Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Welcome to the Wasteland: A Thousand Gas Towns

steve-yegge.medium.com

Post reply on HN
Print view View original

91–100 of 102 posts

Re: Welcome to the Wasteland: A Thousand Gas Towns

#91

Post by planckscnst »

I mostly use LLMs in a zero-touch way - I never actually edit code and I almost never read it. But I do still dive into the details by exploring it with targeted questions through the LLM. Sometimes I go through ridiculously long sessions to get the LLM to "see" the correct/optimal/simplest/etc solution itself. There are many times when it simply never gets there no matter how close I get the horse to the water. I recently did one of these sessions yesterday and it reinforced my impression that systems like gastown and pure Ralph loop style is just not ever going to have the quality I'm looking for, and it's going to cost a lot of money not to get there.

I've honed a relatively decent flow that requires interaction from me for important parts (mostly) while making its own decisions at the not-important parts (mostly). This results in being able to send the agent off on an hours-long dev cycle and have relatively decent results after that need a few minor fixes. I think this is the best style for the current generation of AI

Re: Welcome to the Wasteland: A Thousand Gas Towns

#92

Post by g-b-r »

DragonStrength wrote: post #78
senordevnyc wrote: post #66

It's saddening to see folks here on HN direct so much contempt towards this. I have no intention of messing with Gas Town or Wasteland, but I think it's cool that folks like Steve with lots of money and time are using it to build stuff they find cool and interesting. I doubt these projects are the future of AI agent orchestration, but I do think they're probably going to help us collectively learn better how to work…

I see a lot of folks lamenting how Yegge has ignited a fire under leadership for egoless, factory worker style engineers. I don't think many would care much otherwise, but his post here is lamenting AI is not enough of a factory worker yet. On a site mostly populated by folks who put most of their lives into becoming skilled professionals, hearing "we think your work should be the kind of work we send to the cheapest…

> It's just incredible how much impact he has had with this little hustle, given his products are basically turtles all the way down.

What impact did he have?

Re: Welcome to the Wasteland: A Thousand Gas Towns

#93

Post by g-b-r »

hardwaregeek wrote: post #6

I really want to host a vibe coding competition and see what can actually be made with these systems. Like if we’re doing insane token spends, it better be in service of creating amazing stuff. Can we make an entirely new programming language? Can we make an OS?

LLMs have a chance to make anything they've been trained on, by outputting it verbatim (or with small modifications).

The question is the degree to which they can produce original things.

Re: Welcome to the Wasteland: A Thousand Gas Towns

#94

Post by mieubrisse »

cadamsdotcom wrote: post #85

There are two bottlenecks in software. First is understanding what to build. Second is getting the details right. The best software is both the right software, and high quality. It’s bloody hard to do! So, we do requirements gathering, and try to get the details right (TDD, continuous delivery etc).. it’s surprising to see Gas Town do none of this and optimistically hope agents will converge on good software by just…

I'm working on this! https://github.com/mieubrisse/agenc

I think Gastown is truly special, but I wanted something more focused on learning as I think that's the real bottleneck. So I built AgenC to make it trivial to roll learnings back into your Claude.

Re: Welcome to the Wasteland: A Thousand Gas Towns

#95

Post by mieubrisse »

hardwaregeek wrote: post #6

I really want to host a vibe coding competition and see what can actually be made with these systems. Like if we’re doing insane token spends, it better be in service of creating amazing stuff. Can we make an entirely new programming language? Can we make an OS?

The guy to watch here is https://github.com/Dicklesworthstone . He's rewritten SQLite in Rust with fixes, written his own Rust async engine with fixes that Tokio doesn't have, generated an insane number of tools for agentic orchestration (indexing of all sessions across all harnesses, on-demand skill storage, agent mail), and is currently building out agent orchestration terminal multiplexer stuff.

Source: been watching both these guys closely, as I've been building my own agent factory focused on security + learning: https://github.com/mieubrisse/agenc

Re: Welcome to the Wasteland: A Thousand Gas Towns

#96

Post by egypturnash »

skybrian wrote: post #46
rektomatic wrote: post #37

Earlier quoted context omitted.

I legitametly thought it was Dolt until you pointed it out

The database actually is named "Dolt." It's their own take on "git." https://www.dolthub.com/

another winner from the school of naming that brought us "The GIMP", then

Re: Welcome to the Wasteland: A Thousand Gas Towns

#97

Post by egypturnash »

zachmu wrote: post #51
egypturnash wrote: post #35

"DoIt" looks like "Dolt" in all too many fonts.

lol it's DOLT, not DoIt. Yegge's Medium uses a serif font so you can tell, but in many faces you can't. (We still get this comment constantly and it's very unfortunate)

maybe you should consider a name change then

Re: Welcome to the Wasteland: A Thousand Gas Towns

#98

Post by senordevnyc »

DragonStrength wrote: post #86
senordevnyc wrote: post #66

It's saddening to see folks here on HN direct so much contempt towards this. I have no intention of messing with Gas Town or Wasteland, but I think it's cool that folks like Steve with lots of money and time are using it to build stuff they find cool and interesting. I doubt these projects are the future of AI agent orchestration, but I do think they're probably going to help us collectively learn better how to work…

If you can point to single product he has made through his "vibe-coding" that isn't for "vibe-coding," I think they would all relent.

Again, who cares?

Re: Welcome to the Wasteland: A Thousand Gas Towns

#99

Post by fc417fc802 »

npilk wrote: post #32

With LLM-based tools that inherently rely so much on the semantics of language, I wonder if there will be differences in code generated for the "wanted board" in the "Wasteland", compared to the "task list" in the "public square" or the "wish list" in the "Utopia".

Would be interesting to see what the outcome would be of assigning an "angel" or a "devout brother" to work on TempleOS.

Re: Welcome to the Wasteland: A Thousand Gas Towns

#100

Post by thinkling »

ufmace wrote: post #89
zachmu wrote: post #70

Earlier quoted context omitted.

Regardless of whether this particular project goes anywhere, it's at least very interesting that Yegge has discovered a way to make multi-agent setups work better. Giving them discrete personas ("you are a senior database engineer with 30 years of experience") and narrower scopes makes them much more effective. This was surprising to me but makes a lot of sense in retrospect.

The part that always struck me as weird about this stuff is that all of these "agents" with their "personas" are the same baseline LLMs with the same training ultimately, just told to basically pretend they're different. How far can that really get you? I'm not actually a database engineer with 30 years of experience. If somebody demanded that I pretend to be one, I guess I'd give it a shot, but I would expect any ac…

Agreed, would really like to understand what this (setting the LLM up to assume a role to improve performance) is doing under the cover and why it works.

Why aren't the labs training models to pick a mantra appropriate to the task and do this themselves? "Huh, a database question. I am going to pretend I'm a database expert with lots of experience. OK, here we go!"

Post reply on HN