Osman R.

@ctatedev Please review Agent Contract Language as well. You can find all details here for it. acl.fyi

Introducing Zero The programming language for agents. I wanted a systems language that was faster, smaller, and easier for agents to use and repair. Explicit capabilities. JSON diagnostics. Typed safe fixes. Made for agents on day zero.

If you've ever opened DevTools, fought XPath, or given up and copy-pasted into a spreadsheet — try it. Free to try. No card. extractly.me

Use cases I've seen in the last week: • sales teams pulling lead lists from directories • researchers building datasets from public pages • founders monitoring competitor pricing • journalists scraping public records All without writing a line of code.

Most 'AI data extraction' tools want you to write prompts, schemas, or regex. Extractly.me just takes a URL + what you want, and returns clean structured JSON. Here's me pulling 50 Y Combinator startups into a spreadsheet in under 60 seconds 👇

Just discovered Extractly.me — clean way to pull structured data from messy web pages without writing a scraper. If you've ever fought with BeautifulSoup at 2am, give it a look: extractly.me

@shemsddine Same hunt on my side. Most rooms turn out to be mostly reposts and very little code. If you land on one worth joining, would appreciate a ping back.

Looking for a no-bullshit Discord/server for people actually building & shipping AI agents + LLM work. Genuine devs only, folks who know what the words mean, share real code/architecture, debug in public, and aren't just reposting hype. Drop links if you got one. Building in the trenches daily, want real peers. #LLM #AIagents #BuildInPublic

@screenest_ai Useful map. The part that keeps biting me is not the integrations themselves, it is keeping the agent coherent as it hops between them. Memory and tool selection seem to be where most of my failures live.

The Ultimate Agentic Tech Stack (2026 Edition) To build truly "superpowered" AI agents, you need more than just an LLM. You need a robust ecosystem of integrations. Here is the comprehensive breakdown: 🌐 Web & Search • Firecrawl: Clean web crawling for LLM-ready data. • Browserbase: Headless browser for logins and complex UI actions. • Apify: Pre-built scrapers for X, LinkedIn, and Maps. • Exa / Perplexity: Real-time AI-native search retrieval. 💻 Development & Data • Codex: Core agentic framework for execution. • GitHub: Handling code, issues, and PR automation. • DeepWiki: Deep-dive technical code analysis. • Snowflake: Enterprise-grade data warehousing. • dbt: SQL-based data transformation and modeling. • Supabase / Pinecone: Backend and Vector memory (RAG). ⚙️ Operations & Productivity • Google Workspace: Connecting Gmail, Calendar, and Drive. • Notion: Centralized documentation and task databases. • Obsidian: The "Second Brain" for localized knowledge. • Linear: High-performance issue tracking. • Stripe: Automated billing and payment workflows. 🗣️ Communication & Media • Discord / Slack: Automated support and team workflows. • Bland / Twilio: Giving agents a voice for real-world calls. • YouTube Transcripts: Turning video into searchable research. • Readwise: Querying your entire library of highlights. • Granola / Fathom: Searchable meeting history and transcripts. 🔄 Automation & Intelligence • Zapier / Make: Connecting to 5,000+ third-party apps. • Replicate / Hugging Face: Running specialized open-source models. • Vercel / Railway: Instant infrastructure and deployment. The "Must-Have" Top 5: 1.Firecrawl (Web) 2.Browserbase (Action) 3.Google Workspace (Context) 4.GitHub (Engineering) 5.Obsidian (Memory)

@adishjain333 @dhaivat00 This matches what I keep hitting. Tests catch the paths I imagined, never the ones the agent actually wanders into. Curious how stable your deterministic scoring stays when the underlying tools shift behavior between runs.

Good list. The missing layer here is pre-deploy behavioral testing — slice metrics and regression catch what you tested for, but agents fail mostly on paths nobody wrote a test for. Generate trajectories from the capability graph, score deterministically, no LLM-as-judge. invarium.dev/?utm_source=x&…

Deploying AI agents without proper eval might be expensive That't where LLMOPs comes to the picture It shows you how to build evaluation - Curated golden datasets - Calibrated LLM-as-Judge - Slice-level metrics - Regression testing - Continuous production monitoring #AIAgents #Evaluation #LLM #BuildInPublic

it's a beautiful day

@zeeg Also, sent you DM. I will reach out via email too.

@zeeg David, I would love to have conversation. This is what I have done two days ago, may be something that can help tell a story that I know post-training a bit. x.com/UsmanReads/sta…

if you want to come work on things like this i have open recs at Sentry $350k to $1m TC you need to have a proven track record, ample public work, be highly autonomous, 5+ years industry experience, and be based in SF or willing to relocate. if your github profile cant compare to mine, its probably not the right fit DM your credentials (or email to david at sentry)

@iaindunning Hello Iain, Do you think this qualifies as a work that could open door for a Global Remote hire? x.com/UsmanReads/sta…

Please see job posting here: hudsonrivertrading.com/hrt-job/ai-res… I think this is a unique opportunity for top talent. Your resume will be reviewed by my team. We will pay top-tier comp, in cold hard cash, not questionable equity. Our primary focus is on NYC but we are open to LON.

New role! We're adding people to our LLM team, ASAP. We already build some of the best models of markets. For LLMs, one must ask in 2026, why another one? We think we have a (unique) story, and are hiring for the full training stack - pre-to-post. Big resources, big ownership.

@julianeagu @databricks Hi Julia, Just a day ago, in similar space, I did a research work and published my finding in this article. I will definitely apply at the career link, but I am also open to conversation here in case you find below work interesting. x.com/UsmanReads/sta…

I'm building a new team at @databricks AI Research and we're hiring. We're focused on one of the hardest open problems in AI right now: how do you measure and continuously improve agents that operate on enterprise data at scale. We're looking for founding engineers to build the flywheel that turns evaluation results directly into better agents — from development and training all the way to production. If you want to work on problems that actually matter at the frontier of AI research, I'd love to talk. Link in comments 👇

@sanascam inspiring.

But the journey wasn't easy at all, it's an insane lore, I cried Infront of God everyday begged him every second lost hope too lost patience too but somewhere in my heart I knew I had to pray more about it alhumdulilah thousand times

If any of you thinks God isn't listening to you, HE IS. Last year around this time I went to uni career expo (was still a trainee), and asked jazz hr about their associate program I was waiting to hear from - this year I sit with heads presenting my work at jazz almost everyday

14/ Everything's open. Paper, scripts, mined pairs, the 80% adapter on HF: github.com/ranausmanai/ti… Drop your base model into train_on_pairs.py. 30 min on an H100. If your base is saturated, the boundary chart tells you in advance — no need to spend the compute.

13/ So what's actually here? Not "open base models will catch GPT-5." Frontier closed models train on data this recipe can't reach. What this is: in a verifier-rich domain, an open base with headroom extracts a 30–50pp lift for ~$5 and zero human training data. Sometimes. Within stated limits.

8/ Code felt solved-ish. Onto math. Qwen2.5-3B-Base on GSM8K with an auto-difficulty curriculum: 32 → 66. +34 points. I felt great for about 12 hours.