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

Astra for Coding: Why Are We Doing This Again?

lucumr.pocoo.org

Post reply on HN
Print view View original

261–270 of 333 posts

Re: Astra for Coding: Why Are We Doing This Again?

#261

Post by d5lt5 »

weird-eye-issue wrote: post #230
d5lt5 wrote: post #121

Earlier quoted context omitted.

Why do you think your agents prefer to create scripts instead of doing tool calls these days? I wonder, is it easier to modify a script that agent wrote before to satisfy your prompt, or is it easier to write a new one from scratch each time a retry happens? Are input tokens more expensive than output tokens?

My god. They are not reusing the scripts. They are adhoc, inline Python scripts just used to make a single edit. You seem to fundamentally not understand what everyone else is talking about

They do reuse scripts, though. Maybe it's you who is too lazy to _comprehend_ the output?

Or, maybe your prompts are not good enough. And it's not my problem to fix, as you claim you are very experienced.

Re: Astra for Coding: Why Are We Doing This Again?

#262

Post by chickensong »

lelanthran wrote: post #222
nvch wrote: post #101

Earlier quoted context omitted.

The harness instructs them to behave this way. Also this approach saves tokens. The scripts allow to edit files in bulk, and most of the session cost is in cache reads (e.g. for 300K context each command costs the same as 30K input tokens).

> The harness instructs them to behave this way. Also this approach saves tokens. The scripts allow to edit files in bulk, and most of the session cost is in cache reads (e.g. for 300K context each command costs the same as 30K input tokens). I understand the reasoning, but at that point wouldn't the LLM be better off creating `sed` commands and executing those? I mean, if it's already executing Python, it can litera…

I've experimented quite a bit with giving agents python vs sed + awk. They make mistakes with both, a lot. The only thing that has stood out is that agents reach for python too quickly if it's available, and that awk causes the least problems, while sed might take several attempts to get results, similar to python.

Re: Astra for Coding: Why Are We Doing This Again?

#263

Post by GTP »

taurath wrote: post #120

When the code is shitty it becomes harder and harder for the models to make changes and this grinds progress down to a halt - this has been my experience with “factories” trying them and doing refining steps every few months. I sincerely don’t understand what the people who say they no longer read any code are doing, because it must be somewhat trivial to not run headlong into these issues that stack up time after ti…

I also struggle to reconcile these things, I guess the only way would be to try and see if it works for me, without caring too much if it works for other people on the internet. I would just like to add that, DwarfStar's author (Salvatore Sanfilippo) is a strong supporter of the idea that you shouldn't read the code, and he says he never read DwarfStar's code. And still, it seems that this project is much more than PoC and actually both usable and useful for people doing local inference (I didn't try it myself, but I saw a lot of positive comments about it). Could be that the crucial point is in how we use those models: instead of giving it a general goal (e.g. build me an inference engine) Sanfilippo, being an experienced programmer, kept pointing the models in the right direction. He also read the papers related to the models he was programming support for in DwarfStar, so that, when he worked on optimizations, he knew what should be done instead of prompting a general "please optimize this". So, I would say that, if you let the agents work on a "feature by feature" basis instead of trying to on3-shot things, you get much better results. Could also be that, by attempting to one-shot large projects, the model starts coding badly due to context window exhaustion.

Sorry for the not so well written comment, I was just throwing in some ideas.

Re: Astra for Coding: Why Are We Doing This Again?

#264

Post by Dlemlo »

xyzsparetimexyz wrote: post #153
Dlemlo wrote: post #11

"But for how much more Fable costs, for how much more Astra costs, I do not feel like the results are there." we are in the middle of the beginning. Its just a weird take to talk about the newest model like this while we are still in a R&D phase. And these points don't matter if you let it search and analyse a bug, for example, or if you have good harness and a good architecture and let it do small PRs or if you do s…

Nobody seriously thinks that AI is still at a R&D phase. It's already heavily entrenched both in companies and the financial world. If it's getting worse for coding then thats a major problem

With this progress, every few month there is a new R&D phase because you need to adjust to the new way of interacting with them.

We also still haven't build everything we expect to happen. Like a proper opensource agent platform, agentic layer etc.

Every week there are new research results from frontierlabs.

Re: Astra for Coding: Why Are We Doing This Again?

#265

Post by neomantra »

On disposable code, I’m waiting on an Adafruit Feather microcontroller to come in the mail. I asked Astra to make a me web-based Feather simulator, kinda like the iOS simulator with screen and buttons, so I could work on my UX while I waited.

Something like that would have been a multi-month project a year ago, but I did it in twenty minutes rather than pay for expedited shipping.

Re: Astra for Coding: Why Are We Doing This Again?

#266

Post by WorldIQ »

Marazan wrote: post #30

I love this dance we are doing where when people write the "AI models are garbage machines that produce garbage and are no where close to the fantasy being pedalled by the Crypto bros who pivoted to AI" it always has to be caveated with "AI models are useful and I am highly productive with them" It feels like people should just be able to say "This article comes with the standard disclaimer" and just dive into the me…

The author first had to proclaim his superiority as a non-American - a non Westerner entirely!

”Your whole half of the world is stupid, here’s an unrelated Chinese word” lol love it

Ok, now we know this guy’s got some real culture and insight!

We are not dealing with some Westerner here who only works on 3D game slop.

He makes software factories!

Well he would if the AI code wasn’t so shitty! >:(

Re: Astra for Coding: Why Are We Doing This Again?

#267

Post by croon »

CBLT wrote: post #166
RamblingCTO wrote: post #94

Earlier quoted context omitted.

For me it also produces totally overengineered tests that are tightly coupled to the implementation. For example testing existence of css classes (in a template based go prooject ...) instead of behaviour.

Can you recommend any model that doesn't do this?

Not GP, but IME it's not fixable by model selection, but being zealous about guiding output and vision, and pushing back on all the bad habits LLM in general has (eg verbose output as a band-aid for emergent intelligence). As soon as something is introduced into your codebase, it will continue being picked up into context until you remove it and any reference to it from any potential context entrypoint. If you don't any model will keep venturing down wrong/bad paths.

Re: Astra for Coding: Why Are We Doing This Again?

#268

Post by Fredkin »

Frontier models have seen more Mathematica and Powershell code than I ever have in their training, yet they really struggle to produce working output. They seem heavily tuned to Linux too. Despite adding skills to rectify this, they still fail to realize they're running on Windows and waste tokens. A human with this much training wouldn't have this problem. There are evidently still some pretty big holes still.

Re: Astra for Coding: Why Are We Doing This Again?

#270

Post by jsenn »

Fredkin wrote: post #268

Frontier models have seen more Mathematica and Powershell code than I ever have in their training, yet they really struggle to produce working output. They seem heavily tuned to Linux too. Despite adding skills to rectify this, they still fail to realize they're running on Windows and waste tokens. A human with this much training wouldn't have this problem. There are evidently still some pretty big holes still.

This is probably a harness problem rather than a model problem. GitHub Copilot will happily and effectively use Powershell while Claude Code struggles in my experience.
Post reply on HN