Jit

This business sends handwritten cards for other businesses. Does $10M Arr! Just wow!

@hwchase17 Dreaming. I was looking into it. It seems to becoming kind of a standard across the ai apps. But there are little to none practical examples with it.

@jit_infinity Space is moving fast! What would you like to see next that we’re not currently doing

big week at langchain, with a lot of launches: 1/ OpenWiki - auto generate a wiki of a github repo 2/ two different voice agent tutorials 3/ Harbor integration and tutorial for long running, stateful evals 4/ programatic subagents in deepagents (RLM like) 🧵with details

Sometimes I leave out context and just sit and watch the agents struggle. Reminds them who's still in charge.

Here’s a truth. By working hard you are just making 3 people rich - 1. Your boss 2. Your bank 3. And the government

Building Brain2 - a tool to save context (articles, tutorials, notes, etc) and make it portable and accessible to all your ai agents through MCP. MCP is finally working in Claude web. github: github.com/byjit/Brain2

@HxrshitYadav I know the prices won’t go down. So waiting for m6 or m7, to get better value at higher prices

I was about to buy a MacBook this month. Then Apple raised prices MacBook Air M5: 1,20,900 to 1,49,900 MacBook Pro: 1,89,900 to 2,39,900 29,000 rupees extra. Guess I am waiting a bit longer. Anyone else delaying their Apple purchase because of this?

The flow: query -> keyword / metadata search -> ranked documents -> optional reranker -> LLM Why it can win: - Cheaper - Lower latency - Easier to maintain - More deterministic And in some cases it beats vector search outright.

Vectorless RAG skips embeddings during retrieval. Instead it leans on traditional information retrieval: - Full-text search (PostgreSQL, Elasticsearch, Solr) - BM25 ranking - SQL queries and ranking - Graph traversal - Metadata filters - Hybrid reranking

Just learned about "Vectorless RAG" - retrieval without embeddings. Sharing what I picked up. In a lot of enterprise systems, exact keyword matching and metadata filtering matter more than semantic similarity. And vector infra gets expensive at scale.

@BHolmesDev supports mdx?

Introducing Hubble: the best notepad for you and your agents. - Feels like Notion or Apples Notes, but for Markdown - Live-reloading editor to collaborate with agents - Supports HTML-based apps to build interactive visualizations Free and open-source 👉 hubble.md

Built a small rust dispatcher for my ai agents. github.com/byjit/rust-rat…

@naymur_dev so cool! Glad I discovered this

We already have that liquid glass effect from day 0 🔥 Already live and production-ready in UI-Layouts: → ui-layouts.com/components/liq…

@theKhanal custom trained?

Real-time Nepali speech-to-text, running 100% offline on my MacBook No GPU. No internet. Meet Swar (स्वर).

@AnishDe10 @Google What's next?

Last Friday marked the end of my 2.5-year journey at @Google I'm grateful for the opportunity to work on Google Maps, build AI-powered systems, and learn from some incredibly talented people along the way. While I'm proud of the impact I made, I've decided to step away and explore paths that are more closely aligned with my long-term goals. At this stage of my career, I value learning, growth, ownership, and meaningful impact more than any particular company tag or title. The next chapter is still taking shape, but it will involve exploring a few new directions I've been passionate about for a long time, including spending more time on trading the markets and other builder-focused pursuits. Excited for what's ahead. Onward and upward 🚀

@sargampoudel Do you think it's risky to try it? I am afraid that I will bring harm upon my iphone

ios 27 feels very smooth. would recommend in case anyone is thinking of upgrading

@Sathibuilds Just kidding

Are you building something you'll use

@Sathibuilds At this point I am the only user

@Tristanrhee3 You are absolutely not right

prove me you're not an ai