haniehz

thanks for asking! The MCP server hosts the web assets by either serving them directly (like sending the built React files) or linking to an external web client that hosts them. Once loaded, ChatGPT renders that content inside the chat for users to interact with.

based on the article, it seems like a good reasoning model like gpt5 or opus 4.1 might be good choices for the planner. I wonder if the gpt oss reasoning models would do well

pretty soon! I think it's already happening. Just a matter of time for people to adapt.

I fed ElevenLabs Music a single prompt about our open-source MCP agent framework and got back a complete song: vocals, instrumentation, arrangement, the works. Zero post-processing.

Here's what caught me off guard: the vocal phrasing. Not just the melody, but the micro-timing, breath placement, and emotional inflection. The model placed emphasis on "composable" in a way that actually reinforced the technical meaning. It added vocal runs that felt intentional, not algorithmic.

Technical details that worked:

Prompt structure: [Genre] [Mood] [Key technical terms] [Narrative structure] Generated: 2:04 track with verse/chorus/bridge structure Quality: Comparable to demo-level indie recordings

What this means: Voice synthesis was the laggard in generative AI. That's changing rapidly. We're moving from "impressive for AI" to "actually usable in production workflows." Non-English limitations: I tested it with different languages and hit a wall — very patchy results, nowhere near the English quality. Anyone have experience with non-English lyrics? Curious about phoneme handling across languages.

The gap between human and AI musical performance is shrinking faster than I expected. Worth paying attention to.

Watching LLMs mimic reasoning so well definitely makes you wonder how much of our own thinking is just pattern prediction with extra steps.

How are you thinking about scaling topic and source curation without turning into another aggregator that makes assumptions for the user?

Great question and those are solid tools! Where this differs is the orchestration. Instead of switching tabs or manually checking five sources, this bundles everything into one interactive, LLM-assisted report, plus it remembers your preferences, investment style, and context. You can even swap out GPT for a local model if you’re privacy-conscious or budget-sensitive.

I didn’t build this to beat the market...it started as a way to reduce my own decision fatigue and make sure I wasn’t missing obvious signals. That said, it’s more of a research agent than a trading bot. For ROI, it’s not auto-trading, but it has helped me avoid a few bad calls and made my reports more consistent and thorough.