News for hackers. Stuff that matters.
[dead]
Re:In 27,280 coding agent runs, 64% of inference spend went ... by radlad on Tuesday November 17, 2026 @10:49AM (#50533068)
Hi Guanlan, I vouched for your comment since you are the author, but you should be aware that your posts have been flagged, likely for AI patterns in your writing.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by latentsea on Tuesday November 17, 2026 @11:08AM (#50533426)
This is why for open-weights models at least harness aware post-training is becoming a thing. Also closed source harnesses that aren't fully editable are basically an anti-pattern. The labs should just hurry up and productionize self-evolving harnesses, and have the models internalize harness engineering, so they can optimise the harness for the given task. These are solvable problems.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by Alwayshasbeeb on Wednesday November 18, 2026 @05:14AM (#50544345)
This. I have coworkers who are really smug about claude code being "the best" option for AI assisted development and who refuse to try anything else. They also constantly submit pull requests littered with trash "trace of thinking" slop comments and when I point it out they say "yeah I usually try to make it not do that but this one got through". Meanwhile in pi this was trivial to fix and I'm often mining session history to refine the efficiency.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by Majromax on Tuesday November 17, 2026 @11:05AM (#50533354)
Saying that 'steps per pass' explains cost seems nearly tautological. Run costs are proportional to the number of tokens produced, and the number of tokens is strongly correlated to the number of steps.

The 'steps per pass' criterion is also not something under the harness's direct control in the way that its tool collection and system prompt (/start tokens) are. Even the cache hit rate is hard to control beyond a limit, since modern harnesses rarely rewrite history.

I'm also suspicious about the run-to-run variance between models. Given the high variability within the presented results and the observation that failed runs (which can spin for a long time before delivering a wrong answer) account for a majority of the cost, I'm not sure that many of the conclusions here would hold up under a second sample.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by Zigurd on Tuesday November 17, 2026 @11:12AM (#50533520)
I find myself pulled in two directions by the performance of coding agents: First, keeping them on a short leash seems sensible. I don't get failed results very often. When I first started using coding agents, this seemed not just sensible but required in order to not deal with a lot of hallucinated weirdness. Secondly though, it's pretty cool to vaguely wave your hands in a certain direction and get both user interface in business logic that's objectively nicer than I would've gotten from the effort I would've put into manually coding. Whenever I cautiously use a coding agent, I wonder what I am missing by not letting it cook.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by embedding-shape on Tuesday November 17, 2026 @11:15AM (#50533575)
Not every problem should (or even could sometimes) be solved the same way. Are you planning to maintain the produced code for more than a year? No, then wave away. But if you're expecting others to depend on it, and you want to be able to continue evolving what it is, then a tight leash and full understanding of how it actually works, is very much needed, even if you don't actually type characters into an editor anymore.

Otherwise you still end up with balls of spaghetti where each change takes longer and longer to work into the codebase. With a bit of planning, you can keep tackling changes and they keep taking the same amount of time more or less.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by smallnamespace on Tuesday November 17, 2026 @11:21AM (#50533689)
[delayed]
Re:In 27,280 coding agent runs, 64% of inference spend went ... by catdog on Tuesday November 17, 2026 @11:31AM (#50533864)
> The cost of AI is rapidly falling while its engineering capabilities keep rising.

Wouldn't count on it. There likely is a ceiling, we might or might not be near it.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by reddit_clone on Tuesday November 17, 2026 @12:48PM (#50535259)
Definitely with the cost. I strongly believe the price now is heavily subsidized and will shoot up in the future. (Not new. First hit is free..)
Re:In 27,280 coding agent runs, 64% of inference spend went ... by datsci_est_2015 on Tuesday November 17, 2026 @11:31AM (#50533882)
Robust systems experience very little turnover in nearly every non-software discipline that involves systems. Why the hell would we insert risk into our system-building processes by needlessly turning over code?
Re:In 27,280 coding agent runs, 64% of inference spend went ... by smallnamespace on Tuesday November 17, 2026 @08:44PM (#50540831)
[delayed]
Re:In 27,280 coding agent runs, 64% of inference spend went ... by Zigurd on Tuesday November 17, 2026 @11:34AM (#50533929)
> The requirements will be more stable than the code.

These kinds of changes to the way teams are formed and managed, and how products are developed is what really fascinates me about the impact of coding agents.

Even if it works, for some definition of works, to dump a huge spec into a coding agent context window, I suspect this is where a lot of horror stories about agentic coding come from.

Teams, and the management structure above them, are going to have to get used to more collaborative and iterative product definition and development in order to extract the most value from coding agents.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by wtch98 on Tuesday November 17, 2026 @11:54AM (#50534320)
> The cost of AI is rapidly falling while its engineering capabilities keep rising. This implies that code is becoming more fungible, and you should expect any code you write today to mostly be replaced in the near future.

I have code I wrote in 2006 which is still in daily use adding business value at my company. Why would it need replacing?

Re:In 27,280 coding agent runs, 64% of inference spend went ... by smallnamespace on Tuesday November 17, 2026 @08:54PM (#50540911)
[delayed]
Re:In 27,280 coding agent runs, 64% of inference spend went ... by wtch98 on Wednesday November 18, 2026 @04:24AM (#50543941)
> One reason is because a lot of older code is insecure

As is a lot of newer code. Depends on the threat model

> Another reason is that if you can find a better way to accomplish the goal, then you should eventually switch to it.

If it's financially better and the gains outweigh the risks, sure

> If your code from 2006 is already perfect and there is no room for improvement

Or more realistically if the cost of improving it would be higher than the business benefit gained from improving it

> which is that any code from 2006 that you’re running has survived the quality filter of time

Not really, it's terrible code. But it does the job, because it was written to solve a business problem, not written to make beautiful code.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by tempfile on Tuesday November 17, 2026 @12:16PM (#50534712)
> This implies that code is becoming more fungible

Code has always been fungible, in the sense that it could be refactored.

> you should expect any code you write today to mostly be replaced in the near future

Why on earth do you think this? It does not follow from your other points. Just because you could turn your business into a rube goldberg code juggling machine, why would you?

Re:In 27,280 coding agent runs, 64% of inference spend went ... by skydhash on Tuesday November 17, 2026 @01:06PM (#50535566)
it's amazing that you could read a book like "The Design and Implementation of the 4.4 BSD Operating Systems" (1996) and still find it relevant to browse the code of OpenBSD and NetBSD, three decades later. While good software gets plenty of updates and additions, the core ideas barely bulge over time. A lot of churn is people playing around with no reason instead of solving the problem properly the first time.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by krrrh on Tuesday November 17, 2026 @12:11PM (#50534628)
A shift is realizing that refactors go from something that was costly and infrequent to something that can happen inexpensively and frequently. It takes some rebalancing.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by rrr_oh_man on Tuesday November 17, 2026 @11:19AM (#50533650)
> objectively nicer than I would've gotten from the effort I would've put into manually coding

I find that only works with topics I know very little about.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by datsci_est_2015 on Tuesday November 17, 2026 @11:29AM (#50533849)
This comment could be interpreted at least two ways, and I’m not sure which it is:

  - tongue in cheek: in topics you know very little about, you don’t have the ability to tell whether the LLM output is trash or not
  - very serious: LLMs allow you to create things you could not have created before due to not having the knowledge
Sort of like a Dunning-Kruger or Gell-Mann amnesia.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by devmor on Tuesday November 17, 2026 @11:38AM (#50534004)
I don't see why both can't be true.

LLMs are allowing people to create things with very little effort compared to before, and those creations are of questionable quality - indeterminable by the LLM-user due to a lack of experience.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by ygjb on Tuesday November 17, 2026 @11:46AM (#50534184)
So, basically, the same as before, but LLMs provide the creations of questionable quality at a faster rate :P

I think the quality of the results vary with the rigor of the process. If you one-shot prompt a new service (even with Fable5) "build me an airtable competitor, no bugs", you will get some real code camp quality results, but a well defined set of prompts and a workflow that leverages multiple coding agents reviewing each others work, and implementing test cases at a relatively granular level yields radically different results.

Couple that with a rigorous design/spec session, and multiple checkpoints (say at release milestone) to monitor drift from the original design makes the process a bit more intentional and measurable, while still being far faster than a typical engineer can build.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by yugoslavia4ever on Tuesday November 17, 2026 @11:55AM (#50534357)
Yes, but before you learned as you went, researching and finding out what good (or at least better) looked like; what learning is done when you wave your hand?
Re:In 27,280 coding agent runs, 64% of inference spend went ... by ygjb on Tuesday November 17, 2026 @10:29PM (#50541493)
Honestly? I have more time to learn now than I did before. I can also do something that was really challenging before LLMs as well, which is turn my lack of comprehension into an demand, guided tour through a Socratic exploration of the unknown topic, including generating labs specific to the skills I want to develop and with the resources I have available.

It's like anything else with LLMs, if you treat it like a magic wand, you haven't really learned or built anything, but using it as a true assistant, it can be a powerful and flexible tool.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by contravariant on Tuesday November 17, 2026 @11:31AM (#50533876)
I think Conway's law applies. In an ideal situation your code mimics your organisational structure.

The places where you let an agent do its thing is bound by the (implicit) interface your prompt creates. In an ideal situation this interface is a simple isolated thing in your application.

If you just let it do its thing, expect a monolith. You shouldn't rely on it coming up with interfaces that mesh well with what you had in mind, and each of those interfaces represents a decision that it took, restricting future iterations of the code. So a monolith is not only what you can expect, but also the best you can hope for.

If you want to let it cook, create more places for code that does a thing and you don't care how.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by Ericson2314 on Tuesday November 17, 2026 @11:34AM (#50533920)
BTW the usual point of Conway's structure is to try to adjust the org / who is working on the software, so you don't get a bad software architecture.

The "objective" nature of code is revealing a flaw in the human division of labor.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by contravariant on Tuesday November 17, 2026 @04:05PM (#50538173)
I think the law is more powerful if you view it as a force of nature. You don't have to make your organisation conform to the shape of the software, you can often do it the other way around.

The trouble starts when the requirements for your software are at odds with the requirements of your organisational structure.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by pjc50 on Tuesday November 17, 2026 @11:43AM (#50534106)
Oh that's a good insight. The agent counts as part of the "organization" for this purpose.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by dcss_gardener on Tuesday November 17, 2026 @01:32PM (#50536011)
You can steer it a lot with implicit context too, like ants following pheromones. If you put an empty docs/ dir in your repo agents will put documentation there without being told. If you put one inline test, they'll write every test inline. If you seed a decision record they'll record their decisions. If you have a lowercase sql query they'll use lowercase sql. If you write a test for wcag AA contrast they'll make ui with wcag AA contrast. And so on.

With some up front care you can create an environment where they mostly do what you want when you let them rip without oversight. As the codebase grows it becomes less likely they'll see any one example, and more rarely expressed ones can get missed. And then codebase predating LLMs or designed without this consideration you just have to accept what you get or wrangle them manually.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by nolok on Tuesday November 17, 2026 @11:53AM (#50534318)
I don't understand why that idea was adopted that IA had to be that "give one command have a perfect run and correct answer" thing when it's clearly not how it works and an eternal "the last % is worth 99%" issue.

I know why the US frontier labs push it of course, you can't push a trillion dollar plus idea if you also say it should be a cheap commodity and the best and latest are now the answer, a bazillion mid levels are.

But most work of value I do is with a swarm of "not the latest nor the brightest" models, with a smart one controlling the dance. Most decisions or review is better done through consensus. The myth that "if we keep going it will reach a point where it's always correct" seems so obvious to me, it doesn't work like that, for that matter we don't work like that.

This is not a jab at it. I work with it. I love it. But I think the story imperative are a mismatch for the actual needs of most companies, and it's hiding between the "frontier needs" of special research & co, or the governemental needs.

On the same line of thought: it's pretty obvious as great as these things are, no they're not going to "replace all employee", and no employee without them won't be able to compete either. They're a supercharger, a self multiplying smart assistant with capabilities in any field, and it feels so wrong to me that they've been presented in a complelty different way.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by wat10000 on Tuesday November 17, 2026 @12:04PM (#50534499)
I recently tried some personal evals to see how some different models performed. I was trying Opus, Sonnet, Gemini 3.8 (this was pre-Argon) and a couple of others. I picked a pull request I'd recently made with heavy Claude assistance, one that seemed simple enough to be quick to solve but with enough subtleties that I could differentiate the results. Set up the agents to run clean without knowledge of the solution, and set them loose.

To my surprise, none of them came close to producing an acceptable PR. Quality varied, but even the very best results were missing significant pieces.

My takeaway is that the human in the loop is still massively important for many things. And that makes it really hard to come up with a useful benchmark.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by krrrh on Tuesday November 17, 2026 @12:06PM (#50534516)
> it doesn't work like that, for that matter we don't work like that.

The parallels with self-driving cars are interesting. We demand that agents don’t just produce better code than most humans (which the frontier models do), but that they perform much much much better.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by sl0wbrainRot on Tuesday November 17, 2026 @02:38PM (#50537010)
Keeping them on a short leash is leetcoding. Handing them one recursive function at a time

Devs are the PMs structuring the sprints now cause they know when/where/how to pivot

See discourse wrt who Nvy should pick as ceo (if JH doesn't come back anyway); at least from what I seen the theme is if they are not an engineer who can keep up with tech and know why Nvidia needs to change course Nvidia is toast

Same with software; everywhere I worked PMs that were not engineers sucked to be at as the PMs Duning-Kruger called the shots not the demands of technical problems before the devs

Re:In 27,280 coding agent runs, 64% of inference spend went ... by Zigurd on Tuesday November 17, 2026 @03:12PM (#50537485)
> Devs are the PMs structuring the sprints now cause they know when/where/how to pivot

You need not just the PMs but the management structure on top of them to get with the idea that product development should be iterative and experimental, not spec driven. But yeah that's how things are going.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by sl0wbrainRot on Tuesday November 17, 2026 @03:42PM (#50537888)
Look at all the small businesses where a sole proprietor and 1-2 others handle it after directly interfacing with their customers which number in the dozens

All those layers appeared in response to ZIRP to fuel job creation. They did not appear out of real necessity and iteration as you call out. You're blithely rattling off a startup spec sheet.

You need that perhaps, or believe you do from extensive prior experience.

2010s are gone; public opinion on such an economy has effectively turned the corner for good.

Same way devs complain about others getting into their flow, computer users at large are tired of a dev riding shotgun, acting like a helicopter parent. They don't want a subscription to you.

A forced obligation of your having reach into their lives while they have no such leverage feels invasive af.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by cronin101 on Tuesday November 17, 2026 @11:22AM (#50533699)
I'm far behind the curve with local/open weight models, but I thought it was common practice that you should do some last-mile fine-tuning on the particular harness that you're using and would get a reasonable return on investment for doing so (particularly on quantized models)?

Can anyone confirm/deny this is the case?

If so, it seems a shame not to include this practice into the reporting methodology for full transparency.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by jrmg on Tuesday November 17, 2026 @11:26AM (#50533789)
What does "last-mile fine-tuning on [a] harness" mean?
Re:In 27,280 coding agent runs, 64% of inference spend went ... by cronin101 on Tuesday November 17, 2026 @11:29AM (#50533829)
Taking the open weight model and altering those weights in order to overfit for success on the tool shape / quirks of the harness you are running the model within.

Last-mile aka doing in yourself (on box) vs the compute cluster that created those open weights.

https://en.wikipedia.org/wiki/Fine-tuning_(deep_learning)

Re:In 27,280 coding agent runs, 64% of inference spend went ... by ford on Tuesday November 17, 2026 @11:32AM (#50533887)
I don't think this is common, though it would be interesting if ThinkingMachines/HuggingFace/OpenRouter/BaseTen offered fine-tuned variants of the most popular open models that worked with specific harnesses
Re:In 27,280 coding agent runs, 64% of inference spend went ... by cronin101 on Tuesday November 17, 2026 @11:37AM (#50533989)
In hindsight I think I overestimated how common/achievable this is with the logic that if PewDiePie could do it, so could a layman https://felloai.com/pewdiepies-ai-model-beats-gpt-4o/

However, the hardware requirements are still quite enormous.

I could see a gap in the market for crowdfunded (gofundme?) post-training of open-weight models, per harness. I estimate that the one-time compute cost is probably $x000 so this scales nicely with a large number of interested developers that could reuse the weights.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by hedgehog on Tuesday November 17, 2026 @12:07PM (#50534539)
It's typically easier to tune the harness around the models' existing behavior. In my experience the primary failures in models > 5B params on coding tasks are just that they aren't able to translate instructions into a correct description or implementation of the problem (will write wrong code, wrong tests, and wrong documentation). The basic tool calling and syntax of generated code is fine.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by bisonbear on Tuesday November 17, 2026 @12:16PM (#50534707)
In MiMo's v2.6 public RL training run, they explicitly trained on 23 different harnesses in order to minimize overfitting on any single harness. I imagine that this technique has to be common in RL training runs given that model providers have little control over where their models are used.

https://mimo.xiaomi.com/rl/#overview

Re:In 27,280 coding agent runs, 64% of inference spend went ... by jrmg on Tuesday November 17, 2026 @11:25AM (#50533768)
"In 27,280 coding agent runs, 64% of inference spend went to runs that failed" (the title this submission was submitted with) is a pretty editorialized title.

The page's title is: "Introducing FrontierHarness Eval v1.1"

Re:In 27,280 coding agent runs, 64% of inference spend went ... by ford on Tuesday November 17, 2026 @11:29AM (#50533839)
To be clear - "Failed" here means "Did not meet all pass criteria on a given benchmark task", not "Failed to generate an answer"
Re:In 27,280 coding agent runs, 64% of inference spend went ... by scubazealous on Tuesday November 17, 2026 @11:32AM (#50533890)
While this could be better, I wonder how much time the average developer spends on runs that fail? How much time do students spend on runs that fail?
Re:In 27,280 coding agent runs, 64% of inference spend went ... by seizethecheese on Tuesday November 17, 2026 @11:36AM (#50533954)
The headline here is confusing. They ran a bunch of harness and agent combinations on terminal bench tasks, and 64% of the spending was on tasks that didn’t pass terminal bench verifiers.

It’s unclear whether that’s supposed to be surprising against the selected models for the benchmark. I suspect it’s not and the headline is written to pander to “wasting money on AI” upvotes.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by seizethecheese on Tuesday November 17, 2026 @11:45AM (#50534140)
Okay, I wanted to know “which tasks” and dig into the methodology, and its documented nowhere!

A serious effort here would use a specific set of terminal bench questions. Instead these look like cherry picked questions. How were they picked and why? Without an answer we should assume they were picked to produce some result.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by jzemeocala on Tuesday November 17, 2026 @11:57AM (#50534379)
For real, i remember setting up advanced Linux configurations back before the 3.0 kernel and I'm pretty sure I had a higher failure rate.

And even if they did try to test them on a lot of gotcha tasks; I'm perennially reminded of the ever-relevant quote about AI and other tech: "this is the worst it will ever be"

Re:In 27,280 coding agent runs, 64% of inference spend went ... by dotancohen on Tuesday November 17, 2026 @12:55PM (#50535378)
That quote is outdated. My phone nags me now twice daily to update, and the update removes features I use so I'm careful to tread around it.

My Windows work desktop has two orders of magnitude more memory and more processing power than the last time I used Windows, yet boots slower and is laggier than any Windows I've ever used.

Almost all useful information online is concentrated into YouTube and Reddit, everything is advertising and trying to steal attention. I'm not even welcome on Reddit anymore due to the media campaigns against my country and race.

We've definitely passed the point where the derivative has gone negative.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by seizethecheese on Tuesday November 17, 2026 @12:00PM (#50534425)
Confirmed. This is 17 TB4 tasks and 45 TB2.1 tasks. Why?
Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @12:03PM (#50534473)
We want to see how performance diverges between TB2.1 and TB4 while keeping costs manageable, since running the entire dataset is really expensive.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by seizethecheese on Tuesday November 17, 2026 @12:07PM (#50534538)
How were the specific questions chosen?
Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @01:07PM (#50535581)
You can see the task in https://hub.harborframework.com/datasets/frontier-harness-ev...

We kept the category unified so we have tasks for every domain like ML , science and others.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by seizethecheese on Tuesday November 17, 2026 @01:44PM (#50536192)
[delayed]
Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @12:19PM (#50534759)
Sorry for the inconvenience. We ran 27,280 total trials on Runta (1–4 vCPU, 2–8 GB RAM, 32 GB disk) using the 62-task [email protected] dataset across 5 models and 22 harnesses on Baseten, with 4 attempts per task. Every harness and model ran at max effort with a 131,072 max output token limit per request and per-task agent timeouts ranging from 15 minutes to 8 hours.

For scoring, final passes are divided by scored attempts, where infrastructure errors are excluded from the denominator, timeouts are still graded by the verifier, and reward hacking (such as looking up solutions online) counts as a failure. Costs are calculated using list price multiplied by reported tokens.

https://hub.harborframework.com/datasets/frontier-harness-ev...

Re:In 27,280 coding agent runs, 64% of inference spend went ... by seizethecheese on Tuesday November 17, 2026 @12:25PM (#50534870)
This doesn't answer the question at all.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by dotancohen on Tuesday November 17, 2026 @12:47PM (#50535250)
It's still the most informative thing to come out of that account or the linked article.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by numeri on Tuesday November 17, 2026 @06:33PM (#50539751)
unless they edited it in later, the link at the bottom lists which tasks they ran it on. It definitely doesn't answer why those tasks or why specifically 62 of them
Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @06:39PM (#50539822)
We just want to keep the category unified so we have tasks for every domain like ML, science and others, and also have a chance to see the divergence from tb2.1 to tb4.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by visarga on Wednesday November 18, 2026 @12:00AM (#50542074)
These kind of benchmarks declare "best harness" but measure just individual tasks with zero human interactivity, not project level efficiency. I don't do individual tasks, I work on larger projects made of 100s of tasks. The work across tasks, keeping intent alignment straight and testing reliable, remembering that detail from 5 days ago, that is also important.

This pretended equivalence is like claiming good LeetCode score means good engineering. I know it is much harder to measure project-wide performance but that is what we need.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @12:20PM (#50534783)
I'm also building a leaderboard in the meantime, which will release in a few days with more detailed harness version info and methodology.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by plufz on Tuesday November 17, 2026 @11:46AM (#50534177)
Yeah exactly, compared to what baseline? People here will probably be stunned when I say this, but my pre-agent code was usually not bug free on my first attempts either.
[dead]
Re:In 27,280 coding agent runs, 64% of inference spend went ... by warkdarrior on Tuesday November 17, 2026 @01:23PM (#50535858)
I wouldn't call it "burning of tokens and compute" if you get a useful signal out of it.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by themgt on Tuesday November 17, 2026 @11:59AM (#50534405)
Over the 22 harnesses, the rank correlation of the steps per pass and the cost per pass was +0.51 (p = 0.015). For the other three factors, it was −0.03 to +0.02 (p ≥ 0.88). Within each model, tool calls tracked cost as well as steps. Table 4 below and Table 5 in the appendix give the values.

For harness developers: cut steps, not just prompt tokens. Prompt size, tool count and cache hit rate did not predict cost per pass. Steps did, because every step sends the whole context again.

Incredible, none of these harness developers considered just asking the LLM to solve the problem right away! Don't siege Leningrad, take it immediately!

Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @12:04PM (#50534490)
Cutting the system prompt and tools is easy. Cutting steps requires understanding the model's RL strategy and how it was trained to explore.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by bisonbear on Tuesday November 17, 2026 @12:17PM (#50534736)
I wonder if Claude Code would do better given they just introduced code mode, which should significantly cut the number of steps required
Re:In 27,280 coding agent runs, 64% of inference spend went ... by adrithmetiqa on Tuesday November 17, 2026 @12:46PM (#50535232)
Agreed. Does this help anyone decide what model or harness to use. Nope.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @01:08PM (#50535603)
Good finding! Model preference is interesting.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by onlyrealcuzzo on Tuesday November 17, 2026 @03:08PM (#50537409)
There's thousands of benchmarks, and nobody can agree on what the best models are if you run all of the benchmarks which costs tens of thousands of dollars for frontier models.

Then, somebody runs a small subset of benchmarks - with not enough data to have any statistical significance - and makes further claims.

How are we supposed to derive value from this?

Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @04:15PM (#50538298)
Although people disagree with the best models, the worst model is obvious.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by andai on Tuesday November 17, 2026 @11:39AM (#50534019)
>Through our experiments, we found that you should consider switching harnesses when switching models to maximize your experience.

I've seen this for years, and seen several papers that studied it: the same prompt will help with one LLM and harm performance for a different LLM on the same task.

And on benchmarks which test LLM x Harness, they find that LLMs tend to do way better with their "own" harness. (Or in the case of Chinese LLMs, the harness of the LLM whose training data they borrowed :)

Re:In 27,280 coding agent runs, 64% of inference spend went ... by shevy-java on Tuesday November 17, 2026 @11:39AM (#50534026)
Well, it is a milking system. You have to pay for those tokens anyway.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by colinmarc on Tuesday November 17, 2026 @11:41AM (#50534057)
How are people even using Baseten for coding agents? I am using Pi (so very small requests) in a "normal" single-agent way and I hit the rate limits on throughput immediately.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by bisonbear on Tuesday November 17, 2026 @11:44AM (#50534120)
These are cool results and I'm glad that tokens are being invested to eval harness capabilities.

However, I am hesitant on the correlation between pass/fail in evals which are a binary 0/1 determined by tests, and real world usage, where users can steer, run code review, and care about more than just test pass/fail.

I wrote about the problem of determining "success" for coding agents, and a lot of it boils down to the fact that agents can write dumb code that passes tests, and write smart code that fails tests, and we need to shift to rewarding quality of code and alignment with intent rather than just signal on held out tests (which is still important!)

https://www.stet.sh/blog/cost-per-successful-task

Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @12:01PM (#50534445)
"Cost per pass" is distinct from "cost per successful task."

Cost per pass is simply the average cost per task divided by the pass rate.

If you draw a line from the origin to any harness on that chart, its slope (pass rate ÷ cost) represents passes per dollar. Cost per pass is the inverse of that slope, meaning the lowest cost per pass gives you the steepest line.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by bisonbear on Tuesday November 17, 2026 @12:13PM (#50534665)
In this case, I am not entirely sure how cost per pass is distinct from cost per successful task given that success is defined as pass/fail.

To illustrate the difference, the author of FrontierHarness included "Figure 3. Cost and pass rate of each harness with each model", in which we can see that Codex CLI has $0.937 cost per pass, but a $0.573 cost per task. Measuring cost per task also changes the winner to OpenCode

Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @01:35PM (#50536043)
We use cost per pass, not average cost per task, because we want to see the harness keep a high pass rate and low cost at the same time.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by fennecfoxy on Tuesday November 17, 2026 @11:51AM (#50534274)
Yeah, developers sometimes go down the wrong path as well. It's a non deterministic system exploring a huge problem space...
Re:In 27,280 coding agent runs, 64% of inference spend went ... by bityard on Tuesday November 17, 2026 @11:58AM (#50534398)
This is a good point. I'm (so far) only a hobbyist developer and _most_ of my non-trivial projects took at least 2 prototypes before getting to something that was fully usable with few caveats or technical debt.

I think this is not too different from physical products, either. You often hear of companies spending multiple years and iterating through dozens of prototypes in order to bring something new to the market.

I don't think there is a shorter path: you never truly understand a problem until you take up the engineering work to solve it.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by augment_me on Tuesday November 17, 2026 @11:55AM (#50534350)
This is completely expected. Agents are pseudo-greedy search functions that brute-force solutions using their context.

If you want perfect search, you need lossless compression, which defeats the point of the current systems that are lossy L2-optimized representations of all information.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by underdeserver on Tuesday November 17, 2026 @12:00PM (#50534427)
Wait till you hear what the human success rate used to be.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by maerF0x0 on Tuesday November 17, 2026 @12:00PM (#50534435)
So? I'm sure exceedingly high amount of EC2 cpus are idling, or lost to discarded branch predictions etc.

We waste whatever is cheap to gain what is dear.

(Of course we iterate towards better, but that doesn't mean we delay shipment/usage once it's profitable / better than status quo)

Re:In 27,280 coding agent runs, 64% of inference spend went ... by bob1029 on Tuesday November 17, 2026 @12:05PM (#50534504)
The entire point of this AI thing is that the cost of failure is now ~zero. You can try again in a few minutes / hours. A traditional software business might only complete one iteration before time runs out.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by evalstate on Tuesday November 17, 2026 @12:13PM (#50534650)
Just use a harness that has been properly integrated and tested with each model, with published traces and comparisons (author of) https://fast-agent.ai/benchmarks/

GitHub and Copilot CLI have started to do the same btw (not sure if they publish the traces/reward hack results). But we should normalise this as consumers shouldn't be forced in to playing this game.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by WorldPeas on Tuesday November 17, 2026 @12:14PM (#50534672)
evolving harnesses are also a cool development I've seen recently though I feel they'll have diminishing returns https://www.alphaxiv.org/abs/2609.01437
Re:In 27,280 coding agent runs, 64% of inference spend went ... by jedisct1 on Tuesday November 17, 2026 @12:18PM (#50534751)
It would have been nice to see swival.dev in that list.

But unsurprisingly, Codex works pretty well, too (and not just with OpenAI models), and Claude Code is just a waste of tokens and money.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by samueloph on Tuesday November 17, 2026 @12:25PM (#50534879)
Am I'm missing something or at this point the distinction between "harness" and "agent" doesn't exist anymore because too many people use them interchangeably?

Is this post mixing them up or is my understanding of these concepts off?

Re:In 27,280 coding agent runs, 64% of inference spend went ... by layer8 on Tuesday November 17, 2026 @12:29PM (#50534958)
harness + model = agent (plus sub-agents)

Do you have a different understanding?

Re:In 27,280 coding agent runs, 64% of inference spend went ... by ogmutlu on Tuesday November 17, 2026 @12:31PM (#50534992)
They mean the same thing?
Re:In 27,280 coding agent runs, 64% of inference spend went ... by jimbob45 on Wednesday November 18, 2026 @12:13AM (#50542139)
I’m struggling to understand the difference between “harness” and LangChain, which seems to have already been a failure.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by 0xbadcafebee on Tuesday November 17, 2026 @12:32PM (#50535001)
[delayed]
Re:In 27,280 coding agent runs, 64% of inference spend went ... by BugsJustFindMe on Tuesday November 17, 2026 @12:34PM (#50535025)
This is nice and all, but I really want to see the best harness for ChatGPT and Claude, not just open weight models.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by justincormack on Tuesday November 17, 2026 @12:34PM (#50535029)
You do need to optimise prompts too, eg I found that openai models just gave up in Pi, but adding a line to the prompt matching Codex eg "Once implementation is authorized, the agent should persist through the bounded change, tests, review and final validation. Complexity, expected duration and context compaction alone are not reasons to stop." fixes that. Otherwise it would look at tasks and say they were too hard.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by xyzzy_plugh on Tuesday November 17, 2026 @12:52PM (#50535332)
I don't like this framing.

For a given model, the best harness depends on the problem. As models become increasingly capable, the harness appears to matter more than the model. This is perhaps controversial but allow me to explain.

Most of my problems are in a single domain, using specific tools and programming languages. I give whatever model de jour the specific tools it needs to solve the problem. I don't let it execute arbitrary scripts or programs (though it can, if it really tries). I give it a restricted number of calls and I feed that information back into it (budget). I let it farm out work to other models, using the same harness. I give everything a budget, which is tracked and enforced, but it's not real, and refills at different times.

This bespoke, specialized harness absolutely smokes any other mainstream harness for these tasks. Significantly faster, cheaper, more correct.

But I can't build a mobile app with this harness, or make a movie.

These tests are doing the same thing. They measure a handful of scenarios. I can't easily tell if those map to my needs or not. In my experience, very few of them do.

Can these general purpose harnesses solve my problems? Yes, if I'm willing to let the model churn for a long time and burn a lot of tokens. But that doesn't feel like a sustainable practice.

Maybe I'm too early, maybe it's premature to optimize the harness this way, but I'm spending fewer tokens than my peers and getting a lot more done, and that matters.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @01:10PM (#50535646)
I am thinking about how to measure the efficiency of ultilzing existing context in one specific domain.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by botanrice on Tuesday November 17, 2026 @01:19PM (#50535790)
hey! would you be so kind as to share your method or point me in the direction towards resources on how to setup a harness as you describe?

I've been tinkering with harnesses recently but have mostly used github copilot on vscode mostly thus far as my ai tool and feel like the harness setup on there is lacking (or really confusing...). many thanks if so :)

just trying to learn without googling "how to setup a harness" and get bombarded by a trillion yt vids or competing harness providers lol

Re:In 27,280 coding agent runs, 64% of inference spend went ... by xyzzy_plugh on Tuesday November 17, 2026 @02:58PM (#50537247)
Just ask one to write the other.

"Write me a harness in $LANGUAGE that uses the $MODEL API and has tools for X Y and Z" etc etc

I added remote management and web console over tailscale in about 20 minutes last week.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by NichoPaolucci on Tuesday November 17, 2026 @05:24PM (#50539118)
I would also consider just grabbing pi and tailoring it to your needs. It's super direct, no batteries, and I added a handful of smaller tools that blow everything else out of the water.

I've got some people on my team using 80-100 skills / plugins and their setup takes FOREVER to run even a simple modification.

Mine is just as correct and the feedback loop is lightyears faster, which is what I'm personally after.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by SAI_Peregrinus on Tuesday November 17, 2026 @01:03PM (#50535508)
When a task passes all tests, it's done. More tokens should be spent on failing runs, because most runs where tokens are used should fail. For the same reason that your lost keys are always in the last place you look for them.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by brookritz on Tuesday November 17, 2026 @01:24PM (#50535875)
The TLDR saya: Basesten, Harneas and Leaderboard and still have no clue.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by winddude on Tuesday November 17, 2026 @01:29PM (#50535963)
Reading a lot of the comments everyone is talking about state actors, etc.

Let's not forget multiple departments like the Pentagon, FAA, Homeland Security etc, have been raising alarm that things like this were becoming a heightened risk since Trump and Musk cut ~300k jobs with DOGE, and slashed civilian contractors across nearly every department.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by worik on Tuesday November 17, 2026 @01:33PM (#50536025)
Given the variation between results for different, but similar capability models it is hard to take this seriously
Re:In 27,280 coding agent runs, 64% of inference spend went ... by Olscore on Tuesday November 17, 2026 @01:49PM (#50536287)
The harness doesn't matter. You need a leash.

Once you have sufficient organization around the agent/harness you start to see how a harness should be designed (for the leash.) Everybody seems intent on building a better PR-loop, or a better way to drive an IDE. To me this is backwards, and I'm trying to invert it. No IDE. No PR-loop. Outcome only.

https://github.com/pullboard-dev/pullboard

Re:In 27,280 coding agent runs, 64% of inference spend went ... by markasoftware on Tuesday November 17, 2026 @02:41PM (#50537053)
From the Hacker News Guidelines [1]:

> Please don't use HN primarily for promotion. It's ok to post your own stuff part of the time, but the primary use of the site should be for curiosity.

Your post history is almost entirely promotion of your product. Please stop. I'm flagging your post.

[1]: https://news.ycombinator.com/newsguidelines.html

Re:In 27,280 coding agent runs, 64% of inference spend went ... by elAhmo on Tuesday November 17, 2026 @01:51PM (#50536310)
So, on my daily coding tasks choice of provider, model family, model version, effort/thinking level and harness all have an effect. Some of these are multiplicative (like harness), so this is all extremely confusing for the end user about what the hell should I use do do my tasks.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by bensyverson on Tuesday November 17, 2026 @01:58PM (#50536428)
This is interesting and valuable, but I'd love to see a follow-up that tests a more parallel or subagent-heavy workflow.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by aleqs on Tuesday November 17, 2026 @02:04PM (#50536510)
Another bullshit result obviously meant to promote a specific company/product. Codex is garbage btw.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by scosman on Tuesday November 17, 2026 @02:41PM (#50537055)
We need this benchmark for non-coding harnesses. Harnesses with custom knowledge, custom tools, specific guidance, etc. A ton of the task-specific performance comes down to the skills/tool-set.

I've been building a tuner that tunes the harness (prompts, skills, tools, etc) instead of the weights of models: https://github.com/Kiln-AI/harness-tuner -- Might try to publish some "untuned" vs "tuned" comparisons, the numbers are dramatic.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by d-sanoj on Tuesday November 17, 2026 @03:05PM (#50537349)
I think the harness matters less when we are measuring the work done. There are all kinds of benchmarks in differ cases showing one model is better than the other. The existing ones like Claude Code or codex work well with their models.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @05:27PM (#50539163)
I think Claude Code does not care about token efficiency, the more users use, the more users pay.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by cpeth on Tuesday November 17, 2026 @03:36PM (#50537810)
Harnesses are all fairly simple loops from the ones I have looked at; there is not much to them.

As someone who regularly uses Claude Opus 5.5 from Claude Code and from Copilot, I would say harness barely matters in terms overall performance and behavior. Opus is nearly identical in both harness. The opposite is also true, if I switch Copilot to something like Gemini 4 Argon it's completely different even though I'm in the same harness.

Just some andecdata.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by drivebyhooting on Tuesday November 17, 2026 @04:54PM (#50538768)
Doesn’t the harness include a bunch of prompts for when to create subagents, their communication patterns, and how to verify their outputs?

That can make a huge difference, as evidenced by the agent swarms success at cracking maths.

Re:In 27,280 coding agent runs, 64% of inference spend went ... by chickensong on Tuesday November 17, 2026 @06:19PM (#50539613)
The harness is what you make it, or not. The OP claiming all harnesses are simple loops is looking at it from a simple perspective. As soon as you start customizing or building a bespoke harness, YMMV.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @07:41PM (#50540356)
I don't mean the harness, just the loop. Memory could be a part of the harness.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by chickensong on Tuesday November 17, 2026 @08:15PM (#50540624)
Sorry, I misused OP there, was referring to @cpeth in this comment thread.

Thanks for your eval work, it seems like a tough area to measure!

Re:In 27,280 coding agent runs, 64% of inference spend went ... by shepherdjerred on Tuesday November 17, 2026 @08:40PM (#50540796)
The harness is incredibly important when you're working in parallel with multiple agents. In this regard Claude Code is the best I've found.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by zodwick on Tuesday November 17, 2026 @03:42PM (#50537898)
I wonder if there is corelative improvement between the same lab's own harness and model.

For example I remember Codex having a different line number format than CC and every other harness had to have a two format tool to maximise performance.

I wonder if sucha thing exists still or do they RL on all possible combinations now

Re:In 27,280 coding agent runs, 64% of inference spend went ... by franzb on Tuesday November 17, 2026 @03:45PM (#50537933)
Original title: "In 27,280 coding agent runs, 64% of inference spend went to runs that failed"
Re:In 27,280 coding agent runs, 64% of inference spend went ... by alok-g on Tuesday November 17, 2026 @03:50PM (#50537994)
Genuinely asking: Why is Kilo code (https://kilo.ai) not present in this and most such analysis? It does not also get talked about much on HN in general.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @05:57PM (#50539442)
Kilo does not seem that popular. Point it out if I am wrong.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by reversengineer on Tuesday November 17, 2026 @04:51PM (#50538736)
Is it me, or do the results seem to show a lack of significant difference between both models? And, because of the potential non-determinism in Generative AI to create a huge degree of variance in outcomes, I am dubious about the repeatability of these results.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by Edward40 on Tuesday November 17, 2026 @05:07PM (#50538927)
We ran 4 attempts per task, that's just like when you roll dice 4 times.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by hi_hi on Tuesday November 17, 2026 @05:32PM (#50539211)
I see cost and speed being measured, but not intelligence. Is there a measure of how these harnesses impact the effectiveness of the LLMs?
Re:In 27,280 coding agent runs, 64% of inference spend went ... by krawczstef on Tuesday November 17, 2026 @10:48PM (#50541623)
There's no IP in a harness, and they don't matter that much.

Harnesses simply make different choices around certain context engineering things, what they name tools, how they compact, their system prompt, etc.

So it's not surprising that they work "better" for some models than others, but I'd bet that if you made each harness have same system prompt and tool names, the variance would shrink further...

So I would say know what you're optimizing for and construct the harness you need for it instead (it's just a while loop with an llm call and function calling after all) ...

Re:In 27,280 coding agent runs, 64% of inference spend went ... by lhk931122 on Tuesday November 17, 2026 @11:50PM (#50542012)
But I feel like later on the model itself will end up learning even the harness, so I wonder if there's any point in setting up a harness well on your own. I don't know if OpenAI or Anthropic built their harness well or upgraded their model well, but I ended up throwing away almost all of the harness I set up personally, and the time I spent on it became a waste of time. It seems like wasted effort, same as prompt engineering. From the point of view of each individual user. Since an already optimized system prompt is in place.
Re:In 27,280 coding agent runs, 64% of inference spend went ... by anigbrowl on Wednesday November 18, 2026 @12:23AM (#50542199)
[delayed]
Re:In 27,280 coding agent runs, 64% of inference spend went ... by 0xcoffee on Wednesday November 18, 2026 @12:57AM (#50542394)
[delayed]
Re:In 27,280 coding agent runs, 64% of inference spend went ... by ricardobeat on Wednesday November 18, 2026 @05:48AM (#50544618)
Surprised that Crush [1] was not included in the benchmarks. It’s my default harness for open models, and with the latest subagents update very efficient for orchestration. The built in LSP support helps models a lot in reducing the amount of tokens spent on code syntax errors.

[1] https://github.com/charmbracelet/crush

Re:In 27,280 coding agent runs, 64% of inference spend went ... by andreynering on Wednesday November 18, 2026 @07:53AM (#50545734)
Thanks! I help maintain Crush, and sub-agents are indeed pretty fun to use, and also very useful of course.