EdwardDiego

Very smart. But also a good example of why macros are brittle.

> I don't have to navigate around the codebase to trace data flow.

How big is your codebase?

No true Scotsman, eh?

What path do you propose?

> People worry that not being seen as mindless, uncritical AI cheerleaders will be a career-limiting move in the current environment of enforced conformity within tech

Yes. I am seeing this right now, but it's not just enough to join the chanting around the altar, you've got to be seen enthusiastically using AI as much as possible and it's turned the contributions of some of the most productive high level developers in my company into... ...well, slop.

I'm dealing with a bug from one such right now, where they wrote code that used my library, and it's not working for them, and they lack the understanding of their code to fix it, and the reason it's a bug I'm working on? Cursor couldn't fix it for them, and they lack the mental model to do so, because it's hard to mentally model code you didn't write.

I'm sure there's an inevitable "well, your company/colleague is doing it wrong then" critique incoming. And I agree.

But given that "doing it right" is often being defined as "using it as much as possible" by business leaders across the industry, then we get these paradoxical outcomes where doing so reduces productivity, but no-one is ready to admit that.

You've got to be AI-first, or AI-native, at least if you want the investors to stay invested.

I've never experienced such a collective adherence to tech hype (but probably only because my company couldn't figure out how to jam a blockchain and/or NFT in previously). Not even during the days when everyone was serverless, or "cloud-first" / "cloud-native".

It's wild.

TL;DR - we've always as an industry used appropriate tools for appropriate problems, right now it feels like so many of us are throwing AI slop at walls and seeing what sticks.

> I don’t think you understand how primitive American society is compared to Asia.

Lol, Asia is a big and diverse place. Are you really claiming that American society is more primitive than that of farmers in the arse end of Gansu?

Hint, one of those areas is more likely to have flush toilets.

"Huh, turns out we could replace it all with a 4 line Perl script doing linear regression."

> One team suggested that instead of text-to-SQL, we should build semantic business logic layers, “show me Q4 revenue” should map to a verified calculation, not raw SQL generation.

Okay, how would that work though? Verified by who and calculated by what?

I need deets.

I'm assuming you can identify by the heavy drinking to drown feelings of self-loathing.

> A slightly harder way (but one that gives you more control) is to make your pet idea available for an existing political campaign.

I have become very good at hitching my wagon to passing VP trains, it's sad, but it works.

Universal income you say?

My country will pay for your food if you can't afford any, yep. I'm glad to pay taxes to cover it, because extreme poverty isn't great for a society.

Remember kids, social security keeps society secure against poor people with nothing left to lose!

Just looking at the README, I'm very excited as I have a Hackathon coming up and I am looking to bring more type love to a legacy Django codebase, and was going to utilise MonkeyType again, but this time I'll use this!

I'll look into it, cheers!

Clay pigeon / skeet shooting is hard too, it's a competitive sport for a reason, but then, so is shooting a target with a rifle at 400 yards, but the armed forces train you up on that. So designated drone marksman might become an interesting new speciality in the modern army.

It was a joke.

Sure, and for those reasons of probability of a successful hit, hunters don't tend to shoot birds on the wing with rifles either. I mean it's doable, but it's a very hard shot to land with a MK-1 Eyeball.

But we've seen infantry effectively using shotguns in Ukraine as a defensive weapon against drones for the same reason hunters use shotguns to shoot birds on the wing, within your weapon's range the odds of a disabling hit are pretty high if you're trained.

So not sure why it has to be either a rifle bullet _or_ a massively overengineered defense contractor's very expensive super-duper-shotgun round.

Haha, wait, it's flak guns again, just redone for the 21st century. Ah bless.

If the choice is restricted by stand-off range, that is a different kettle of fish, maybe do bring back the, once considered entirely obsoleted by high altitude aircraft and missiles flak.

But in the here and now, regular shotguns are being used defensively, with a certain level of success, by infantry against drones in Ukraine.

Yep, I've used this pretty successfully, the ideal is to run it under realistic prod traffic over time to capture as many types as can flow into a given function, but a good set of unit/integration tests can also provide good coverage.

And if you re-use the same type store (SQLite DB) across multiple instrumented runs, you can further improve it.

https://github.com/Instagram/MonkeyType

A binturong looks like Paddington Bear's cousin who moved onto harder drugs than marmalade, and hasn't had a hit a in a few days, and is doing it hard.

https://upload.wikimedia.org/wikipedia/commons/a/a7/Binturon...

I suspect it was added by the LLM that wrote this MCP.

At the risk of giving too much value benefit of the doubt, maybe an LLM put that BCC in to debug failures and well, it's hard to PR code you have no mental model of.

> It was silly yet satisfying when Claude (the PR reviewer) agreed the change looked good and was ready to merge.

> "Approved - Ship It!” and ‘Great work on this!”

This pat on the head from an algorithm gives me the creeps, and I'm really struggling to put my finger on why.

Maybe it's because it's emulated approval, yet generating real feelings of pleasure in the author?

If you don't already, check out Admiral Cloudberg's write-ups. She started doing research for MentourPilot last year, but has a few years of previous articles that'll wiki-hole you real good.

> AI 100% does currently improve productivity

What definition of productivity are you using?

> That's why you tell claude code to write tests, and use them

I've seen Python unit tests emitted by LLM that, for a given class under test, start with.

    def test_foo_can_be_imported(self):
        try:
            from a.b.c import Foo
        except ImportError:
            self.fail()


    def test_foo_can_be_instantiated(self):
        from a.b.c import Foo
        instance = Foo()
        self.assertNotNull(instance)
        self.assertTrue(isinstance(instance, Foo)

   def test_other_stuff_that_relies_on_importing_and_instantiating_foo(self)
        ...

And I've watched Cursor do multiple rounds of

"1: The tests failed! I better change the code. 2: The tests failed! I better change the tests. GOTO 1"

until it gets passing tests, sometimes by straight out deleting tests, or hardcoding values to make them pass.

So I don't have the same faith in LLM-authored tests as you do.

> I feel like hallucinations have become a cop out, an excuse, for people who don't want to learn how to use these new tools anyway.

I feel like you've taken that attitude so you can dismiss concerns you don't agree with, without having to engage with them. It's disappointing.

I have spent plenty, rest assured.

Thanks for the insightful reply that showed me where I went astray.

Except humans learn from your PR comments and in other interactions with more experienced people, and so inexperienced devs become experienced devs eventually. LLMs are not so trainable.

Also used to do work for Qatar, hence why it's not corrupt at all for Trump to accept a plane from Qatar.

Yeah, I was laughing in Kiwi at this. When you live in a country where you can sunburn in less than 10 minutes in spring, skin cancer is a far higher risk.

True, they've gone all in on Flink, which can also do what you want. And I suspect SparkSQL can do it also these days, but I haven't looked at that for ages.

Good luck!