Archive

Quoting @SpringStreetNYC · Mar 13, 2026

Long response but, I have mixed emtions about a lot of it, including my own decisions.

For me, a bunch of stuff came together at the same time:

Disclaimer:
a.) I was still using OpenClaw at this point (but the same happened with nanoclaw)
b.) I'm not affiliated with Anthropic in any way other than being a subscriber

During setup, I saw the option to use Claude Code. I figured... "Hey, I got the 90€ Max subscription and in one year, I've never reached its token limits. So, how bad could it be?"

So I chose to use my subscription auth option over plopping in an Anthropic API key. I'm thinking nothing of it. And then, you know, used it for a couple of days quiet heavily.

Shock 1: I looked at the stats and saw: I had burned through 18.4M tokens in one day. Eighteen. Million. Tokens. For basically configuring an agent to use WhatsApp and a cron job? See t.co/zs6VMJexpc for details.

I was like "wat."

Then I saw the "Anthropic officially bans using subscription auth for third party use" thread on HN. Somehow, I found TZubiri's "all you can eat buffet" analogy the most surprising:

A similar scenario would be if you go to an all you can eat buffet, you pay for a single person, but then you actually unload an army of little clones that start eating the whole buffet. Technically it was an all you can eat buffet and you paid the price right? Well no, come on, don't play dumb. - t.co/WMshusbl7n

Shock 2: Oh, that 90€/mo. plan (that has completely changed how I work) is heavily subsidized by Anthropic assuming fair-use.

Shock 3: And oh, without noticing, I had started carting off plate after plate after plate from the "all you can eat buffet" that assumes fair-use.

Shock 4: And ohhhhh shit, they could have unplugged my subscription because of all that.

All together, this threw me off and I tried to process it here: t.co/RJsaawzofc

So. For me it was all of these coincidences that really made me think about my part in the fair-use and decided that I would...

a.) rather keep my subscription intact
b.) start using the actual API
c.) keep an eye on cost
d.) make a conscious decision of cost vs. value a few weeks later

And long story short, I personally feel better about myself by playing by (what I believe to be) fair rules.

And it's funny because does this really matter?

Do I, as the subscriber, really have to take into account how my actions affect the "feelings" of a "gazillion-dollar company" that I'm paying for a service?

Is the "all you can eat" buffet analogy even true?

I don't know and I'm still conflicted about it all.

What I do know is:

That I'm paying for my "fair-use" API-based nanoclaw assistant ~$2 per day which feels completely above board.

It costs an extra $10 (one-time) to add a new skill or improve the existing ones every few weeks. Again, above board.

I feel like I'm doing "the right thing" but I could totally be wrong, too!

Quoting @SpringStreetNYC · Mar 13, 2026

Long response but, I have mixed emtions about a lot of it, including my own decisions.

For me, a bunch of stuff came together at the same time:

Disclaimer:
a.) I was still using OpenClaw at this point (but the same happened with nanoclaw)
b.) I'm not affiliated with Anthropic in any way other than being a subscriber

During setup, I saw the option to use Claude Code. I figured... "Hey, I got the 90€ Max subscription and in one year, I've never reached its token limits. So, how bad could it be?"

So I chose to use my subscription auth option over plopping in an Anthropic API key. I'm thinking nothing of it. And then, you know, used it for a couple of days quiet heavily.

Shock 1: I looked at the stats and saw: I had burned through 18.4M tokens in one day. Eighteen. Million. Tokens. For basically configuring an agent to use WhatsApp and a cron job? See t.co/zs6VMJexpc for details.

I was like "wat."

Then I saw the "Anthropic officially bans using subscription auth for third party use" thread on HN. Somehow, I found TZubiri's "all you can eat buffet" analogy the most surprising:

A similar scenario would be if you go to an all you can eat buffet, you pay for a single person, but then you actually unload an army of little clones that start eating the whole buffet. Technically it was an all you can eat buffet and you paid the price right? Well no, come on, don't play dumb. - t.co/WMshusbl7n

Shock 2: Oh, that 90€/mo. plan (that has completely changed how I work) is heavily subsidized by Anthropic assuming fair-use.

Shock 3: And oh, without noticing, I had started carting off plate after plate after plate from the "all you can eat buffet" that assumes fair-use.

Shock 4: And ohhhhh shit, they could have unplugged my subscription because of all that.

All together, this threw me off and I tried to process it here: t.co/RJsaawzofc

So. For me it was all of these coincidences that really made me think about my part in the fair-use and decided that I would...

a.) rather keep my subscription intact
b.) start using the actual API
c.) keep an eye on cost
d.) make a conscious decision of cost vs. value a few weeks later

And long story short, I personally feel better about myself by playing by (what I believe to be) fair rules.

And it's funny because does this really matter?

Do I, as the subscriber, really have to take into account how my actions affect the "feelings" of a "gazillion-dollar company" that I'm paying for a service?

Is the "all you can eat" buffet analogy even true?

I don't know and I'm still conflicted about it all.

What I do know is:

That I'm paying for my "fair-use" API-based nanoclaw assistant ~$2 per day which feels completely above board.

It costs an extra $10 (one-time) to add a new skill or improve the existing ones every few weeks. Again, above board.

I feel like I'm doing "the right thing" but I could totally be wrong, too!

Quoting @karpathy · Mar 7, 2026

I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then:

• the human iterates on the prompt (.md)
• the AI agent iterates on the training code (.py)

The goal is to engineer your agents to make the fastest research progress indefinitely and without any of your own involvement. In the image, every dot is a complete LLM training run that lasts exactly 5 minutes. The agent works in an autonomous loop on a git feature branch and accumulates git commits to the training script as it finds better settings (of lower validation loss by the end) of the neural network architecture, the optimizer, all the hyperparameters, etc. You can imagine comparing the research progress of different prompts, different agents, etc.

t.co/YCvOwwjOzF
Part code, part sci-fi, and a pinch of psychosis :)

Quoting @SpringStreetNYC · Mar 5, 2026

~$1.68 per day. That's what having a nanoclaw-based assistant costs me. t.co/eMhYVjkIqZ

Quoting @nikitabier · Feb 11, 2026

Prediction: In less than 90 days, all channels that we thought were safe from spam & automation will be so flooded that they will no longer be usable in any functional sense: iMessage, phone calls, Gmail.

And we will have no way to stop it.