Is it just me, or does the AWS MCP server just slow down frontier models and consume more tokens? I've seen it save one or two misconfigurations, but typically, it just takes forever and doesn't seem to improve the output.
Are schools just going to limit students to GPT 3.5 or something like they did to us with the TI-83 Calculator when the TI-89 came out? Then charge $100 for a specialty harness?
I've got to give @Railway credit for iterating so fast with the AI landscape. It is an absolute PLEASURE spinning up new products using @claudeai and the Railway MCP server. Keep it up!
There are always anecdotes about Claude models being nerfed. This seems to always happen around new model releases. Well, here's another anecdote. If @AnthropicAI needs a day to look into a model being awful, please look at Opus like right freaking now. It is absolute trash.
As someone who has had my SIM blacklisted by Verizon because they thought I stole my own iPhone by moving it from a biz to personal account, sometimes, blaming the vendor for an outage is cool.
Honestly, if @Railway emailed me pre this @googlecloud issue and said "We're going multicloud for reliability" I would have worried more about complexity-related outages. The unlikelihood of a major cloud hyperscaler cutting a client like this should approach infinity.
If you pay for a more expensive Uber due to proximity, you should be refunded the difference if it ends up taking longer than its UberX counterpart after a driver is found.
Went on a ride. Forgot to sync the camera's time to my phone. @insta360 doesn't use the EXIF timestamp that everything else uses. Claude and I built a tool to modify the internal timestamp so I can sync with my GPS stats! github.com/morethancertified…