rmonvfer
Joined 5 years ago
Joined 5 years ago
Cloudflare’s authoritative DNS uses EDNS Client Subnet (ECS) to return different IP pools based on where the query originates. Spanish resolvers get IPs from a range that La Liga blocks. If your recursive resolver is physically outside Spain (or you use DoH/DoT to tunnel to one), Cloudflare returns a different, unblocked pool.
AdGuard DNS works well for this.
Also, in my experience (having built in this space before), regulations aren’t really the issue. Court rulings are, because there’s no open data for them in Spain. And the potential users for a paid product (legal professionals) already know the law; the key players (big law firms) have their own databases of annotated and verified court rulings and other documents.
Maybe it’s a little bit overkill but I set it all up once and only have to whitelist sites every once in a while so it’s not really an annoyance. Besides, I’m not 100% sure now but I’d say that just using uBlock is enough (if properly configured) to prevent cookie-based tracking so my setup is definitely over-engineered.
Start times are atrocious, long conversations literally break rendering (and I’m using the desktop app on a MacBook Pro with an M3 Pro CPU and 36 GB of RAM so an electron app shouldn’t feel so janky so frequently, right?).
IMHO they just don’t care right now (and I get it from their perspective) but I’m pretty sure there’s a lot of low hanging fruit they could fix to make the experience 5x or even 10x smoother. Claude Code (TUI) is also another memory hog and I’ve even had Ghostty literally crash with enough open sessions.
Fortunately Anthropic models are also incredibly so at least I’ve been able to speedrun my own desktop client using Tauri and fix many of the issues I encounter every day. I might release it at some point.
I was spending unholy amounts of money and tokens (subsidized cloud credits tho) forcing Opus for everything but I’m very happy with this new setup. I’ve also experimented with OpenCode and their Zen subscription to test Kimi K2.5 an similar models and they also seem like a very good alternative for some tasks.
What I cannot stand tho is using sonnet directly (it’s fine as a subagent), I’ve found it to be hard to control and doesn’t follow detailed instructions.
Software engineer based in Spain, I'm currently building agentmode.co
You can reach me at ramon<at>agentmode.co