Live data from Hacker News

Scaling long-running autonomous coding

cursor.com

11–20 of 203 posts

Re: Scaling long-running autonomous coding

#11
post #8

This is going to sound sarcastic, but I mean this fully: why haven't they merged that PR. The implied future here is _unreal cool_. Swarms of coding agents that can build anything, with little oversight. Long-running projects that converge on high-quality, complex projects. But the examples feel thin. Web browsers, Excel, and Windows 7 exist, and they specifically exist in the LLM's training sets. The closest to real…

Pretty much everything exists in the training sets. All non-research software is just a mishmash of various standard modules and algorithms.

Re: Scaling long-running autonomous coding

#12
post #9
post #3

"To test this system, we pointed it at an ambitious goal: building a web browser from scratch." I shared my LLM predictions last week, and one of them was that by 2029 "Someone will build a new browser using mainly AI-assisted coding and it won’t even be a surprise" https://simonwillison.net/2026/Jan/8/llm-predictions-for-202... and https://www.youtube.com/watch?v=lVDhQMiAbR8&t=3913s This project from Cursor is the s…

2029? I have no idea why you would think this is so far off. More like Q2 2026.

You're either overestimating the capabilities of current AI models or underestimating the complexity of building a web browser. There are tons of tiny edge cases and standards to comply with where implementing one standard will break 3 others if not done carefully. AI can't do that right now.

Re: Scaling long-running autonomous coding

#13
I would love to know the cost of building this browser. I think that multi-agent orchestration systems will probably be the theme for systems this year.

I think the north-star metric for a multi-agent orchestrator system would be how much did it cost to get this done. how much better could we have done? should we have used a cheaper model for doing a trivial task and an expensive one to monitor it?

Re: Scaling long-running autonomous coding

#14
Supposing agents and their organization improve, it seems like we’re approaching a point where the cost of a piece of software will be driven down to the cost of running the hardware, and the cost of the tokens required to replicate it.

The tokens were “expensive” from the minds of humans …

Re: Scaling long-running autonomous coding

#15
This looks like extremely brittle code to my eyes. Look at https://github.com/wilsonzlin/fastrender/blob/main/crates/fa...

What is `FrameState::render_placeholder`?

``` pub fn render_placeholder(&self, frame_id: FrameId) -> Result { let (width, height) = self.viewport_css; let len = (width as usize) .checked_mul(height as usize) .and_then(|px| px.checked_mul(4)) .ok_or_else(|| "viewport size overflow".to_string())?;

    if len > MAX_FRAME_BYTES {
      return Err(format!(
        "requested frame buffer too large: {width}x{height} => {len} bytes"
      ));
    }

    // Deterministic per-frame fill color to help catch cross-talk in tests/debugging.
    let id = frame_id.0;
    let url_hash = match self.navigation.as_ref() {
      Some(IframeNavigation::Url(url)) => Self::url_hash(url),
      Some(IframeNavigation::AboutBlank) => Self::url_hash("about:blank"),
      Some(IframeNavigation::Srcdoc { content_hash }) => {
        let folded = (*content_hash as u32) ^ ((*content_hash >> 32) as u32);
        Self::url_hash("about:srcdoc") ^ folded
      }
      None => 0,
    };
    let r = (id as u8) ^ (url_hash as u8);
    let g = ((id >> 8) as u8) ^ ((url_hash >> 8) as u8);
    let b = ((id >> 16) as u8) ^ ((url_hash >> 16) as u8);
    let a = 0xFF;

    let mut rgba8 = vec![0u8; len];
    for px in rgba8.chunks_exact_mut(4) {
      px[0] = r;
      px[1] = g;
      px[2] = b;
      px[3] = a;
    }

    Ok(FrameBuffer {
      width,
      height,
      rgba8,
    })
  }
} ```

What is it doing in these diffs?

https://github.com/wilsonzlin/fastrender/commit/f4a0974594e3...

I'd be really curious to see the amount of work/rework over time, and the token/time cost for each additional actual completed test case.

Re: Scaling long-running autonomous coding

#17
Did anyone manage to run the tests from the repository itself? The code seems filled with errors and warnings, as far as I can tell none of them because of the platform I'm on (Linux). I went and looked at the Action workflow history for some pages, and seems CI been failing for a while, PRs also all been failing CI but merged. How exactly was this verified to be something to be used as an successful example, or am I misunderstanding what point they are trying to make? They mention a screenshot, but they never actually mention if their goal was successfully met, do they?

I'm not sure the approach of "completely autonomous coding" is the right way to go. I feel like maybe we'll be able to use it more effectively if we think of them as something to be used by a human to accomplish some thing instead, lean into letting the human drive the thing instead, because quality spirals so quickly out of control.

Re: Scaling long-running autonomous coding

#18
post #9
post #3

"To test this system, we pointed it at an ambitious goal: building a web browser from scratch." I shared my LLM predictions last week, and one of them was that by 2029 "Someone will build a new browser using mainly AI-assisted coding and it won’t even be a surprise" https://simonwillison.net/2026/Jan/8/llm-predictions-for-202... and https://www.youtube.com/watch?v=lVDhQMiAbR8&t=3913s This project from Cursor is the s…

2029? I have no idea why you would think this is so far off. More like Q2 2026.

[flagged]

Re: Scaling long-running autonomous coding

#19
post #3

"To test this system, we pointed it at an ambitious goal: building a web browser from scratch." I shared my LLM predictions last week, and one of them was that by 2029 "Someone will build a new browser using mainly AI-assisted coding and it won’t even be a surprise" https://simonwillison.net/2026/Jan/8/llm-predictions-for-202... and https://www.youtube.com/watch?v=lVDhQMiAbR8&t=3913s This project from Cursor is the s…

Time to raise the bar. By 2029 someone will build a new browser using mainly AI-assisted coding and the surprise is that it was designed to be used by pelicans.

Re: Scaling long-running autonomous coding

#20
post #15

This looks like extremely brittle code to my eyes. Look at https://github.com/wilsonzlin/fastrender/blob/main/crates/fa... What is `FrameState::render_placeholder`? ``` pub fn render_placeholder(&self, frame_id: FrameId) -> Result { let (width, height) = self.viewport_css; let len = (width as usize) .checked_mul(height as usize) .and_then(|px| px.checked_mul(4)) .ok_or_else(|| "viewport size overflow".to_string())?;…

this is certainly an interesting way to pull out an attribute from a tag: https://github.com/wilsonzlin/fastrender/blob/main/crates/fa...
Post reply on HN