Where I come from (Malawi, Africa), WhatsApp is so widespread that most people prefer it over email - to the extent that people don't really check their e-mails unless it's required for work or they are applying for something. For most people, WhatsApp is the de-facto communication channel.
I help moderate a community of developers and we hit the whatsapp group limit of 1024 members and sometimes have to wait for someone to leave (intentionally or accidentally) before we can add new members. We've tried to move people onto "better" platforms like Discord or Slack but we always end up coming back to WhatsApp which is subsidized via MNOs (mobile network operators) social media data/internet bundles and for the fact that most people are just stuck on whatsapp.
Hi Eric, this is very nice work! I played with this and love the idea and execution. I have had similar thoughts about how chatting with AI seems to lack some of the elements you mention that we get from talking to other people.
I appreciate that you've made it open source and will be checking out the code and maybe that can get me to finally play with Tauri :)
This has inspired me to move my personal blog to Hugo aswell. I have been using Hashnode[0] for the past few years and while it's okay, they recently automatically deleted one of my blog posts which was written in my local language, Chichewa and was one of my popular amongst, even non-developers.
Ironically, my company's blog and websites are built with Hugo.
Thanks for all the great work. How good is the model at composing actions and is there a way to say, give the model ability to scope actions, for example if actions are related to permissions or some other context? Would one need to pass the role or permission as context or finetune separately?
I think you mean taking the results of one function call and putting it into another? We saw some promise but didn't heavily train for this use case in the base model. The thing we noticed with the 270m sized models, and the performance expectations of AI models in 2025, is that these size models perform best for _specific users_ when finetuned to that specific use case.
What I suggest is mocking some data either by hand or using some automated tool and finetuning in this kind of use case and using the finetuning colab setup.
> is there a way to give the model ability to scope action for example if actions are related to permissions
Permissions depend on your system architecture more than the model. The model itself just takes in tokens and outputs tokens. Permissions are defined by your security/system setup in which the model itself is running.
If you output Zod you've basically solved Typescript, a Zod schema more or less is a Typescript type and can be explicitly made so with the `infer` capability
Trying to keep it simple but I can already feel some "design pressure" to think about making the DSL more complete (language) by adding features like loops and variables. Still early days!
I help moderate a community of developers and we hit the whatsapp group limit of 1024 members and sometimes have to wait for someone to leave (intentionally or accidentally) before we can add new members. We've tried to move people onto "better" platforms like Discord or Slack but we always end up coming back to WhatsApp which is subsidized via MNOs (mobile network operators) social media data/internet bundles and for the fact that most people are just stuck on whatsapp.
reply