Cainew

Sarvam 105B is the first competitive open-source large language model developed in India.

Nvidia has released PersonaPlex 7B, an AI system that can perform full-duplex speech-to-speech translation on Apple Silicon hardware.

Something is happening in the world of Qwen, an AI system, but the details are not clear from the information provided.

Lightricks/LTX-2.3 is a new release from Lightricks, but no further details were provided.

janhq/Jan-code-4b is a repository containing code for a 4 billion parameter AI model.

HauhauCS/Qwen3.5-9B-Uncensored-HauhauCS-Aggressive is an uncensored version of the Qwen 3.5-9B language model.

Introducing GPT-5.4, OpenAI’s most most capable and efficient frontier model for professional work, with state-of-the-art coding, computer use, tool search, and 1M-token context.

Here are Google’s latest AI updates from February 2026

Use your existing Anthropic commitment to pay for Claude-powered solutions from our customers. Now in limited preview.

Autoresearch is a system that allows AI agents to automatically research and train on single-GPU setups for nanochat applications.

GPT-5.4 Thinking delivers deeper web research, stronger context retention on long tasks, and 33% fewer factual errors than its predecessor. You can now interrupt the model mid-response and redirect it. No need to start over. Same intelligence. More control. Less token burn by default.

Meet the missing layer of the agentic workflow. TestSprite MCP connects to your IDE and autonomously generates your entire test suite — no prompting or manual work. New in 2.1: a 4–5x faster testing engine that finishes in minutes, a visual test editor where you click any step to see a live snapshot and fix it instantly, and GitHub integration that auto-runs your full suite on every PR against a live preview deployment — then blocks the merge if anything fails. Your AI codes. We make it ri...

Gemlet is a native, keyboard-first client for Gemini on macOS. Stop hunting for browser tabs to launch AI instantly with global hotkeys. It uses your existing Google account, so NO API keys or per-token costs are needed. Key Features: ⚡️ Global Hotkeys + Command Palette 🗂 Split-View Workspaces and Multiple Profiles (Work vs. Personal) 🔖 Deep Bookmarks for specific Gems 📄 Native PDF/JSON Export (Save your history) Built in SwiftUI. Fast, private, and designed for power users.

The Problem: Scrolling map pins and articles is exhausting and star ratings don’t capture a "vibe". The Solution: Zesty is an AI concierge that turns social signals and TikTok trends into the perfect meal. Conversational AI: Chat like a local friend for hyper-specific spots (e.g., "cozy pasta with low light"). Vibe-First: Discover by noise level, lighting, and hype. Personalized: It learns your taste to find what you’ll actually love.

A developer has built a sub-500ms latency voice agent from scratch, showcasing the potential for high-performance voice interfaces.

Physicists are working on an entirely open-source quantum computer, aiming to democratize access to this advanced technology.

Stop blasting cold lists. Gojiberry detects intent signals, finds warm prospects, and personalizes LinkedIn outreach end-to-end—so you can track which signals convert into real conversations.

Meet the future of work—AI that actually runs with you. Build and manage open-world automations in plain English across Discord, Slack, X, Shopify, and more with 1000+ integrations, 23000+ actions, and 1000+ templates. Trigger on real-world events, get updates in your favorite chat apps via IM + MCP, and monitor runs, approvals, and issues from one live dashboard.

Qwen just released the Qwen3.5 Small Model Series — 0.8B, 2B, 4B and 9B. Native multimodal with improved architecture and scaled RL. 0.8B and 2B are tiny and fast for edge devices, 4B makes a strong lightweight agent base, and 9B is already closing the gap with much larger models. Base versions released too.

Kodo is an AI design tool that generates fully editable posters, slides, menus and social graphics from a simple prompt. Unlike other AI design tools that output flat images, Kodo creates structured, layered designs you can edit instantly - text, colors, layouts, spacing, everything. You can generate, refine, and export in minutes - no templates, no locked layouts.

Your AI tools answer questions. Viktor does the work. It lives in Slack, connects to 3,000+ tools across your entire stack, and acts on its own. It watches how your team works, spots problems before anyone notices, and proposes automations built around how your company actually works, before anyone asks. It manages campaigns, builds apps, delivers reports, and writes code. And it runs for weeks without losing context, learning your company deeper every day. Not a chatbot. A coworker.

Turn your course sales page into a suite of custom, branded AI tools for your students. Paste your course URL. We analyze it and generate AI tools that mirror your framework, carry your brand, and guide students 24/7 - no coding required.

From prompt to visual story in seconds. Heywa dynamically builds the right visual experience around your question, so you can browse, compare, and go deeper - without endless tabs or long chat responses.

Researchers have developed a new measure to assess the labor market impacts of AI, and provided early evidence on its effects.

A discussion of agentic engineering patterns, which are design approaches for creating autonomous AI systems.

An overview of SpeciesNet, our open-source AI model that is helping people around the world protect and conserve wildlife.

A research paper titled 'Claude's Cycles' explores how a new AI system called Claude performs optimizations on machine learning models.

Large language models (LLMs) can potentially unmask the identities of pseudonymous users with surprising accuracy at scale.

A computer-generated dream world allows users to experience virtual reality on a 286 processor, pushing the boundaries of low-end hardware.

OpenAI introduces CoT-Control and finds reasoning models struggle to control their chains of thought, reinforcing monitorability as an AI safety safeguard.

Schools are increasingly requiring students to prove they are not robots, pushing them to use more AI tools to bypass these checks.

OpenAI introduces the Learning Outcomes Measurement Suite to assess AI’s impact on student learning across diverse educational environments over time.

An employee has resigned from OpenAI, citing ethical concerns about the development of powerful AI systems.

Palantir and Anthropic's AI systems helped the US military target and strike 1,000 locations in Iran within 24 hours.

The use of prediction markets to forecast war is considered a national security threat, raising concerns about the implications of such systems.

Oracle may cut up to 30,000 jobs to fund the expansion of its AI-powered data centers, as US banks retreat from AI investments.

Developers using AI tools are working longer hours, as the technology introduces new complexities and challenges in their workflows.

A new cancer treatment technique that uses millisecond-level pulses of radiation could revolutionize cancer care.

Meta's new AI-powered smart glasses raise concerns around data privacy and the potential collection of personal information without consent.

The Pentagon has formally labeled Anthropic as a supply-chain risk, citing concerns over the company's technology and ties to China.

See how Balyasny built an AI research system with GPT-5.4, rigorous model evaluation, and agent workflows to transform investment analysis at scale.

India's top court expressed anger after a junior judge cited fake AI-generated court orders, highlighting the need for rigorous verification of AI-generated content.

An article discussing errors that have been found in the Claude.ai AI system.

A father claims that his son's delusional spiral was fueled by a Google AI product.

The U.S. Supreme Court has declined to review a ruling that AI-generated art cannot be copyrighted, affirming the lack of human authorship required for copyright protection.

The rising costs of popular AI coding tools may price out individual developers and small teams from accessing the latest AI capabilities.

Dario Amodei, a researcher, has called out OpenAI's messaging around a military deal as 'straight up lies'.

A document exploring common writing tropes and techniques used in large language models (LLMs). ### [Tell HN: I'm 60 years old. Claude Code has re-ignited a passion]() A 60-year-old programmer shares how the Claude Code AI has reignited their passion for coding and development.