kvyb

My AI agent is my only employee | OpenTulpa, open source

[kvyb @0xkvyb]

My AI agent just booked a $120 nail appointment at 11pm, answered 3 pricing questions, and cost $0.15 to run. Here's how: 🧵

[kvyb @0xkvyb]

@sflorimm let’s connect! I’m building an AI employee for small businesses: automations etc. And you?

[Charly Wargnier @DataChaz]

Do you realize what's happening here?

🚨 DEEPSEEK V4 PRO IS NOW 75% CHEAPER FOREVER

The new king of cost/token.

> Input: $0.435
> Output: $0.87
> Cached: $0.0036 (almost free!)

How?

V4 paper shows or SemiAnalysis: it takes just 27% compute & 10% cache compared to v3.2 👀↓

[kvyb @0xkvyb]

@DataChaz I wish this price per tokens becomes the new norm 🙏

[kvyb @0xkvyb]

@heyblake let’s connect! looking to follow more builders.

[kvyb @0xkvyb]

@nicdunz that’s great, what is she using it for?

[kvyb @0xkvyb]

@Kappaemme1926 that’s how I do it, works pretty well. Especially if you put it against a review skill: keep working until <skill> happy etc.

[kvyb @0xkvyb]

@argofowl Had to bump it from medium to high, yeah. And even on high it’s fumbling basics. Guess it’s to warm conversion from 5.5 to 5.6

[kvyb @0xkvyb]

@vikitweetus Cursor has new Compose. Claude has mythical mythos 🤔

I’d say Cursor

[kvyb @0xkvyb]

@pangshuo1981 Still too expensive though, they should learn lessons from deepseek haha

[kvyb @0xkvyb]

Qwen-3.7-Max is the model we’ve have been waiting for. Forget Gemini-3.5-Flash, this is the big thing 👇

[kvyb @0xkvyb]

@DeeplimaChtr @upupup27499533 @hiarun02 bench maxing but a bit of a hit for my wallet

[kvyb @0xkvyb]

@bil0090 Dope

[kvyb @0xkvyb]

@DeeplimaChtr @hiarun02 it’s good, but quite a bit expensive. Did you try it?

[kvyb @0xkvyb]

@yyyypoi @hiarun02 true, and that’s the big “oof” moment when using it. Not really viable, especially with painful caching.

[kvyb @0xkvyb]

@dusty_donkeyy @hiarun02 yeah, and that’s great. Competition is good for LLMs. Hopefully will make models more accessible!

[kvyb @0xkvyb]

@shikhr_ you can alias the command, that should work ;)

[kvyb @0xkvyb]

@0xSero interesting what people are using it for? fast roleplay probably.

[kvyb @0xkvyb]

@YashHustle_22 Qwen-3.7-Max for sure 👍

[kvyb @0xkvyb]

@iruletheworldmo I think the first cataclysm will happen when everything runs on slop code that is impossible to refactor or review, and 80% uptime will become the norm.