[lode]

I tried it out yesterday, after reading the enthousiastic article at https://www.macstories.net/stories/clawdbot-showed-me-what-t...

Setting it up was easy enough, but just as I was about to start linking it to some test accounts, I noticed I already had blown through about $5 of Claude tokens in half an hour, and deleted the VPS immediately.

Then today I saw this follow up: https://mastodon.macstories.net/@viticci/115968901926545907 - the author blew through $560 of tokens in a weekend of playing with it.

If you want to run this full time to organise your mailbox and your agenda, it's probably cheaper to hire a real human personal assistant.

[quietsegfault]

Just watch a few videos on Clawdbot. You'll invariably see some influencer's Anthropic key, and just use that. Wokka wokka!

[0xbadcafebee]

If you have an old M1 Macbook lying around, you use that to run a local model. Then it only costs whatever the electricity costs. May not be a frontier model, but local models are insanely good now compared to before. Some people are buying Mac Minis for this, but there's many kinds of old/cheap hardware that works. An old 1U/2U server some company's throwing out with a tech refresh, lots of old RAM, an old GPU off eBay, is pretty perfect. MacBook M1 Max or Mac Mini w/64GB RAM is much quieter, power efficient, compact. But even my ThinkPad T14s runs local models. Then you can start optimizing inference settings and get it to run nearly 2x faster.

(keep in mind with the cost savings: do an initial calculation of your cloud cost first with a low-cost cloud model, not the default ones, and then multiply times 1-2 years, compare that cost to the cost of a local machine + power bill. don't just buy hardware because you think it's cheaper; cloud models are generally cost effective)

[muwtyhg]

> don't just buy hardware because you think it's cheaper

Surely there is also the benefit of data privacy and not having a private company creating yet another ad profile of me to sell later on?

[wartywhoa23]

Huge pyramids are built of relatively small blocks, kudos to everyone contributed.

[Sharlin]

"Pyramid" is an interesting metaphor to use, given the connotations.

[turnsout]

Yeah, I looked at Clawdbot / OpenClaw at the beginning of the week (Monday), but the token use scared me off.

But I was inspired to use Claude Code to create my own personal assistant. It was shocking to see CC bang out an MVP in one Plan execution. I've been iterating it all week, but I've had it be careful with token usage. It defaults to Haiku (more than enough for things like email categorization), properly uses prompt caching, and has a focused set of tools to avoid bloating the context window. The cost is under $1 per check-in, which I'm okay with.

Now I get a morning and afternoon check-in about outstanding items, and my Inbox is clear. I can see this changing my relationship to email completely.

[azinman2]

Post it!

[geek_slop]

I had the same problem. Ask Clawdbot to optimize token usage. It cut my usage in half.

[testdelacc1]

Just imagine what would happen if you asked again.

[ern_ave]

Can't you just point it at a local ollama? It'd be slower, but free (except for your electricity bill).

[itissid]

I think one thing these things could benefit from is an optimization algorithm that creates prompts based on various costs. $$, and what prompts actually gives good results. But it's not an optimization algorithm in the sense gradient descent is, but more like Bandits and RL.

There has been some work around this practically being tried out using it for structured data outputs from LLMs https://docs.boundaryml.com/guide/baml-advanced/prompt-optim...

I won't claim I understand its implementation very well but it seems like the only approach to have a GOFAI style thing where the agent can ask for human help if it blows through a budget

[columk]

That's the sad thing. There are so many millions of talented under-employed people in the world that would gladly run errands or set up automations for you for $200-$1000 per month or whatever people are spending on this bot.

Developers trust lobsters more than humans.

The other wild thing is that many of these expensive automations that are being celebrated on X can already be done by voice using Siri, Google, or any MCP client.

[jauntywundrkind]

Would have been $68 on DeepSeek, which is also imho very good.

I still have Opus review the shit out of & plan my work. But it doesn't need to be hands on keyboard doing the work.

[lurking_swe]

part of me sympathizes, but part of me also rolls my eyes. Am i the only one that’s configuring limits on spend and also alerts? Takes 2 seconds to configure a “project” in OpenAI or Claude and to scope an api key appropriately.

Not doing so feels like asking for trouble.

[lode]

That's what I did, which is why I abandoned my experiment this quickly.

I'd find it hard to write such an article about how this is the next best thing since sliced bread without mentioning it spending so much money.

[jmathai]

Are you all enabling auto reload for personal projects?

I load $20 at a time and wait for it to break and add more.

[iamtheworstdev]

not only that, but clawdbot/moltbot/openclaw/whatever they call themselves tomorrow/etc also tells you your token usage and how much you have left on your plan while you're using it (in the terminal/console). So this is pretty easily tracked...

[guluarte]

you can use your claude max subscription

[swordsith]

oh yeah let me just pull my 200$ monthly subscription out of my back pocket

[preommr]

Isn't that explictly against the TOS? I feel like Anthropic brought out the ban hammer a few days ago for things like opencode because it wasn't using the apis but the max subscriptions that are pretty much only allowed through things like claude code.

[drewstiff]

No you can't, Anthropic keep blocking it