The cheapest available model once you have Theory of Mind (the idea that the other things in the environment might be thinking like you do) is that they're you again.

The Smarties test (What's in this Smarties tube - look it's not Smarties, ok now what does somebody else think is in the tube?) shows that humans need a further step to discover that model isn't enough.

But it's still cheaper and it's pretty good. It will correctly predict that this person you've never met before probably wants cake not death, just like you. It won't reliably predict whether they prefer lemon cake or coffee cake. But it's a good first guess.

Also: some people seem to get an amount of pleasure from pointing out repeats, as if remembering that something was posted before is knowledge enough to make them a better person than the poster, us all, or just the person they thought they were themselves. This is fine when something is posted far too often, or is reposted by a point-farming bot (presumably the users running such bots hope to use the reputation of the account somehow in future), but is often done overzealously.

Personally, I don’t bother searching because I only consume the headlines, on other news sites too, come to think of it. There’s lots of interesting things people post but frankly I’d rather pay for a good book on any subject.

hides from the dreaded downvoters

I used to spend more time browsing when reading an actual newspaper or magazine. The discourse on opinion pieces and such is more thought out too—many people, myself included, post too quickly before thinking because we’re always on the go.

Something about the online experience consuming news is less satisfying. Perhaps a hacker out there can set up a print version of HN archives, and print it on a Gutenberg Press. :)

I'm not sure your complaint is actually that A* is bad, it's that the heuristic function is unfair (to the player, by giving the mob data they shouldn't have). A more sophisticated game could use a more interesting function like an estimate as to what direction the player's movement sound would be heard from.

This isn't an actual solution. Video Games have to do things fast. You can't just sit there and wait for a minute as you try brownian motion on hundreds of units. There are plenty of different solutions to get more realistic and natural pathfinding, typically navmeshes and then local node based movement. But you still need to go from a to b, somehow.

Your example also fails in several really obvious ways. What if there is a pool of mud or water in the middle of the path, it IS traversable but doing so is not ideal? A* you give this a high traversal cost and you'll naturally path around it. Your solution will have the object going through high cost but traversable areas in strange ways. This is going to be worse, as players will just kite and exploit this fact.

But humans do quite intelligently pathfind around objects they're aware of, and update their mental models with new information. The front door is locked? I'll go around the back.

You can achieve exactly what you're describing by hiding information entities do not have from their pathfinding.

Graphs aren't the problem, and thinking along those lines won't get you where you're trying to go.

> Brownian motion until you do or get tired

The point of movement for npcs in a videogame isn't to behave realistically (or to be simulated fully), it's to produce engaging and challenging behavior to the player. In 99% of cases giving characters, like enemies in an action game, some extra information to enable them moving towards the player is the correct choice, people are fine with suspending their disbelief if the alternative is npcs giving up or running around like brownian particles, which does not make for a good experience.

Almost all the time the correct choice in game design is that it's fun and interesting, unless for the exception where you're literally building a real world simulator.

> See the target/know which direction it is? Go that direction unless you see an obstacle, in that case go around the obstacle, eventually even backtracking if it turns out the obstacle was worse than you could see. Don't see/know the target? Brownian motion until you do or get tired.

What a long and convoluted way to try to reinvent the A* algorithm...

I encourage you to build a game with the mechanics you describe, especially something like an RTS, and see if it's any fun to play...

If you want cohesion between entities pathfinding, adjust the cost when you do the pathfinding for tiles that has friendlies on them to be lower than their base cost.

The way to think about water crossing with naval transports, is to consider those things to be conditions. You already have a set of condition when pathfinding. Just add another case for water. Make it so the requirement is that you’re either on a ship or there is a ship on the adjacent tile you checked previously, e.g N-1. If valid, set a flag and now every tile check that is water should be appropriate.

Open AI tried it with Dota 2 and had to limit the gameplay a lot to be competetive against humans.