I've already commented on other posts that having LLMs build deterministic and testable tools is the real unlock.

Even for things like customer service, a LLM that analyzes customer support transcripts and then updates your call tree to better route people is a huge win.

Thanks for the suggestions.

Yeah, some other folks have pointed this out so need to tweak the win conditions/balances.

> But this really points to a deeper problem: in a game like this you have to pick one of two philosophies.

You really hit the nail right on the head here. I haven't built many games before but this has taught me that it's really a rock/paper/scissor balance that is tough to get right.

> I do however absolutely love the vibe of the game. I think with a little more attention it could be absolutely amazing.

THANK YOU! I was totally going for a Soviet vibe and glad people like that part.

Realized after I posted that it's probably really hard to actually win.

> I've played it 100 million times in a monte carlo simulation and even applied two heuristics to hit a winning strategy:

This is the MOST HackerNews comment ever (in the good sense) and it makes me immensely happy that this is the kind of feedback I received.

Btw, going to post a new version with some tweaks to the ruble payout etc.

I did seriously consider if that is really the best example of the Soviet economy ;-)

Summary:

- first publisher said: you can't publish this b/c you could go to prison and we could get in trouble financially

- second publisher: you are going to publish this. If you go to prison, we will support your family financially for the entire time you are in prison. If you lose your house, we will buy them a new house.

You can read about it here: https://tim.blog/2021/06/30/michael-pollan-this-is-your-mind...

People talk a LOT about defending free speech but this story has always stuck out to me as what that really means.

Also, I've been trying to make a remix of SMP but with a princess theme for my daughters. Been experimenting with making other games to practice building out game mechanics and stumbled into making a game where you manage a Soviet Tractor Factory [0].

I'm interested to see if developing parts of Game A lead to Game B which in turn lead you back to interesting ideas for Game A (or maybe better Game C).

0 - https://alexpotato.com/games/tractor47/?l=hn

Comments URL: https://news.ycombinator.com/item?id=48503530

Points: 3

# Comments: 6

Some quick stories about databases in general and type hints in particular:

1. Some of the joins involved 10+ tables AND had query hints

That sounds cool until you realize that the queries were designed for the planner from 3 versions ago

2. Was in multiple conversations with head of the DBA team and application owners that went something like this:

Developer: "This query hasn't changed in YEARS! Why is it not performing well now??"

Me: "Have you considered that maybe the data cardinality has changed?" and then I or the lead DBA would explain how database indices worked.

I mention the above only as a warning to folks newer to DBA land that some of these features can seem great but can also be a crutch. What should have happened is that the schema, indices or application code should have been reviewed by the app teams (with the DBAs) periodically to ensure assumptions were correct.

(This was several years ago so I'm curious what kind of impact the LLMs have had on this)

I work in DevOps at a firm that has been very enthusiastic about using LLMs (in the good sense).

The phases were basically:

- try out having the LLM do "a lot"

- now even more

- now run multiple agents

- back to single agents but have the agents build tools

- tools that are deterministic AND usable by both the humans (EDIT: and the LLMs)

The reasons:

1. Deterministic tools (for both deployments and testing) get you a binary answer and it's repeatable

2. In the event of an outage, you can always fall back to the tool that a human can run

3. It's faster. A quick script can run in <30 seconds but "confabulating" always seemed to take 2-3 minutes.

Really, we are back to this article: https://spawn-queue.acm.org/doi/10.1145/3194653.3197520 aka "make a list of tasks, write scripts for each task, combine the scripts into functions, functions become a system"

-- END of original post --

What I would add:

if you let LLMs do whatever they want, they will happily make code. You can add tests to confirm that the tests work (which you used to do with human code, right?). You can also read the code.

When you read the code, you'll find that they sometimes do totally bananas things that still produce working code (I've seen humans do this too but that's another story).

In other words, you still need to make sure the system being built makes sense.

More succinctly:

Coding may be dead but software engineering is alive and kicking.

0 - https://www.imdb.com/title/tt4819510/

Someone pointed out that Malcolm Gladwell even does "oh, wait, I almost forgot" or "That just reminded me" during his speeches but that it rehearsed and a consistent part of the speech as well.

Similar to how comedians have a "set" that they first put together, polish and then repeat over and over again.

e.g. a friend of mine once met William Shatner and then ran into him again a few months later. When asked "How are you doing?" Shatner answered exactly the same way at both the first and second meeting. I imagine some of this is efficiency since famous people tend to get the same questions over and over again. Tom Wilson even has a business card that answers a lot of these questions [0]

What was more surprising was seeing this in high school. I did a summer program with kids from all over the US. A few months later, I saw one of them at a sports event and, similar to Shatner, he had a canned response. He was from a well to do family and was probably on some kind of "track" to the right college etc. Was still surprising to hear.

If you are curious to see someone busting the cache, there are video compilations of Sean Evans from hot ones asking questions of guests based on deep research and them being incredibly impressed. [1]

Charisma on Command also has a great video on how to ask better question [2]

0 - https://www.upworthy.com/back-to-the-future-actor-has-a-hila...

1 - https://www.youtube.com/watch?v=Endmr-93KOY

2 - https://www.youtube.com/watch?v=cHyYlFCaXPM

e.g.

- DotCom boom was letting companies IPO even if they had no revenue

- Great Recession was due to loosening credit restrictions for mortgages e.g. giving people NINJA (no income, no job) loans

so very curious to see how this plays out.

Case in point:

In the book Angry White Pyjamas [0], the author is British and living in Japan.

He is going through the Tokyo Riot Police training which involves a lot of aikido training. He is also teaching English to high school students.

He points out that the techniques used for training aikido worked well with the students as well.

Specifically:

- show the technique

- have someone try out the technique

- talk about what they did well and what they didn't do well

- have everyone else practice

Highly recommend the book btw if you are interested at all in Japan, martial arts, living abroad etc.

0 - https://amzn.to/4v3rHdq

The thinking is that evolution created error correction for the critical proteins to account for mutations.

Fascinating stuff.

Paul Graham doesn't think so

https://paulgraham.com/brandage.html

They used to charge you by the minute but now it's easier to just charge you a flat charge per month.