Used to be in that eternal battle but a squirrel baffle solved it.

I don’t doubt it’s possible to get better consistency but the juice is really not worth the squeeze for me. You end up churning through huge expensive models, orchestrating sub agents, writing out boilerplate hand-holding instructions (“please don’t break this, stop trying to commit to main, please lint ffs…”).

I don’t use it for Java but that would make sense with rigid enterprise patterns and VeryVerboseNamesThatAreEasierForAModelThanAHumanFactoryClazz {...

I don’t think our career is boned, moreso that all juniors trying to get in are boned. Everyone who knows what going on transition to a more hands-off architect role.

But like I said, our tokens are heavily subsidized right now. When they pull the rug, code monkey jobs will start to get listed again (with lower salaries of course).

Things I’ve realized while working with AI (Claude code):

• It’s fantastic for very small macros and medium length scripts. Think dev ops stuff, pre-commit hooks, transforming data. Keep it small enough to manually review and something you can run without destroying anything important. This can massively boost your codebase QoL. [Double bonus for not wasting tokens to solve the same problem over and over]
• It’s decent-to-good at debugging but not consistent with fixes. It can find some utf encoding edge case that might have taken you 1hr+ but suggest the dumbest bandaid fix you’ve ever seen. Also very good at spinning up unit test suites for basic edge cases.
• Due to obvious training bias, it’s pretty good with common libraries and cloud platform infrastructure. It could probably help with writing a complex cron call, debugging regex or fixing an IaC config. On the flip side it won’t bother to use the latest package version or know your niche/new library.
• It does better with greenfield because exploring your codebase introduces a ton of bias. It might try to fit in an ugly hack when a refactor to simplify everything is way easier.
• It’s absolutely garbage with UI, just throws the most disorganized HTML together that isn’t reactive or reusable. OK enough for ugly internal stuff but God help anyone relying on it for that.
• This is setting up to be the biggest rug pull in history. People that buy into it heavily just to save a couple bucks on engineer payroll are going to be fucked when they start ratcheting up the token price.

All in all it can be useful when used with care but will never be a magic bullet.

Alt: Copilot after dropping your production database

To save a click: May 20th

Ask it ten times to make list of 20-30 random words

This is true on ootb models but not the universal rule. You could adjust the temperature all the way up and get something way more random, probably to the point of incoherence.

The trick is balancing that with keeping the model doing something useful. If you’re clever you could leverage /dev/random or similar as a tool to manually inject randomness while keeping the result deterministic.

…a single instance of a Uyghur being killed

Genocide is more than just killing, it’s the deliberate destruction of a people including its culture and institutions. In fact, the reason you’re so focused on people dying is because imperialist powers felt the need to redefine it to allow their exploitation. Even in its more narrow definition it still includes things like abducting and re-education of children and malicious targeted actions (forced labor, restricted reproduction, relocation, etc…).

China is a massive country, and has always had issues with maintaining control over its more distant and ethnically distinct pockets. This stretches back centuries right up to now. For example: 100+ years of separatism, uprisings and violent incidents in East Turkestan. Or more recently, 160 Tibetans self immolating in protest of government repression since 2009. In this lens, there’s plenty of evidence that could support these accusations.

Meanwhile anyone can visit Xinjiang

This sounds like something people parrot only if they haven’t actually traveled China. For one, Xinjiang is massive. It’s about the size of Iran; 1/6 of China by land area. Saying you can visit it is like saying you could visit somewhere in France + Spain + Germany + Italy.

For two, foreigners require a special permit to visit ~12-15% of China (varying by year). This includes the expected restricted zones (military or government areas), but also the areas along borders (many in Tibet and Xinjiang) and “politically sensitive” areas. There’s no official list published for obvious reasons but they’ll certainly let you know if you’re not welcome.

As a disclaimer for those of you furiously typing whattaboutgaza: Yes that’s a genocide. Yes, many countries have engaged in similar kinds of repression. Yes media will amplify stories that paint rivals in a bad light, no that’s not unique to western media.

That’s kind of a funny example because, on a quick skim, nothing he did was exceptionally clever or unusual (other than workarounds for not having source code). R* basically paid him 10k for some basic profiling that they never bothered to do.

I agree, real code always has tradeoffs. But there’s a difference between a conceptually simple change taking 3 weeks longer than planned and 6 months. The reality is game code is almost always junk and devs have no incentive to do better.

Getting a feature functional and out for launch day is the priority because you don’t have any cash flow until then. This has been exacerbated with digital distribution encouraging a ship-now-fix-later mentality.

This means game devs don’t generally have experience with large scale, living codebases. Code quality and stability doesn’t bring in any money, customer retention is irrelevant unless you’re making an mmo.

When a dev with game dev experience says something should be easy to fix, it’s under the assumption of a reasonable code base. Most games are built off of common engines and you can sometimes infer how things are likely organized if you track how bugs are introduced, how objects interact, how things are loaded, etc…

When something is a 1 day bugfix under ideal conditions, saying it will take 6+ months is admitting one of:

• The codebase is fucked
• All resources are going to new features
• Something external is slowing it down (palworld lawsuit, company sale, C-suite politics, etc…)
• Your current dev team is sub par

Not that any of those is completely undefendable or pure malpractice, but saying it “can’t” be done or blaming complexity is often a cop out.