Just deployed my first fine-tuned model on @nebiustf 👀 Super smooth process, way easier than I expected. Recording a full step-by-step guide next. Stay tuned!

Introducing arXiv Researcher Agent!🕵🏻‍♂️ We built an AI Agent that: - Searches arXiv and generates clean research reports - Saves findings with persistent memory - Recalls and builds on past sessions Built with @nebiustf , @memorilab , @tavilyai and @streamlit Here's the Demo:

AI memory doesn’t need exotic databases @memorilab's Memori uses traditional SQL to power agent memory: - Short-term convos in temp tables - Long-term facts promoted to permanent storage - Rules, preferences, entities stored as structured data Proven, efficient, and scalable👇

I’ve been building a lot with Claude Code lately… But writing code is one thing, shipping reliable code is a whole different game 👀 So I recorded a full breakdown of the exact workflow I use to keep AI-generated code production-ready. Full video here👇

Gemini created it in one Prompt! Crazy to think, How far we have come!

Trying our Google's Antigravity with Gemini 3 Pro!

Just tried Anthropic’s @claudeai Code Web 👀 I pushed it on two tasks: • Fixing API endpoints across a repo • Building an AI Agent team with @AgnoAgi The results were… interesting. Some things worked great, some didn’t. Watch it here👇 youtube.com/watch?v=nIt...

Recently started exploring @EntelligenceAI, and wow, this feature is so underrated 👀 I can instantly check: - PR summaries - Sprint reports - PRs, issues, docs, and recent updates And you can even chat with it! All inside Slack. No tab-switching, no context loss.

Give your AWS Strands agents more focus! With Tool Filtering in the Strands SDK, you can: - Pick only the tools your agent really needs - Skip unused ones to keep it fast - Fine-tune behavior right from the SDK Example: load just read_documentation, search_documentation, or recommend 👇

Just dropped a new guide! Building an AI News Research Assistant with Bright Data + Vercel @aisdk It scrapes global news, bypasses paywalls, detects bias, and gives smart summaries, all in one workflow. Think of it as your personal AI journalist, built from scratch. Read the full breakdown 👇

I compared Composer-1 and Swe-1.5! Watch it now: youtu.be/aFQgUK3pgoA...

Currently Trying out Claude Code Web! Video coming soon!

Most agent frameworks stuff prompts or juggle vector stores. @memorilab Memori takes a different path: 1. Dual memory (short + long term) 2. Intelligent promotion of what matters 3. Relational DBs for structure, joins, and indexing Agents that actually remember 👇

Ready to play with @windsurf 's Aether Alpha, Beta, and Gamma!

Started Fine-tuning Models on @nebiustf! Will share a Detailed Guide Soon!

Make your AWS Strands agents return clean, structured results 🧠 With Structured Output, you can: - Get well-defined responses for any task - Use invoke_async for async execution - Simplify downstream parsing and logic Here's an Example👇

Just realized I’ve uploaded 50 videos on my YouTube channel! Not a huge number, and honestly, I haven’t been that consistent 😅 But the support from the community has been amazing! Time to change gears and stay consistent from here on. Stay tuned!

I recently tested @cursor_ai ’s new Composer-1 and @windsurf ’s SWE-1.5! I ran both on real projects: - A Monkeytype-style Typing Game - A 3D Solar System using Three.js The results were... surprising 👀 Full breakdown + live test in my new video 🎥👇 www.youtube.com/watch?v=aFQ...

Give your AWS Strands agents clean, conflict-free tools! With Tool Name Prefixing, you can: - Avoid name clashes across multiple MCP servers - Keep tool namespaces organized - Run complex setups without confusion Example: prefix tools from each server to stay conflict-free 👇

Tool calling is the foundation of AI agents! → The LLM reads your prompt → picks the right tool → formats the request → gets results → decides the next step. That’s how it thinks and acts autonomously. Now imagine multiple LLMs doing this together, that’s a multi-agent system!

AWS's Strands Agents SDK is a simple, straightforward way to build agents. 15 lines to define an agent with access to an MCP server. I found it fun to work with and appreciated how simple it was to define functions.

200+ pages of pure gold from @huggingface 🤯 They just dropped a full playbook on how to train an LLM, covering pre-training, post-training, infra, and everything in between. What worked, what didn’t, and how to make it run reliably. Link to the Paper: huggingface.co/spaces/Hugg...

AI-generated content has officially surpassed human-created content for the first time. 🤯 Honestly… this was bound to happen. Every time I open Medium or Dev. to, it’s 90% AI slop. Same titles, same tone, same “Top 10 Tools” posts on repeat.

I built my own AI Digital Twin using @HeyGen_Official’s new Avatar IV, and it looks insanely real 👀 From just one video, it generated a lifelike talking avatar with real expressions, gestures, and perfect lip sync. Honestly, it feels like the future of AI video creation. Link in comments below:

I just built my own OCR app using NVIDIA-Nemotron-Nano-12B-v2 Upload an image or PDF and instantly convert it into structured markdown using LLMs! Here’s the stack I used: - Nemotron-Nano-V2-12b on @nebiusaistudio - @streamlit for the UI Try it now 👇 x.com/Arindam_172...

Nvidia Nemotron Nano 12B VL Also knows this ICONIC Scene! I gave the Exact Same Image and the Prompt on @nebiusaistudio playground! Here's the result: x.com/MaziyarPana...

Most memory stacks today reinvent the wheel. Sometimes the wheel (aka relational databases) just works.

Relational DBs have run banks & social networks for 50+ years. Why not AI memory? @heygibsonai Memori builds on SQL for: - Entities (people, projects, tech) - Rules & preferences - Structured short + long-term memory Battle-tested storage for human-like agents 👇

True!

Want to truly understand how AI agents work? Don’t just read, run them. This repo is one of the most practical resources out there: A curated list of open-source agents you can launch, test, and break Here are a few highlights, the full list is in the repo 👇

The future of AI memory may not be new. It may be old tech used in a smarter way: databases.

Build AI Agents the Simple Way! mcp-agent is a lightweight framework to build production-ready agents with MCPs. - Manages MCP server connections for you - Implements composable agent patterns - Supports model-agnostic multi-agent orchestration 100% Open Source 🌟

Starter testing Claude 4.5 Haiku!

The only AI Gateway you’ll ever need! ⚡️ Portkey lets you route to 200+ LLMs with 50+ built-in AI guardrails Built-in observability, retries, caching & more. 100% developer-first.

Your agent can have persistent memory 🧠 Enable Conscious Ingest Mode in @heygibsonai Memori and it will: - Load essential facts once at startup - Keep them available through the session - Skip re-processing on every query Perfect for personal assistants 👇

Your Coding Tools will keep changing Mental models, and problem-solving skills won’t. Learn how to think, not where to click.

It takes just a few lines to give your AI agent full browser powers ✨ With @nebiusaistudio + @browser_use , your agent can: - Open websites - Search & sort results - Return answers in markdown Here’s how 👇

It takes just one line to automate browser actions & scrape any site ✨ @scrapegraphai Agentic Scraper can: - Handle logins & complex actions - Navigate + click + scroll + fill forms - Extract results as markdown - Use AI for structured data Here’s how 👇

Just tried out Google OPAL 👀 It’s still experimental, not a lot we can do yet. But the UI is super clean (way better than OpenAI’s Agent Builder). Built a simple workflow, worked pretty well! Even asked it to improve the flow, and it actually did 😄 Solid start from Google

I just tried Google OPAL! I must say the UI is much better!

Learnt about a New Term, “Context pollution”

Microsoft just launched its Agent Framework! → Combines AutoGen's innovation with Semantic Kernel's stability → MCP, A2A, OpenAPI support → Multiple orchestration modes → Full observability with OpenTel → Human-in-the-loop approvals 100% Open Source

Give your AI agents full memory visibility with AgentOps @heygibsonai Memori records conversations & injects context automatically. AgentOps tracks every memory operation: - When memory is stored - How context is injected - Dialogue flow across sessions Here’s why it matters👇

Scaling AI agents isn’t just about smarter prompts! Like microservices, agents need loose coupling + event-driven design to - share data, - act in real time, and - avoid rigid dependencies. That’s how they truly scale!

Anthropic Launched their Agents SDK!!🔥

Give your @crewAIInc agents a reference library With Docling Knowledge Sources, agents can: - Ingest web pages or docs - Answer with citations - Work with richer context Example: use Docling to load Lilian Weng’s papers into CrewAI 👇

Tried out @CodeRabbit CLI today 🐇⚡ And honestly… I’m impressed. > The file tree view makes the DX great! > It spotted potential issues in my code. > Even explained + suggested AI agent prompts to fix them Overall, big fan of this workflow 👀

Loving this little detail in the @coderabbitai CLI, the file tree view. 🌲 Such a simple thing, but it makes navigating + reviewing way easier

Awesome AI Apps has Crossed 5k Stars on GitHub! Thanks, everyone, for your support! I'll be adding some more projects soon!

Just tried the new @github Copilot CLI 👀 And wow… It’s quite good. I asked it to fetch issues, explain code, and even suggest project ideas. The coolest part? It actually scanned my repo and gave me solid ideas based on them! This looks Promising!