v-yanakiev

As an example flow, you would register a user's unique secret with the service. Then, the service's client-side library would redirect all of the OpenAI requests to itself, and would also pass the unique user secret. If the corresponding user has no funding for the AI request, it would deny it.

Any interest?

https://fly.io/docs/about/pricing/

Instead of paying $0.1722/hour for a 4vcpu machine/container, you'd pay $0.03444/hour.

It'd run on Hetzner infrastructure, using Firecracker microVMs.

It'd have a minimal interface. You'd build and push a Docker image with a CLI. After that, you'd start containers / communicate with them / destroy them over REST.

Any interest?

The extension would log the data automatically when it detects that, for example, Stripe Checkout has been visited. The data would be end-to-end encrypted, and would be accessible from multiple devices (when you login + enter the E2E encryption code).

Would anyone be interested?

    1. A VS Code debug window-lookalike front-end.

    2. A package, which you install on your app, which sets up a debugger over the app's process, and INITIATES a WebSocket connection with a remote proxying service (see below).

    3. A service which proxies the debugger's connection to the front-end, establishing the debug session.

Any interest?

Syntix is an AI code interpreter. It allows you to ask an AI programming questions, and have it answer back in a Dockerfile-defined container, or in one of 8 environments supported out of the box: NodeJS, NodeTS, Python, C#, Java, Go, Rust, pgSQL.

The AI will automatically install any needed dependencies and execute the generated code in its own environment - with full internet access.

This gives the AI instant feedback on its code, circumventing hallucinations.

You can also upload files and folders to the execution environment for processing. The current size limit is about 7GB.

I use fly.io's Fly Machines for the execution environments, as well as for the app's back-end.

Here's a demonstration of the Syntix AI agent fetching the top Hacker News post of the day, based on a user request: https://www.youtube.com/watch?v=_wcJ7mTNJ0A

And here is a demonstration of using a custom environment on Syntix: https://www.youtube.com/watch?v=JlY0KBnzCzA

https://www.hetzner.com/cloud#pricing

https://cloud.google.com/compute/vm-instance-pricing?hl=en

This made me think about running a cheap High Performance Computing service on Hetzner.

Would anyone be interested in such a service? It'd allow you to define (via a Dockerfile) a high performance computing environment, and start it / communicate with it / destroy it via a REST API. Each start would provision a new VM: it'd take 1 minute for the machine to start listening. This would keep costs extremely low.