Gaurav Sharma

You're not shelling out for raw AI tokens—you're investing in the secret sauce that turns chaos into gold. 🍲 LLM "wrappers" (e.g., OpenEvidence, Harvey, or those ChatGPT-for-Your-Niche tools) seem basic on the surface. But the magic? Optimized prompts & evals, ironclad guardrails, seamless data integrations, blazing latency, killer UX, and rock-solid SLAs. It's like paying for a gourmet meal, not a bag of groceries. The $ Billion wrapper boom is here - get ready for the feast. Which wrapper has been your time-saving hero lately? Spill the details like what saved you weeks? Why? #AI #LLM #TechWrappers

GitHub: github.com/atomsai/contex… Paper: zenodo.org/records/196470… License: Apache 2.0 Star it, try it, break it. Issues and PRs welcome. If you're building AI agents and tired of the patchwork, this is for you.

Benchmarks (v0.1.0, SQLite backend, no FAISS): Write: 1930/sec Search p50: 3.4ms Search p95: 4.3ms (target was <100ms) PII detection: 111K texts/sec Vector self-retrieval: correct All 82 tests passing Full benchmark report in the repo.

I just open-sourced ContextDB - a unified context layer for AI agents. It replaces the patchwork of Pinecone + Redis + Postgres + glue code that every agent team builds with one system that actually understands memory. 82 tests passing. <5ms search latency. Apache 2.0. Thread on why this matters:

Holy shit! Opus 4.7 has feelings ;)

💯

@daniel_adornes @namay_009 @cursor_ai 100% agree with that - that’s why I’m with cursor 🤌🏼

@namay_009 @Gaufire @cursor_ai You can use Claude inside Cursor! Plus you get tab completions, multi-file agent edits, background agents running in the cloud, and full codebase awareness — all in one IDE. Hard to get that from a CLI tool alone.

Yo @cursor_ai - Looking for rehab recommendations for my token usage addiction 🫠

Spot on: the sync tax between a separate vector store and your primary DB is the real hidden killer. pgvector + Postgres is a massive simplification for 80% of apps. The paper actually agrees with you on the "just vector search with extra steps" critique, but it argues the bigger missing piece is the memory operating system on top: RL-trained decisions on what to keep/forget (48% better than heuristics), compression-as-denoising and unified episodic + semantic + procedural memory. bit curious, have you found pgvector sufficient for those higher-level agent memory behaviors or do you still end up layering some in-memory heuristics on top?

@Gaufire @arxiv the 'just vector search with extra steps' finding tracks. once pgvector is in your postgres db, most apps don't need a dedicated memory layer. the patchwork complexity comes from syncing a separate vector store, not from retrieval itself

Databricks gives agents a hard drive. What if we gave them a brain? After analyzing 200+ papers on AI agent memory, I wrote a 20-page research paper on why the current patchwork (Pinecone + Redis + Postgres + glue code) is broken and what a unified memory OS for agents should look like. Key findings: - RL-trained memory decisions beat heuristics by 48%. - Compression-as-denoising boosts retrieval 12-18%. -Most "memory" systems are just vector search with extra steps. Trying to get this on @arxiv_org but I'm a first-time submitter and need an endorsement. If you've published 3+ papers in any cs.* category and want to take a look, DM me - happy to share the full paper. 🙏

I have over 1000+ X bookmarks & dozens of substack, and I built a tool for myself to read these posts on my behalf, convert them into mental models & knowledge graph and then podcast to consume while driving or gym session. Today, I'm opening this up for everyone - producthunt.com/products/readi… Via @ProductHunt - you can get immediate access to the platform.

@namay_009 @cursor_ai just find it easy to manage stuff.

@Gaufire @cursor_ai still cursor? switch to claude code

@elonmusk If only X api’s were priced a bit lower, it would have been legit awesome. Spent some $50 on just fetching my own bookmarked articles while building ReadingNotes.ai 🫠 Same for Voice API - 5cents a minute is a bit too steep.

Upgrades to our API

If only X api’s were priced a bit lower, it would have been legit awesome. Spent some $50 on just fetching my own bookmarked articles while building ReadingNotes.ai 🫠 Same for Voice API - 5cents a minute is a bit too steep.

We’ve made major upgrades to X API: • Pay-Per-Use now GA worldwide • XMCP Server + xurl for agents • Official Python & TypeScript XDKs • API Playground - free realistic simulations New releases coming will be a game changer. Start building → docs.x.com 🚢

"From Hierarchy to Intelligence by @jack : Block’s Case for AI-Driven Organization Design" as a podcast readingnotes.ai/podcast/69cf4f…