🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Conformal prediction on time-series data with arbitrary distribution shifts over time - paper & code Paper leverages regret minimization algorithms arxiv.org/abs/2302.07869 Python GitHub: github.com/salesforce/onl… Salesforce open-source, dependency on Merlion

Manipulating LLM Knowledge -Learns a map from text -> LLM fact encodings Encodings can be: -Knowledge Editors (modify outputs to be consistent w/ facts) -Model Probes (see what properties LLM attributes to entities) Links probing, prompting & editing arxiv.org/abs/2304.00740

We started working on the Workflow Patterns almost 25 years ago to create a language to talk about workflow capabilities in a vendor- and language-independent manner. For many years, our website workflowpatterns.com/was the most visited website on workflow management. (1/5)

LLMs Meet New Knowledge While approaches like RAG are proving effective in enriching LLM workflows, one common issue is knowledge mismatch. LLMs are generally static (with a training cutoff) so injecting new knowledge could create issues with performance and factuality. This

⏰ TimeGPT: docs.nixtla.io 📈 Nixtlaverse: nixtlaverse.nixtla.io 📒 Get started with TimeGPT: docs.nixtla.io/docs/getting-s…

Introducing Privatemode AI—the always-encrypted AI service! Privatemode ensures your data stays secure with confidential computing. Available as: 🔹 Privatemode Chat (Windows & macOS) 🔹 Privatemode API Learn more: privatemode.ai

Two weeks ago, we announced Privatemode AI, the only AI service that keeps your data (Prompts, responses, metadata) always encrypted. 🔒 Do you want to know why did we build Privatemode and how we conceptualized it? Read the related blog post: eu1.hubs.ly/H0hdRGg0

See how this milestone supercharges our roadmap for privacy-first apps and Confidential AI 👇 🔗 blog.opensecret.cloud/opensecret-joi…

AI agents can make process orchestration and automation much more dynamic, but only when AI is operationalized in a way that is composable, scalable, and flexible. See agentic orchestration in action in this demo by Senior Developer Advocate Niall Deehan: bit.ly/3Hu6SnS

China didn’t like that Maple AI encrypts user data away from prying eyes 👀 Maple is banned in China ❌

Welcome to the OpenSSF family, OpenBao A Series of LF Projects, LLC Project! 🎉 OpenBao is a new sandbox project focused on secure secrets and encryption management—originally forked from Vault & now evolving to serve open source communities even better. Read the blog: openssf.org/blog/2025/06/1…

This is not possible with Maple AI - end to end encryption for all AI inference. In fact, if you get our models to say something particularly edgy, please share the prompts with me! I’m always on the hunt for better jailbreaks 😂

Researchers create benchmarks that decide AI model worth, yet capture 0 value from their work. Evault uses Ritual's TEE infra for ❖ verifiable evaluations ❖ where creators monetize their expertise. the cutest Ξngineer on the current state of benchmarking and how Ritual fixes it ↓

Tired of chasing down lineage across emails, Slack threads, and asking around? The new DataHub MCP Server puts metadata context right where your team works. Now live in v0.3.12. 📖 Read the release blog: hubs.la/Q03v3S4K0 #Metadata #AI #DataHub #MCP #AIReadiness

SLMs are already capable of commonsense reasoning, instruction following, and code/tool interaction at levels comparable to 30–70B models, with orders of magnitude better throughput. Examples include Phi-3, Hymba-1.5B, DeepSeek-R1-Distill, and RETRO-7.5B.

🛡️ Introducing Maple Teams: Secure AI for Your Organization If you're handling sensitive data in your company or organization, this is for you. No more compromising privacy for productivity. Let's dive in. 👇

what if chatgpt could forget what you typed… like actually? as you can see most "private ai" today is cap. you type something, and somewhere, the model just saw your entire message, raw, full and nothing obfuscated. even if it’s encrypted in transit, the model itself still

🚨NVIDIA CEO just redpilled the MIT study claiming AI makes people dumber > “I haven’t read the MIT research lol” > “I literally use AI every day and my cognitive skills are actually advancing” > most of his job as CEO = asking good questions > “I’m not asking a machine to teach

The paper provides a taxonomy of context engineering in LLMs categorized into foundational components, system implementations, evaluation methodologies, and future directions.

Traditional agent planning approaches often fail in enterprise scenarios due to unstructured plans, weak instruction following, and tool selection errors. Routine provides a clear and modular format for LLM agents to follow multi-step plans, reducing ambiguity and improving tool