Stop listening to Anthropic.
Stop listening to Anthropic. From “loops” to “goals” to now “tagging”, it feels like about once a month Anthropic will add some new feature to the Claude tech stack, and claim it’s the future. What should be more apparent to everyone is t…
Stop listening to Anthropic.
From “loops” to “goals” to now “tagging”, it feels like about once a month Anthropic will add some new feature to the Claude tech stack, and claim it’s the future.
What should be more apparent to everyone is the subtext to all of these new features.
They are incredibly token hungry.
Don’t be surprised when Claude goes down a rabbit hole, and to achieve a goal, chooses a roundabout hacky solution that you will need to replace.
And don’t be surprised at your usage bill after the agents guzzled tokens.
Claude Tag for Slack is no different.
What Anthropic isn’t promoting: 1) You can only use Opus, no smaller models. 2) Even on a Claude Teams plan, no subsidized usage. Only API usage. We asked it to summarize a single channel, and before it could even provide a response, it hit the $5 cap we set for testing. 3) It’s an AI governance nightmare.
Next time you see an Andrej Karpathy tweet telling you this new Anthropic feature is the next evolution of AI, remember he’s paid to market Anthropic products, and remember Anthropic is desperately trying to become profitable.