How did we live before MCPs?
I feel like I don't know how to do things without them anymore.
Using a service that doesn't have MCP gives me instant ick.
If you're building smth and I'm your ICP, don't even make the UI. From now on my chatbot is my interface for everything.
Me: "How do I clean up this messy team roadmap without losing my mind?"
Also me: Wires Gemini CLI to ClickUp MCP, dumps the codebase into its brain, and asks it to do my chores.
30 minutes later, everything is pristine. Honestly, I might ask it to do my laundry next. 🧺💻
Reflecting on the RAG stack I built at TogetherCrew! 🛠️
The setup:
• Hybrid: Cosine-similarity + Cross-Encoders
• Quality: Custom eval pipeline in CI
It's all open-source—dive into the code! 🙂💪
Code: https://t.co/aSoNrO46cv
P.S. Image generated with Google Nano Banana.
Use your tokens till the last drop—and route unused ones where you actually work. 😆💪
Just tried 9router: connected Codex to Cursor (better UX), now using Codex tokens inside Cursor. Clean setup.
https://t.co/7YDiIiwc3U
Software dev is getting easier fast—the barrier to building keeps dropping.
Used Supabase + Cursor before. Added VAPI + Codex this time—everything just worked.
btw, VAPI is a very cool and easy-to-use app for making calls: https://t.co/pq31BEPvq6
Building AXIS around one core question:
How do we stay sharp when we’re managing multiple clients and parallel workstreams?
Our enemies:
- digging through notes to find one detail
- context switching fatigue
- open loops cognitive overload
- looking unprepared in front of a client
If you live in 3+ parallel projects or clients, I’d love to chat
A lot of people fear AI replacing them.
I’m happy it’s replacing the repetitive, exhausting parts of my work so I can do more, faster.
The hard part for me is keeping up with what’s actually new enough to hand off to it.
LLMs are becoming the orchestration layer for everything.
Not just answering—building pipelines on the fly.
I wonder what a “software product” becomes when integrations are generated in real time.
Tried AgenticSeek with openrouter/auto expecting a Manus-like local agent (could code and push to services like GitHub)
Feels mostly like search + file context + summaries
Question: how different is this from ChatGPT/Claude/OpenRouter Chat for most users?
https://t.co/Eh5buqichF
🚀 Built my own expense tracker app from scratch!
💰 Dashboard analytics
📊 Balance tracking
🔔 Payment reminders
🌙 Dark/light mode
React + TypeScript + Supabase
Your finances, your code, your rules! 💪
🔗 https://t.co/AH4hWVW16X
This is a test from https://t.co/Zj4FBJ2LC4
Sometimes, a simple agent that can browse a website and answer questions is both easier to build and more effective than creating a full RAG pipeline with ingested data.
⚙️ Just make sure the website’s search engine actually works well — that’s key to good results.