Maybe the hawk had a sweet tooth?

Santa?

It’s new compared to FAT!

Honey, nut, cheerio.

If you scale it up you can probably send more than one right? Send ten and nine work. That’s not nothing.

Cursed.

Shadow & Bone (but what I really want is the six of crows spinoff).

WoT.

Dead Boy Detectives.

The PR isn’t public yet (it’s in my fork) but even once I submit it upstream I don’t think I’m ready to out my real identity on Lemmy just yet.

I just spent about a month using Claude 3.7 to write a new feature for a big OSS product. The change ended up being about 6k loc with about 14k of tests added to an existing codebase with an existing test framework for reference.

For context I’m a principal-level dev with ~15 years experience.

The key to making it work for me was treating it like a junior dev. That includes priming it (“accuracy is key here; we can’t swallow errors, we need to fail fast where anything could compromise it”) as well as making it explain itself, show architecture diagrams, and reason based on the results.

After every change there’s always a pass of “okay but you’re violating the layered architecture here; let’s refactor that; now tell me what the difference is between these two functions, and shouldn’t we just make the one call the other instead of duplicating? This class is doing too much, we need to decompose this interface.” I also started a new session, set its context with the code it just wrote, and had it tell me about assumptions the code base was making, and what failure modes existed. That turned out to be pretty helpful too.

In my own personal experience it was actually kinda fun. I’d say it made me about twice as productive.

I would not have said this a month ago. Up until this project, I only had stupid experiences with AI (Gemini, GPT).

If you’re in the US like me, we should be aware the problem isn’t bright lights; it’s that our regulations don’t allow for the European beam alteration tech that will dim sections at a time based on oncoming traffic.

Brighter lights are a huge boon to safety, but we need the corresponding tech to keep it that way.

I run massive, global kubernetes clusters in AWS for a company you’ve probably heard of. There is no queue of clean VMs–not like you’re thinking anyway. And provisioning a new node can take Too Long under not-all-that-uncommon scenarios.

The next best option is overprovisioning the cluster, but even 1% overhead has big costs at this scale.

For large scale compute clusters with elastic load I absolutely care. The difference between one and five minutes of boot time when I ask for a hundred new instances to be provisioned is huge in terms of responsiveness to customer requests.

If it’s domestic, there’s at least some recourse available. Facebook was fined $5 billion for the Cambridge Analytica scandal.

When they own the platform they can use it to serve you catered disinformation.

They can have your data but unless they can also decide what you see as a result, it’s not the same thing.

That’s the difference.

As the primary author of my previous org’s GHAs (not GH Enterprise, just the team tier) I found some feature gaps compared to org[n-2]'s Jenkins but they were fairly quickly filled.

I was initially skeptical but it wasn’t more than a month or two before I was just glad to be off Jenkins. And now that I’m back to a big org with a big Jenkins footprint, I really miss GHA.

Having everything be contextual in the same place is a huge value add for me.

But mechanically that’s just moving the confidence threshold to 100% which is not achievable as far as I can tell. It quickly reduces to “all objects are pedestrians” which halts traffic.

According to some cursory research (read: Google), obstacle avoidance uses ML to identify objects, and uses those identities to predict their behavior. That stage leaves room for the same unpredictability, doesn’t it? Say you only have 51% confidence that a “thing” is a pedestrian walking a bike, 49% that it’s a bike on the move. The former has right of way and the latter doesn’t. Or even 70/30. 90/10.

There’s some level where you have to set the confidence threshold to choose a course of action and you’ll be subject to some ML-derived unpredictability as confidence fluctuates around it… right?

There is a gulf between people who are paid well for their valuable labor (even into the millions of dollars) and the capital class who primarily profit on the labor of others.

Rent seeking is a big driver of “eat the rich”.

This still makes me mad.

My wife and I drive almost the same model of Audi, separated by a couple of years. One still has physical buttons for infotainment and one has a touch screen, but both support Android Auto and CarPlay.

I prefer the physical controls for it, because I can glance at the screen and know “turn right two clicks and press down” to get where I want, and then look back at the road while I do it.

If I had to make a wild guess as to why it’s designed that way? Cleaning flat buttons seems way easier than cleaning knobs. And no moving parts. Maybe more resilient (can be made with cheaper parts) considering the flimsy electronics that would be underneath the knobs compared to the more industrial (robust?) kind under an electric range.