Thanks for flagging, summary is:
Charts are from a seeded random draw.
And the instinct is pretty spot on players state ~74% average confidence but hit ~54%.
Accuracy is basically flat across confidence brackets.
Up only strategy quietly beats the coin flipping monkeys (stocks generally go up and to the right) but players call down 43% of the time.
Companies will keep doing what they're doing now - they'll just use another verb than "buy" or "purchase". Which would be a small improvement, I guess...
Not an RPG and no quests but it has some of the most well-done game AI in any game I have ever played. Basically every mob in the game has its own goals and goes about fulfilling them regardless of what the player is doing, and where the player is etc. It's more like a simulated environment and the game drops you in the middle of it and you have to learn how the world works, navigate and survive it.
It's not even any kind of advanced technology, just how fun game AI can be if a developer gives it sufficient attention, instead of basically bodging together some behaviour trees and calling it a day while spending all the compute and development budget for graphics. That's one reason why game AI sucks so much most of the time.
Another reason is that players themselves don't want convincing enemies in a convincing environment. I always recall a review of Rain World where the game critic threw his controller because he thought the enemies respawn randomly when the player dies, thus depriving the game critic from the opportunity to memorise their "patterns". In truth, the enemies don't "respawn" they just go about their business while the player is regenerating and so they're not always where you last left them when you return at the spot you died. The world keeps turning when you're gone. So you do have to learn their "patterns" except those patterns are not trivial patrol cycles but actual, you know, behaviours.
But, no. That won't do. Give us a game where we can memorise every telegraphed attack so we know when to press which button with millisecond precision like a mindless automaton.
Sorry, little rant there. I'm saying that many gamers and many devs don't actually want decent, convincing game AI.
Very sad to see this (and cancelled my subscription). My home-grown web-based orchestrator will now have to rely on just codex and Gemini. One of the reasons that I was wary of turning it into a product.
Eventually Google and OpenAI will probably follow Anthropic, but they’re not quite as capacity-constrained.
I'm sorry, that's absolutely bullshit. In fact, I wish we had left everyone who complained behind—the python community would have been happier and healthier for it. Absolute crybabies who wanted to be catered to without caring for how intractable the problems with python2 were—e.g. dealing with unicode was a royal pain in the ass, and the bytes/string divide completely fixed it. IMO, it was the best-executed breaking change I've ever witnessed in a language.
In comparison, e.g. Scala 2 -> Scala 3 was an absolute nightmare—it just didn't have the same vocal wailing from maintainers in the community (or, I suppose, a fraction of Python's popularity to begin with).
Being to aggressive in breaking stuff gets you a shitshow like Node.js or Ruby. Long-term source code compatibility is a very useful feature for open source and a sign of a mature eco system. Feel free to add stuff, but once it's part of a stable release it has to be maintained long after a "better" way to do it comes along.
Javascript would heavily benefit from breaking changes. The reason why it still sucks ass to use today is because this won't ever happen.
I can't speak for node.js specifically but who gives a shit
> Long-term source code compatibility is a very useful feature for open source
Sure, until you need affordable maintainers. Maintainability must be balanced with patience for bad software. Cf the insane maintenance cost of perl scripts
nodejs itself doesn't have very many breakages; i have plenty of code that is unchanged from 0.12 to 24. npm is a whole other kettle of fish but I don't think you can blame the core project for the sins of everyone that publishes to the package manager. Python2 -> Python3 on the other hand had a lot of breakage in "standard" code.
Sure, but it's pretty trivial to generate a CLI application that talks to that API.
That's how I let agents access my database too. Letting them access psql is a recipe for disaster, but a CLI executable that contains the credentials, and provides access to a number of predefined queries and commands? That's pretty convenient.
Yes. But are you letting your agent make the decision of when and how to call that CLI? And presumably you're invoking it via the Bash tool. In which case your agent is free to write ad-hoc bash orchestration around your CLI calls. And perhaps you don't just have one such CLI but rather N for N different services.
And so we've arrived at the world of ad-hoc on-the-fly bash scripting that teams writing backend agentic applications in more "traditional"/conservative companies are not going to want.
Don't get me wrong, it's great for claude-code-type local computer automation use cases -- I do the same as you there.
