I’ve been doing some AI training and evaluation work recently through platforms like **Handshake, Snorkel, and DataAnnotation**, so I thought I’d share my experience in case anyone else is looking into this kind of work. Honestly, I wasn’t expecting much at first. I’ve tried a few online task platforms before, and some of them didn’t really feel worth the time once you looked at the pay. The AI-related work has been more interesting for me. A lot of the tasks involve things like reviewing AI responses, checking whether the reasoning makes sense, testing code, comparing different answers, and working on prompts. I tend to prefer the technical projects because they actually make use of my programming background rather than just being repetitive data-entry work. I also like that the schedule is pretty flexible. There are times when I have a few free hours and can pick up some work without having to commit to a fixed schedule. At the same time, the amount of work definitely isn't consistent, so I wouldn't look at it as something with guaranteed income. For me, the earnings have been pretty variable depending on how much work was available and which projects I was able to get into. I've had months in the **$4,500–$6,000 range**, with some stronger months getting closer to **$8,000**. Over time, I've made roughly **$30,000–$40,000** from this type of work. I don't want to make it sound easier or more consistent than it actually is, though. Some periods are much slower, and getting onto the better projects can involve fairly difficult assessments. The work also isn't necessarily available whenever you want it. Overall, I've found it useful as a way to make some additional income while doing work that's actually related to AI and programming. If anyone else has experience with **AI training, LLM evaluation, prompt work, or data annotation**, I'd be interested to hear how it's been for you.

Getting an agent to run clean with Claude Code or an MCP server in your local setup feels great until you try to make it run an end-to-end production workflow. The agent executes its narrow task fine, but the handoff between your logic, database calls, and the actual project steps still ends up messy. You spend less time writing boilerplate and more time gluing disconnected tools together just to ship one feature. When you move an agent past the demo stage, where does orchestration usually start slipping for you?

Some great talks here at The Linux Foundation's Model Context Protocol Summit in Toronto, and great people as well! Great to see what people are working on too - MCP has proven to have staying power in a market that moves very quickly. It is at the heart of the client engagements I am doing right now, because it not only provides utility and value at multiple levels, but also has a great deployment story. If you're building with MCP, or thinking about it, feel free to DM or leave a comment 🙂