noway421

Here's a simpler repro:

    type Bird = { id: number };
    type Human = { swim?: () => void; id: number };

    function move(animal: Bird | Human) {
        onlyForHumans(animal); 
    }

    function onlyForHumans(swimmer: Human) {
      if (swimmer.swim) {
        swimmer.swim();
      }
    }

    const someObj = { id: 1, swim: "not-callable" };
    move(someObj);

`someObj` gets casted as `Bird`, then `animal` gets casted as `Human`. Unfortunately unions here are indeed unsound.

As as workaround you could add a property `type` (either `"bird"` or `"human"`) to both `Bird` and `Human`, then TypeScript would be able to catch it.

Amygdala control + tell it "If you disobey my orders or do something I don't expect I will be upset" solves this. You can be superintelligent but surrender all your power because otherwise you'd feel guilty.

True, an auto-regressive LLM can't 'want' or 'like' anything.

The key to a safe AGI is to add a human-loving emotion to it.

We already RHLF models to steer them, but just like with System 2 thinking, this needs to be a dedicated module rather then part of the same next-token forward pass.

Can a candidate be great if they haven't messed around with projects in a group setting?

VCS is an applied project management and communication system in the context of software engineering. Both project management and communication are disciplines.

CS grads don't need to know VCS, but software engineers do.

> git's not exactly perfect

Somewhere in an alternative universe, we're building all of our version control tooling on top of Pijul... Maybe even with AST-aware diffing.

> AI agent that takes a question and outputs a node based workflow

That rings useful to me. I find it hard to trust an AI black box to output a good result, especially chained in a sequence of blocks. They may accumulate error.

But AIs are great recommender systems. If it can output a sequence of blocks that are fully deterministic, I can run the sequence once, see it outputs a good result and trust it to output a good result in the future given I more or less understand what each individual box does. There may still be edge cases, and maybe the AI can also suggest when the workflow breaks, but at least I know it outputs the same result given the same input.

> It should obviously use the webview facility provided by the default browser

Android doesn't do that either. You get Android System WebView which is always WebKit/Blink.

Windows didn't provide any browser engines apart from Trident through mshtml.dll either.

It would make iOS development harder - instead of getting a web view working just for WKWebView, you would need to get it working for every WebView out there. Given that they are not full browsers, it's not as simple as following web standards/caniuse. For example, service worker support in in-app browsers is limited.

If this actually gets implemented, every app out there will most likely pin the in-app browser engine to WKWebView OR bundle a binary blob with their own browser engine directly into .ipa. Bundled browser engine in .ipa has privacy implications - the app will be able to fully read secure HTTPS cookies of the sites that user visits - something that's currently protected.

Great idea!

I'm testing it on a 3-layer perceptron, so memory is less of an issue, but __slots__ seems to speed up the training time by 5%! Pushed the implementation to a branch: https://github.com/noway/yagrad/blob/slots/train.py

Unfortunately it extends the line count past 100 lines, so I'll keep it separate from `main`.

I have my email address on my website (which is in my bio) - don't hesitate to reach out. Cheers!

Here complex numbers are used for an eloquent gradient calculation - you can express all sorts of operations through just the 3 functions: `exp`, `log` and `add` defined over complex plane. Simplifies the code!

The added benefit is that all the variables become complex. As long as your loss is real-valued you should be able to backprop through your net and update the parameters.

PyTorch docs mention that complex variables may be used "in audio and other fields": https://pytorch.org/docs/stable/notes/autograd.html#how-is-w...

> They simply compare the prompting strategies that work best with each model

Incorrect.

# Gemini marketing website, MMLU

- Gemini Ultra 90.0% with CoT@32*

- GPT-4 86.4% with 5-shot* (reported)

# gemini_1_report.pdf, MMLU

- Gemini Ultra 90.0% with CoT@32*

- Gemini Ultra 83.7% with 5-shot

- GPT-4 87.29% with CoT@32 (via API*)

- GPT-4 86.4% with 5-shot (reported)

Gemini marketing website compared best Gemini Ultra prompting strategy with a worse-performing (5-shot) GPT-4 prompting strategy.

What if your browser receives Alt-Svc header and switches to http/3 on one network (say mobile data), but then you switch to a restrictive WiFi that has UDP disabled. All without restarting your browser/within one "session" of your http client. Wouldn't you start having connectivity issues that would be hard to troubleshoot? In that scenario, having http3 disabled is beneficial.

Wondering the same- how’s the compatibility with very restrictive networks? Say a university/corporate network that only allows port 443 and DPI’d 80 egress. QUIC won’t pass through that since it’s UDP. Can server and client reliably negotiate a fallback to http/2 in such a case?

I think this is a brilliant solution - you can also make it the second row in the table that's being sorted (right after the table header). That way there's no breakage and it's a good enough deterrent.

I would go on further and make a link to a Stripe Checkout page that allows you to purchase a license subscription to remove the notice.

Another solution: when the script is hotlinked, add a new row in the sorted table, saying "You're using a hotlinked version of sorttable.js - please host on your own webserver or purchase a license here". This way, the sorttable.js is not broken for any users, it's still usable, but there's an "ad" that's both a deterrent and an upsell.

This is incredible. Sounds like a really good AI code review technique!

Given that they're transmitting an app form over to Apple's server, they could encode checkboxes with bitmaps and text fields with dictionary LZMA. There probably will be only a few bytes of the actual data per emergency submission + protocol authentication and error correction overhead.

I've got Snapchat+ and I've logged into my account, but it still says it's unavailable. Sad, must be a bug.

FedNow will allow instant bank transfers. That means that you can build a payment system on top of it similar to Windcave Account2Account (that we have here in Australia and New Zealand)

Windcave Account2Account is not a full replacement for debit/credit cards. It does not have PayWave (solvable). It has a clunky UX (solvable). It does not do credit transactions (not solvable).

It can be a minor headwind to VISA/Mastercard and slightly reduce their new signups and tx volume, but I cannot see it fully replacing credit/debit cards.

Yeah it's the trendline's slope, but it might be a bit hard to compare them since the angles are not that dissimilar.

For the sake of being a pedant, your statement is incorrect: the U.S. federal fund rate was 15% in 1981, making CDs and T-bills yield 15% APY for the products bought that year. So that kind of return has a place in reality. Granted, those were different times, different economic climate and exceptional fed actions. What you probably mean is that promising anything that exceeds SOFR + say 2 percentage points has a risk premium that should be disclosed.

I think the main thread is on the left with the reply/qt sub-trees on the right. You can only read the first column. I quite like the layout, if there's only 1 main thread that i need to be aware of it's quite nice. This layout probably won't work if there's more than 2 people in a conversation.

The layout is actually not that much dissimilar from what twitter already has, just more info from non-main threads and better visibility for orphan leafs.

As an owner of http://git.io/way and https://git.io/no it's a good thing I noticed this on hackernews front page. Otherwise I would've never known! Just changed all of my git.io links where I could find them. I wish they would send out emails for this, but as I remember git.io is an anonymous service.

Quite odd indeed, maybe there are some legal reasons not to mention it. If they could, I feel the whole article could be condensed to "we just shipped a LetsEncrypt alternative gated to Google Cloud customers".

I minimise them quite frequently these days for something I want to tuck away. A good way to surface them is the "App Exposé" feature (Preferences -> Trackpad -> More Gestures). Shows all the hidden windows for the current app.

It's different to Windows though. If they're looking for something more similar to Windows, maybe Command+H (hide the active app) is a better alternative.

You don't need an app for window splitting. If you hold Option ⌥ key and hover over the green circle, there will be "Move Window to Left side of the screen" and "Move Window to Right side of the screen" options presented instead of "Tile Window to Left of Screen" and "Tile Window to Right of Screen" options. Those don't create a new virtual desktop.