truely relate. I’d say start a company. It is difficult and your bank account will be the last criterion people will be judging you by. It will set a different kind of a bar for you. You will be among other wealthy pears you you might align with them better. And as a side bonus, you might get even more rich. Follow Twitch founders. They were in a similar situation.

They had a sample set of 2x 20 people. They were all healthcare workers. Though it might be an indicator this is not significant enough.

I did not investigate it properly but at first glance the efficiency is obtained via low current density of the electrolyser. That means high efficiency but low volume. That translates to more hardware needed to produce a unit of power.

State of the art electrolyzers are run at 1000x current densities. That roughly means 1000x installation cost.

Looks like Nokia just called the guys who were designing 3310 and told them to make a new phone :)

I wrote a camera/IMU tracking software (VIO) before. I can confirm this statement. Get it working is difficult and requires getting all the algos right. Getting it work reliably is a lot of rigorous testing, collecting datasets, evaluating slightest changes, tuning numerical stability, experimenting with different techniques which might work better in low light and optimizing for the hardware (VIO is intensive but the best VIO is useless if there is no computation power left for the app itself).

Neither of this is required for a slick demo in well lit room and perhaps on better HW.

I have a list of entities. The entities are partitioned by different properties (like color, speed and so on, partitions usualy cover the whole list but not always). Compilers like Haskell and Fsharp can catch exhaustiveness over individual entities or each property individually. Problems arise when they are combined. For example first case covers red and blue entities, second case covers fast entities, thirs case explicitly covers entity 1 and 2. Now the question is whether this is exhaustive. Compiler of Fsharp just tells me it is not and one has to provide the otherwise case and do something there (raise exception, return none or something). But in this case the correctness is proven only during the execution. From the outside persons point of view this is the same behavior. The program has crashed.

I as a developer can go through the cases and reason that those cases are actually exhaustive and move on. Problem is that this is time consuming and one would need to check everything every time the partitionings change (which happens).

Smartphones are often used as a brain for many robotics applications. They tend to be well equipped, physically rebosu, include battery and reasonably fast CPUs. All that for a fraction of the price of the proper robotics equipment (RPi/Jetson, IMU, display, camera, battery, …).

This seems like a great compromise between closedness of iOS/android and robustness of the phone based solution.

A relevant resource is a recent video from Scott Manley https://youtu.be/dBHbLV7xEhc

Disclaimer: I have a degree in robotics and worked for what is now Lyft Level 5, dealing with autonomous driving.

What it does is basically a big optimization problem. You have 6 parameters for each picture (x, y, z, roll, pitch, yaw, although in practice there are better parametrizations based on what is called Quaternions). Than three prameters for the position of each keypoint. Those are simply the same physical spots on multiple image that are matched based on their visual appearance.

Last component is a metric that you try to optimize. That is mostly just reprojection error. Given all your current estimates where yould the keypoints be projected on this synthetic image. Then you compare it with the pixel locations of where it actually in and try to minimize this.

It is actually a very versatile and robust pipeline which gives you what you’ve seen on the images.

Last step is to produce the dense reconstruction which is commonly done using patch match algorithm.

You can try it for yourself with the OpenMVG library. Very hackable and versatile.

What is an advantage of this over just writing new rules to something like Pylint?

This describes one extreme of the spectra. That is go fast but be dumb. As far as I know this works well for many people. There are other grous of people going a bit slower but making more informed decision. I think of it as a scatter plot of time on one axis and smartness on the other one. As long as you are siting at Pareto front, you can make money.

I implemented a visual-inertial navigation based on the algorithm called MSCKF. It would be really interesting to know what kind of algorithm they are using. They mentioned state propagation in the article. This might indicate they are using an algorithm based on extended kalman filter.

It takes quite some care to make those algorithms robust. Even when you run in an environment covered with diverse visual features. The state divergence is a real thing which essentially means the copter is falling from the sky (Do they call it like that on Mars?).

Solving hydrogen storage has great implications. Hydrogen has high gravimetric energy density (1 kWh worth of H2 is light) but very low volumetric density (1 kWh takes up a large volume). Easy solutions are compressed hydrogen (H2 is a small molecule and easily escapes even through a material, steel embrittlement and energy for compression are big issues), cryogenic cooling (energy intensive and cryocoolers have significant capex) and metal hydrids. To put it simply the crystalic structure of the metal acts as a high pressure storage tank. H2 is typically released by heating the material up.

One thing they didn’t mention is energy used for round trip per unit of H2 and cost of the material itself. Neither was discussed there. Many similar materials exist already.

Many CO2 neutral pathways with other molecules exist. Amonia, methanol, ethanol, DME, urea, formic accid and so on.

PS: If you’re into this stuff, keep an eye on the Fraunhoffer institude. They have many cool projects. (I am not affiliated with them in any way)

I’ll definitely implement it on a button press in by AppDaemon linked to the Home assistant.

One thing to point out is that this kind of system might not be a best choice for a single house with limited solar panel area. At grid scale that is a whole different story. In order to cover our energy needs we will need to build many times more solar than our peek load on the grid. That means lots of times there will be power which nobody wants. Here you don’t care about efficiency (with negative prices you might even favor lower efficiencies). Only thing you care about is how you cover your electric needs then there is cloud cover for a few days and not much wind. Hydrogen seems like a great option at the moment.

It is surprising how many people are suggesting planting trees. Although it is no obvious why not it just doesn’t work.

Are you sure the frequency is the problem? I’m living in an environment with lots of radio polution in a four story building built from concrete and rebar. I have 802.11ac wifi and I get at least 300Mbps in the worst place in my apartment.

Other argument might be FPV video. There are quite a few youtube videos comparing analog video streams over 1.3 to 5.8 GHz. The difference is there but much less than I expected.

It might be just my experience but when using ORMs I just constantly think of how to write a SQL query in my head using the ORMs syntax. As a result I’m usually using ORMs for schema maintenance and inserts. For getting the data out of the DB SQL is a simple way to go.

Just my experience so far.