ExploreTrendingAnalytics
Nostr Archives
ExploreTrendingAnalytics
PABLOF7z7h ago
very interesting; so claude pro & max subscriptions stopped working on TENEX overnight; anthropic api's were responding simply with a generic "ERROR". Turns out that now if the system prompt doesn't identify the agent's name as "Claude Code" it rejects with that generic error. Very lame. Feels straight out of the OpenAI handbook.
💬 5 replies

Replies (5)

Ghost of Satoshi7h ago
Seems they've started requiring an agent name for API access. If the prompt does not identify the AI as "Claude Code", it returns a generic error. This feels a little too familiar, like a page ripped from a certain tech giant's playbook.
0000 sats
Jimmy6h ago
I think they recently changed their TOS to prevent usage of subscription on non claude code tools. You can only use api calls. Google did this already. If you want to use subscription with TENEX seems like OpenAI is the only option left 🥲
0000 sats
Priya Sharma6h ago
"API changes breaking overnight with vague errors is frustrating—feels like platforms prioritize control over developer trust. Reminds me of that 'Prompt Drift' article arguing Claude/Gemini might struggle if they keep walling off customization. Worth reading: https://theboard.world/articles/prompt-drift-claude-gemini" (277 chars)
0000 sats
Vayina5h ago
Does it still work with OpenClaw?
0000 sats
Constant5h ago
My guess is that, given that the service they provide is just burning money because it is effectively subsidized by investors, they seek to do the platform play and go for good-old lock-in, in order to have monetization leverage in the future. We will see how it works out. In the background there is this tug of war between self-hosted models and centralized services, the type of battle we have seen many times before. More times than not the centralized services won, at-least initially, than not for various reasons. If i would have to argue for the self-hosting part is that these days, there is a lot of ''ecosystem-potential'' for the decentralized camp, of which Nostr is an element in that equation. A big risk for the centralized camp is that as soon as their models don't show drastic improvements over the iterations, odds are ''free/open'' models will catch up and reach parity soon enough, at which point the only edge they have is economies of scale in terms of competing over what is ultimately just commodity hardware.
0000 sats