Hacker News Reader

Comment by holistio | Hacker News Reader

You pay $200/month to Anthropic, $200/month to OpenAI, $200/month to Cursor, $200/month to $200/month to Google, and seeing that it didn't come to a nice round $1024/month, you pay $200/month to Sakana to coordinate it all, because why not.

While you're at it, feel free to send me $200 as well, I'll generate a crypto address ending with "AI".

View on HN

My current setup:

    $20/month: Claude Code
    $10/month: Minimax
    $16/month: Xiaomi Mimo
    $10/month: Opencode Go

Opus at low/medium effort generates plans. Then several coordinator/worker pairs are possible: DeepSeek v4 Pro + Minimax M3, Mimo v2.5 Pro + Mimo v2.5, Mimo + Minimax, Sonnet 4.6 + Haiku. I've been running hundreds of long multi-agent sessions, topped up extra credits here and theere, but haven't reached $200/month spend yet. Relying entirely on Claude/Codex feels like a waste of cash now.

TIL: I just found out that base58 disallows I (capital i), l (lowercase L), O (capital o) and 0 (zero), so I could only generate GrxoJt4eNXE2QaQ55iPSa7hhiYdzCo8ZeAuokmh2Cai.

(don't send anything, sharing only because of the base58 fun fact I didn't know)

More fun facts:

Omitting those characters makes it good for generating passwords if they need to be typed in by hand.

Double-clicking a base58 string always selects the whole string and it doesn't wrap accidentally, thanks to missing / and +, so it's also convenient to copy and paste.

Unfortunately, no special characters means that a base58 string will often be rejected as a secure enough password.

at this point I might just try Neuralwatt and see how much request I can get with GLM5.2. I've read a lot of reviews that its very cheap to run using Neuralwatt cloud

Pay $0 to run a local model or even a cheap DeepSeek V4 model via their API which is close to free per million tokens.

These prices are just going to get raced to $0.

Maybe. But for now it's fascinating how $200/month has kind of become a normal tier.

It's similar to how AirPods normalised all of us having $300+ headphones. All of us would have scoffed at the idea a decade ago.

Many people here spent a lot more than $300 on headphones long before AirPods appeared.

I had a really nice Sennheiser before that, too. But now you hop on the subway and everybody sports one.

Those were hobbyists, audiophiles, professionals, artists (recording, performing, etc.).

They are talking about a much larger group of people.

I think OP meant noise-cancelling headphones, which were fairly ubiquitous in tech circles in open offices; before Apple launched AirPods.

Airpods Inc. would be very high up SP500 as a standalone business.

The Sony WH-1000XM series and the Bose QC35 were the standard quality headphones years before AirPods were a thing, and both retailed at $300+.

Of course, premium headphones existed before. I have a WH-1000XM4 sitting right next to me.

But your aunt Josie didn't have one. Now Apple is selling 80 million units / year and the ~$300 price tag has become normal. Before that, most people had headphones that were 10 times cheaper.

$300 isn’t what AirPods cost though. You can get a pair of AirPods 4 for $129 on Apple.com, and I presume that is still the most popular model. If you’re paying ~$300, you are buying premium headphones.

The base model where I live (Central Europe) is $194. The Pro is $357. The Max is $779.

I just averaged it out.

But, it is not all about cost: models like DeepSeek v4 flash (I use the US company Fireworks.ai and also buy tokens directly from DeepSeek) is very fast, very low latency while working.

Would you want to use a text editor that updates the screen very slowly? Kind of the same thing for using agentic systems as coding assistants: don’t want a ‘sluggish’ experience.

I have, mostly, long running autonomous tasks, so it doesn't matter how slow inference is. If I optimize for latency it means I'm turning into the limiting factor.

Not while the hardware required to run a local model at an acceptable speed costs way more than $200.

Guess what, the big players are hoarding all the RAM and GPUs so that other people can't afford decent hardware. It's working out beautifully for them!

> Not while the hardware required to run a local model at an acceptable speed costs way more than $200

It's $200/month. You have to take into account energy costs and all the rest of a system, but if you break even within 1-2 years ($2400-$4800) it'd be a pretty good deal. And $4000 buys you a pretty decent system.

Sure, if you're going to keep using it long term.

But it's a hefty upfront investment for people who just want to experiment. The good thing about $200/month subscriptions is that you can cancel them any time and cut your losses. Not so with a $4000 computer that loses half of its resale value as soon as you plug it in.

I think the current sweet spot for people who don't already own a high-end gaming PC is to rent a server with a beefy GPU from Hetzner et al. and run local models there.

I used to have a $20/mo ChatGPT subscription and now I spend $12 per year using Kimi models on OpenRouter, and that's with zero-data-retention-only providers (some models sometimes have free providers with scary tracking). Maybe I just don't use that many tokens, I don't fill the context with more than what's needed for a specific request, but it goes to show how these subscriptions can be an absolute ripoff. The thought of spending 200x that is insane to me

The beauty of your approach: when people are not paying for an expensive subscription, they can decide to use models less and not feel like they are leaving money on the table.

Not everyone can run local models. It is also expensive will be outdated soon as the model evolves.

Where do you acquire free hardware and free electricity such that you can host local models for $0?

I wish I only paid $200/mo for Anthropic! Multiply that by 20x.

What are you getting out of it at $4000/month?

i burned ~20k+/mo on codex.

Did you make those money back?

It is hard to stretch every single token to a win but …

The major two deals it was purposed to are still up on the air , if we win sure , 60x win

Redesigned and rearchitected our company's backend and mobile apps in about 2 months with 1 engineer and 1 PM trained to work with Claude. We used a custom-built multi-agent harness to crank through lots of parallel tasks, which necessitated the API usage.

Does it work? I’m less interested in economics than fit with an MVP.

https://news.ycombinator.com/item?id=48625727

Or use openrouter and switch to model you want to use..(i think so)

Or TrustedRouter if you want privacy and open source

You ought to realize that shilling your product in the comments doesn't exactly come across as trustworthy.

Disclosing affiliation hasn’t been a legal thing for a while. It’s reputational. Knowing that firm spams is a black mark.

Oh! I thought TrustedRouter was a joke/sarcasm. Very wrong placement of the comment.

It’s all open source and I say that it’s mine in all the sibling comments above

Happy user here, pairing it with Composer 2.5, with Fugu Ultra as advisor and Fugur as planner. For scope/architecture it’s on par with useful Fable-style orchestration than one chat thread.

I've been shipping production on archive.tw with Fugu Ultra in /advisor on oh-my-pi.

Advisor doesn’t slow the loop if the driver stays fast. Worth it if your harness can split advisor from worker.

Which software are you using to do that?

Edit: nevermind, but which plugin or so?

Yo dawg, I heard you like agents, so we put agents in yo agents so you can burn tokens while you burn tokens.

Reply on news.ycombinator.com