Hacker News Reader

Comment by thevinter | Hacker News Reader

Personal anecdote: I exhausted my Claude usage yesterday so I decided to spend 20$ to try Kimi while I was at it. Logged in, paid, downloaded Kimi Code, set it to use K3 and prompted something along the lines of: "Check this repository and find all the settings that can be passed as input related to hardware, I/O, thread control or networking. Produce a report".

It thought for about 12 minutes and then told me I had exhausted my daily quota. (The next day Fable did the same task in 3m)

If you want to buy a plan for K3 do NOT buy the 20$ one.

View on HN

I had the _exact_ same problem. Got the 20 USD/month subscription from Kimi.com (paid annually), exhausted my 5-hour quota with a simple prompt on OpenCode + Kimi K2.7 through their API. Cursor got the same prompt done in minutes.

According to their web interface, I'm also on 23% of my weekly usage. That feels crazy, as I do a lot more than that on the 20 USD Cursor plan, and never got even a warning.

I thought OpenCode was doing something wrong, but you had the same issue with Kimi Code.

Moonshot AI has two subscriptions archetypes and the one they link to from their main page is the subscription for the web interface/research/agents/OpenClaw (that they host for you)

There is a small link below the tiles for these plans saying something "looking to use it for code? Use Kimi code" and upon clicking you'll get a basically identically looking subscription page with different sub prices and more quota

Non coding plans:

   * Plus (19 "base agent" -- wahtever that might be)
   * Pro (39 x2 )
   * Max (99 x6)
   * Ultra (199 x12)

> https://www.kimi.com/membership/pricing

Coding plans:

    * Starter (19 - "base coding")
    * Explorer (49 - x4)
    * Expert (99 x10)
    * Master (219 x25)

> https://www.kimi.com/code

I stumbled on that when i tried Kimi K2.5 back when that was released.

It's also worth mentioning that the $20 get you a 0.25 priority for API usage - because the others get x4 priority. So get ready for a _slow_ experience

Thank you so much. I was about to ask you for links, since they have multiple domains that look similar (kimi.ai, kimi.com, moonshot.ai, etc.). I wasn't sure which one was the legit one, or if they all are.

The plan I got is called "Moderato", which somehow isn't any of those 8... I'll try to read more about it.

That doesn’t appear to be the case anymore. They just have one set of plans: Allegreto, modereto, etc which apply to both the chat app and the coding apps.

Can confirm, I see these monthly prices on both pages:

Moderato: 19 USD or 15 USD annually

Allegretto: 39 USD or 31 USD annually

Allegro: 99 USD or 79 USD annually

Vivace: 199 USD or 159 USD annually

However, they're saying that the plans might get split up again:

> New Membership Plans Coming Soon

> Kimi and Kimi Code benefits will be separated. Existing subscribers are unaffected.

> You can still buy the current plan before launch.

Same as with the Tweet:

> Going forward, we'll also split membership into two more focused plans: Kimi Membership for Kimi Web, App, and Work; and Kimi Code Membership for coding workflows. This will help us match compute more precisely and keep the experience stable.

Problem is I don't know whether to pull the trigger on the annual Vivace now or wait for after the split - whether it's gonna be a cash grab afterwards on the account of the demand, or whether I might save some money if I buy it after the split.

So I am likely coming from a completely different world because I use entirely self hosted models - but is it really common in your workflows to just wait for 12 minutes without any feedback? I am constantly watching my model and following the path it’s taking and cutting it off or steering it if it’s headed down a dead path. I was just curious if this is really common to just have no feedback for 12 minutes?

I have Claude workflows that it takes 30+ minutes to get any feedback while it thinks

Well, to be clear, there was some feedback (e.g. it was calling subagents and printing some thinking messages).

That said, yes, it is pretty common for me to wait 10, 20 and sometimes even 30 minutes without steering the model or looking at what its doing. I usually write a pretty detailed prompt at the start that describes the issue, the usecases, the testing to do and the definition of success; then I sometimes ask it to draft a plan and give it a read, but after that I just press enter, let it run and come back when it's done.

In the end I do a manual review, both of the code and the functionality, but most of the time I get exactly what I wanted.

I recently had GPT 5.6 work on something for 54 minutes while I went to sleep

I had two agents work from 7pm to 6am today... a routine for me.

Today we needed to migrate to a cheaper CI/CD provider and do a better robotic for a particular task. GPT-5.6 Sol.

Usually, the longer the AI works on something the crappier its output because that means the context is getting filled up.

If my AI prompt takes more than 5 minutes to execute, I'm almost certain to stop it and break up the task or be more specific about what files it should look at or what subtask it should tackle first. The people who let it run for hours are absolutely not getting anything worth keeping in the end.

Exactly.

I've found 6 minutes or so the sweet spot for upper bound with 5.6 Sol.

And it sounds like the OPs query above requires scanning throughout a large portion of the codebase, which will inherently consume a large number of input tokens. No locality to it.

>Usually, the longer the AI works on something the crappier its output because that means the context is getting filled up.

This was definitely truer with older models but isn't necessarily the case now.

They frequently do other things apart from navel-gazing that take a lot of time but get good results, like spinning up subagents to solve some hairy task in a loop.

Stopping/distrusting long-running AIs is a habit I've had to unlearn myself.

I frequently get good results from a 30+ minute Fable session, when I've asked it to do something complex (e.g. run the QA tester in a loop and eliminate all crashes, one commit per crash fixed)

This is my experience as well. Lots of success with running one hour Fable sessions (one hour of API time). Not using a loop here, but there's often lots of documentation for Claude to go through first before building out the project.

The main failure point for me with long running Fable sessions now is just that it might hit a safety guardrail and downgrade to Opus midway.

I have workflows that run for 8-10 hours unattended regularly. This is on frontier models though.

But if I were to look in on them they’d be giving feedback during that time, but they wouldn’t need steering.

I have the same on local models.

I assume a lot of these 8-10 hours are spend calling tools and getting feedback from that?

Yeah absolutely. And from other context windows for agents.

What do these workflows do? I can't think of anything that needs to run for so long.

Pretty straight ahead coding workflows. Most of my operational workflows are much simpler (because diagnosing an operational issue is much more time sensitive obviously).

As a casual AI code user, I can’t even go 30 seconds without claude grinding to a halt asking for permission to run a bash script for the 10th time.

The first step is building a sandbox (vm, container, whatever) that allows you to trust the agent in its most permissive mode.

This varies based on the maturity of my projects but most of my sessions live in VMs with very specific egress rules and permissions.

I think this is an area where the developers of the agent harnesses should be spending a lot of effort. It's not reasonable to expect your users to set up and maintain a VM just to safely use a tool. When I run cmake, I don't have to worry about it accidentally deleting my root filesystem or posting my account credentials to github. Most of the command line coding tools I use don't even need/have a --dry-run option.

Could be. But a) I don’t know why I’d trust them to do that when existing tooling means I don’t have to and b) this is not a new paradigm for me. I’ve been running most of my dev environments in sandboxed mode for a long time. I don’t trust npm either.

Frustratingly none of the coding harnesses seem to default to just firing up a docker container with the working directory mounted and letting the AI run wild inside the container.

You can get that working for yourself trivially. My agent environment now has 3 different kinds of compute abstractions, 3 different agents and a nascent workflow engine in it.

It works how _i_ want, which is really refreshing.

Oh I've built the same too. But it took me a few hours to work out the quirks, when it should just be the default.

What local model?

DeepSeek V4 Flash

K3 is relatively slow, but 12 minutes points to the harness not the model IMO.

Codex works for this, it'll continue thinking and will answer the prompt even if your limit is reached.

I feel like this is the right way to do it. Especially if you're using your own credits. Unless it's obnoxiously long task.

So often it'll cut off just as it's about to finish.

Buys $10 worth of credits.

> Codex, translate the entire PostgreSQL codebase to idiomatic Rust, make no mistakes.

Next day do it in Zig.

The day after that, Zig broke compatibility, fix that.

I had the same experience with Claude the other day. I asked it for something fairly simple and it used up my whole 5 hour quota without producing the answer. Felt like a total waste of time.

What's the app? nodejs todo app with a "database back end" in a text file, or a 1m line spaghetti code COBOL app from 1971?

kinda surprising to me after the macos web demo was apparently done on the $40 monthly coding plan which was like a 6 hour task

Reply on news.ycombinator.com