1
Socials: - github.com/GeiserX - mastodon:mastodon.social/geiser - https://geiser.cloud ---
I built/vibe-coded VPNBypass because I was tired of manually managing routes every time I connected to my VPN. Some services (streaming, banking, local network) need to bypass the VPN—but especially, and the trigger for this whole project: my VPN was blocking my own site, lynxprompt.com. Doing this manually with route add commands is tedious. Some VPNs implement this and call it split-tunneling, but not mine.
VPNBypass sits in your menu bar and automatically detects your VPN connection and gateway, resolves domains to IPs and adds the appropriate routes, refreshes DNS periodically (since IPs change), and updates /etc/hosts for proper DNS resolution. It comes with 50+ pre-configured services (YouTube, Telegram, etc.) and you can add custom domains.
On the technical side: it's built with Swift/SwiftUI as a native macOS app (13+). It uses a privileged helper via SMAppService to avoid repeated sudo prompts. It also detects Tailscale exit nodes automatically, monitors network changes, auto-applies routes on startup and VPN connection, verifies routes... etc.
Would love feedback—especially on services I should add to the default list, edge cases with different VPN setups (OpenVPN, WireGuard, corporate VPNs), or any issues you encounter. I've only tested it with GlobalProtect, so please report if you see any problems with other providers.
GitHub: https://github.com/GeiserX/VPNBypass
License: GPL-3.0
Thank you!
I built it because I got tired of rewriting the same “how I want the AI to code” rules every time I started a new repo. LynxPrompt generates and manages AI coding rules/config files in a portable way across IDEs and other AI-enabled coding tools. It also lets you save, share & discover blueprints (also called templates/AI configs/prompts — the industry has many names for these) made by other devs.
Honestly, I know: yet another AI tool. But my problem was very specific: keeping AI coding rules consistent across projects and tools without relying too much on “memory” features. LynxPrompt focuses on bootstrapping a repo config quickly and making those rules portable and versionable.
What it does:
- Wizard generator: bootstrap an AI config for an existing repo or a new project in minutes
- Portable rules: keep your AI coding preferences consistent across coding sessions
- Blueprints: publish/share (and optionally sell) your team or personal setup
One feature I like a lot is having the API enabled in LynxPrompt, so your AI of choice can self-update its coding rules and save/version them in the platform (also, if you configure it that way in the wizard).
I’m posting to get feedback (and ideally a few early users): Does the “portable AI coding rules” idea make sense? or what would you need to trust shared/paid blueprints (previews, diffs, versioning, ratings, etc.)?... What is your real pain here?
Links below. Thank you and "happy" vibe-coding — at least, we have LynxPrompt ;).
- First blog post: https://lynxprompt.com/blog/thrilled-to-welcome-you
- Docs: https://lynxprompt.com/docs
- Ideas/bugs/support (please, show some love): https://lynxprompt.com/support
- Wizard (requires sign-in—sorry, I decided to do this to prevent abuse): https://lynxprompt.com/wizard
I’m looking for a recent post (around 3-6 months old) in which grid animations are shown to demonstrate the process of heart muscle cell activation, and on a later stage, how a clump of dead cells can provoke a “vortex” and electrical reanimation is needed. I tried searching for it with Algolia but it’s not showing up.
Thank you in advance.
It's three months already and we have a stuck serverless export. Rebooting the instance does nothing. This impedes us to create new exports, so we are only relying currently on snapshots.
Is paying Google Support the only way forward? Are we really being forced to pay to get this solved? This should be free.
Additionally, who's going to pay for all the minutes in which the serverless export is stuck doing nothing?