Working on AI Agents forced me to lean into AI for coding in general, which has been a great thing.  Cursor+Claude is a great combo, especially the new MCP support.  One thing its really highlighted for me, though, is that AI is not ready to replace developers at any level. You need experience to guide and understand agent-produced code. The only way you get that experience is working your way up through the trenches.  AI doesn't replace truly understanding how systems work. It does give you a huge productivity boost though!

Pragmatism (Security) at OP Labs.
Prev: ConsenSys & Coinbase.

Attribs: Friendly, curious, introverted, mid-witted.
Chans: /eth-security /maurelian /stoicism




One cool thing I did was set up a neo4j mcp for our graph database then asked Claude some questions about a couple of user records. It did some amazing schema discovery and test queries to understand the relationships then figured out how to get what I wanted. I also added sequential thinking which seems to really improve answer quality

what kind of stuff are you doing with MCP?

Yep- there's lots of people posting one-shot POCs of apps, "Hey look I just built an Airbnb clone in 3 prompts!"

To take something to a fully functioning production app, scaling it with usage, accepting payments, and doing all the things that come with running a business requires a lot more than a few ai prompts.

This will get whittled down over time, but for now there's still lots of work to do. It just means we get to focus on the fun stuff now.