curious_curios

Specifically it’s data useful for training language models. Using text generated by another model has been shown to be counterproductive. This is why so many companies are trying to lock down access to their user generated content and set up data sale agreements for access to data prior to the major release of large language models.

It’s now possible to implement a rudimentary readability system in GitHub via CODEOWNERS[1] or external utility like policy-bot[2].

1: https://docs.github.com/en/repositories/managing-your-reposi...

2: https://github.com/palantir/policy-bot

Git definitely feels like it was made for power users, not beginners.

You may want to try mercurial or pijul to see if those approaches are more intuitive for you.

During a goldrush those selling pickaxes make the most.

Realistically the best thing you can do is let it go and have the satisfaction you were right. No interview is perfect, there are false positive and negative hires. Don’t take it personally.

If you’re on campus you might want to consider jobs like TA, desk attendant or lab monitor. Most of them have lots of downtime and will pay decent so you can work on your own studies/projects.

You definitely don’t need those things to be in decent shape, I’m just saying they help tremendously.

I think it’s the other way around. Give the average person an insanely high salary to spend on healthy food, designer prescription drugs and supplements, the ability to direct their own work hours, a great health plan and subsidized gym membership and see what happens.

- Reach out to investors in your greater area.

- See if your university has an entrepreneurial program or club.

- Look up uni alumni who have started their own company which has failed and see if they want to do another.

- Apply as an early employee to a startup and start building your network.

- Forgo a co-founder, just do it yourself.

Usually RSUs can be provided in either shares or cash equivalent. What is often done is a portion of the vesting for the month will be paid in cash to offset the taxes withheld. This is something that a company should walk any candidates through so they understand what they’re getting into.

They really should have no problem with providing a recent valuation. You should say that you’re trying to plan out your tax liability, because depending on your jurisdiction it’s going to be taxed. From that standpoint if it’s significant and there’s no liquidity it’s actually a liability.

You’re glossing over a lot of details here, which is where most of the pain is.

Properly chunking the data, handling non-standard text formatting in source documents, not even having OCR’d text in source documents, having disparate indexes available per client, minimizing hallucinations even with properly context data, and more.

Everything after the plus is considered a sub address and is just an arbitrary tag. You can include any text there and it’ll get delivered to the primary address. It’s useful for filtering emails, and is part of the email spec.

That’s a pretty big leap in logic.

Languages evolve and are in constant flux as words come and go and their meanings change.

It’s the same reason that we haven’t all evolved to a single universal organism.

Languages are primarily used to communicate, not discourage communication. They are also not usually unique to a single country.

People will find all sorts of reasons to create in and out groups, but I can’t imagine a country setting a policy of limiting learning of their official language. To protect their culture they’d want it learned so it doesn’t get supplanted by another external language.

Yep, it’s based on html/css/javascript usually. With things like WebGL you have a full 3D rendering engine in the browser. They’ll often use WebRTC or web sockets for game updates from the server/other players.

There’s a list of HTML game engines if you want to dig deeper: https://html5gameengine.com/

If you hadn’t built it out to start it’s pretty hard. There are a lot of nuances in how you format strings by language so you can’t just translate every string you find. Also keep in mind some strings shown to end users come from calls to the backend, so you’ll need a localization story there.

This also means that you need to add a localization gate to all feature releases, slowing down velocity.

Refactoring for i18n after the fact in a large codebase is surprisingly difficult. Many of these places have done some research and made the call they would profit more from continued feature work.

The banks are liable if they don’t police their users, just like you would be liable if you in turn gave your credentials out to someone else.

Society apparently has deemed those beneficial at large as there is not a critical mass of opposition to get representatives to push for changes. In this case I’m equating special interest groups and PAC’s with money to spend to influence as also beneficial to society as they have been allowed to prosper.

There are AI models, which are free to download and run locally, that are uncensored. That is the reality today.

Then there are platforms that host models they have developed. Is it not ok for those proprietary hosting platforms to apply guardrails to what their models respond with?

Taking this to the extreme, should providers of other services not censor their products? This includes KYC/AML for financial institutions and section 230 like provisions for social media. It seems like society deems those to be beneficial, why not similar for hosted AI products?

Note that self-hosted models do not have these restrictions.

This comes off as a political hot-take trying to score quick points, which is not interesting to me as an engineer.

What WOULD be interesting is a discussion around what the actual indicators they used and whether we could expect a nation-state actor to leave behind concrete evidence. That discussion would be deeply nuanced and constructive to future endeavors.

If you want simple, free and flexible you can build a static site blog with Hugo and host on GitHub Pages.

If you can present a reasonably coherent case that something is tied to a targeted business outcome without ethical concerns my team will get it done no matter how unsexy it is.

The names are verified, but are also verified to have been spreading misinformation about COVID as well as operating a fairly shady medical supply company capitalizing on that misinformation during COVID as well as other disasters like the East Palestine train derailment.

- Adding new vendors without regard for ongoing maintenance and compliance work.

- Ill defined requirements.

- Setting something as the highest priority without cutting something else.

- Anything from a biz person who’s burned the engineering team publicly.

It’s probably being employed to detect bots, scrapers, and users behind VPNs.

From a technical standpoint there’s probably no value as it would offer no benefit for the insane cost.

From an ethical standpoint probably impossible from the start.

Maybe it’s just me but the things I’ve tried to get ChatGPT (and other related tools) to do has fallen over really fast after you get past the initial toy examples. We’re talking complete inability to resolve introduced errors, grossly insecure APIs, misinterpreting requirements and straight up hallucinating libraries where it couldn’t figure out how to implement something itself.

For light research and rapid prototyping it’s just another tool that augments a human and I think the fundamental underlying technology will not be able to fully resolve those issues until new architectures come out.

Use the API directly is an option.