Dayshine

Or just use the programming (or other) lens for documentation

See section about the new definition under GDPR.

Like I said, modern privacy standards. The US is still quite far behind: I've had to do HIPAA training and it shows.

I'd advise you to set your bar a lot higher than your national standards if you ever plan to do international products: nobody in the US will complain if you use stricter definitions but you'll instantly be rejected in Europe if you call that pseudonymization.

By definition in any modern privacy specification, the only difference between anonymous and pseudonymous is that you can recover the true identity.

I believe a decade plus ago Pseudonymized meant simply making something less identifiable, but not any more.

Pseudonymized data is for all intents and purposes completely anonymous as long as the people you share the data with cannot access the linking table.

That constraint can be enforced technically, contractually, or legally, depending on how important the breach would be.

I wish I could install Linux on my laptop, but without HDR what's the point!

Have you ever experienced recurring or ongoing debilitating pain? Going from 1-2 days of pain per month to more doesn't sound like "no side effects".

However, it turns out it's really annoying to use text based config for most things. Primarily because integrations can't edit your config, so you have to do a lot more manual migration of config. Yaml also doesn't have any proper ide support, so it's an annoying loop of restarting the app and staring at logs looking for typos.

And finally the version control was useless because so much of what I wanted to keep was state, which can't be in config.

I've recently realised that I only actually needed the complex rules and UI to be in code. And they still are: now through the ui with instant validation.

If carton took a TF/pytorch model and just dealt with the conversion into a real runtime, somehow using custom ops for the bits that don't convert, that would be amazing though.

Porting the same extension to sublime text was a nightmare. Python has poor language support in all ides, the docs were lacking, there was no good debugger and I couldn't just check the code.

9.8

It's always 9.8.

To be fair, this is an application, so the score might be more grounded in reality. The CVEs I encounter as a developer are always insane and generally massive overreach from security to product design. It's not a vulnerability against all downstream libraries and applications that using some obviously unsafe C++ call might cause a DoS due to poor performance.

All of my banks use a mobile app for confirming transactions, which requires me to login. Sometimes that requires reauth not just biometrics. I'm not going to go home and try and type a 20-30 character password into a phone when trying to pay for car parking.

With a flight there is 2 hours of standing queueing or waiting in a "gate", then another 30 minutes of dead time at takeoff and landing where you can't do anything either.

Nothing that actually helps you use it in any real context!

I had to use a proper training tool (vim adventures) before I got enough practice for things to click. And I had to discover a handful of plugins before it was at all usable in a work environment (acejump, code aware motions, IDE integration).

Vimtutor is like a jira tutorial. Yes, you're telling me what I can do with you, but what I need to know is how to use you.

Tensorflow has breaking changes in model behaviour on patch versions!

Pytorch and tensorflow require a huge python environment to run, and how on earth do we get that onto a client system without them having virtualisation?

Worst of all, they both have very significant changes in prediction depending on the CPU (or god forbid GPU) on the system.

Once we export to Onnx we find we get reliable output and performance across runtimes, which seems mandatory for running any kind of product.

Is saving a tiny bit of money that much more important than a smooth handover? Or are employees just not willing to work their notice periods?

It sounds like in this case it's actively going to hurt the business.

A model saved in 2.0 should work in any later 2.x version.

In practice, they've changed the implementation of the batch normalisation layer at least twice, causing 3 sig fig differences in a unet.

They just don't seem to care about production use of tensorflow.

The final image had no plagerism.