AI News
AI community posts, news and release updates
Latest updates
Screenpipe records your screen to analyze your workflow and then generates automation agents. Ideal for AI practitioners looking to boost productivity.
- YC
- Productivity
- agents
- screen recording
OpenWorker is an open-source AI agent that doesn't just chat—it delivers polished documents, sends Slack messages, updates calendar entries, and more. Runs on Mac, supports multiple models (GPT, Claude, Gemini, open-weight, Ollama), uses your own API key, and keeps data local for privacy.
- agent
- privacy
- open-source
- multi-model
- OpenWorker
- News
Kimi CEO Builds Full App Solo in 11 Minutes with K3: Barrier Drops to Lunch Money
Published Jul 23, 2026Kimi's CEO walked on stage and built a complete app solo in 11 minutes using K3. He noted that what used to require a team, budget, and months now needs a weekend and a $15 plan, reducing idea testing cost to the price of lunch.
- AI编程
- 效率
- 应用开发
- Kimi K3
- CEO演示
- Discussion
The $2 Trillion AI Circular Financing Problem: What If Chinese LLMs Eat the Frontier's Lunch?
Published Jul 23, 2026An analysis argues that the $2 trillion circular AI financing hinges on frontier companies like Anthropic and OpenAI making good on $1.5+ trillion in unfunded commitments. If Chinese open-source LLMs capture market share, they may fail, explaining the pushback against Chinese LLMs.
- frontier models
- AI financing
- Chinese LLMs
A passing test suite doesn't guarantee model quality—it may hide serious flaws, sparking a rethink on evaluation practices.
- 测试
- 陷阱
- 模型评估
- Launch
Core MCP launches: AI agents can use your crypto wallet with your guardrails
Published Jul 23, 2026Core MCP connects Claude, ChatGPT, or any MCP-compatible agent to your Core seedless wallet, with customizable guardrails, enabling AI agents to use your crypto wallet.
- MCP
- AI代理
- 加密钱包
SpaceXAI has released Grok 4.5 across all platforms (Web, X, iOS, Android), claiming it's their most intelligent model yet.
- 模型发布
- SpaceXAI
- Grok 4.5
- Launch
Gemini 3.5 Flash Cyber: Fast and Efficient Code Security Analysis, Catches Complex Vulnerabilities
Published Jul 23, 2026Google introduced Gemini 3.5 Flash Cyber, a specialized lightweight model for security teams. It checks significantly more code paths and consistently caught complex vulnerabilities that standard models missed in tests on Chrome and Android codebases.
- Gemini
- security
- code analysis
- 3.5 Flash Cyber
- Launch
Gemini 3.5 Flash Cyber: Lightweight Security Model for Vulnerability Detection
Published Jul 23, 2026Google released Gemini 3.5 Flash Cyber, a specialized lightweight model designed to help security teams spot and patch vulnerabilities before they can be exploited.
- Gemini
- security
- vulnerability
- 3.5 Flash Cyber
- Launch
Merge API Launches Fusion: Multiple Models Collaborate, Outperforming Single Frontier Models
Published Jul 23, 2026Merge API launches Fusion, a service that sends a single prompt to multiple models independently, then a judge model consolidates the answers. Benchmarks show an open panel (GLM-5.2+DeepSeek V4 Pro+Qwen 3.7) beats GPT-5.5 solo by 10.5 points and matches Claude Fable 5 quality at roughly a quarter of the cost. Each model answers the full prompt, akin to an expert panel with an editor.
- multi-model
- Fusion
- ensemble
- merge