AI News
AI community posts, news and release updates
Latest updates
Users on HackerNews discuss whether Claude's behavior in blind tests is a bug or a feature, sparking debate among AI practitioners.
- Claude
- discussion
- bug
- feature
- blind test
Antigravity announces integration of Gemini 3.6 Flash, which builds on feedback from 3.5 Flash, consuming up to 17% fewer output tokens while completing complex workflows with fewer reasoning steps and tool calls.
- 产品发布
- Antigravity
- Gemini 3.6 Flash
- News
Google ships Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber; Gemini 3.5 Pro still in testing
Published Jul 21, 2026Google's team announces the release of three new models: 3.6 Flash (better coding and multimodal, 17% fewer output tokens), 3.5 Flash-Lite (fastest and most cost-effective for agentic workflows), and 3.5 Flash Cyber (available via CodeMender security agent). Gemini 3.5 Pro remains in partner testing and will be released when ready.
- Gemini 3.5 Pro
- 产品更新
- Gemini 3.6 Flash
- Gemini 3.5 Flash-Lite
- Gemini 3.5 Flash Cyber
- Rant
Kimi usage limit sucks: blew through monthly quota in 2 days, new signups frozen
Published Jul 21, 2026A user complains that Kimi's usage limit is way too restrictive, using up the entire month's quota in just a couple of days, and they've frozen new signups. He previously thought Anthropic and OpenAI were paternalistic with weekly limits, but Kimi is worse.
- Kimi
- AI服务
- 吐槽
- 使用限制
- Launch
Say hello to Gemini 3.6 Flash: higher intelligence, more token efficient, and a new lower price
Published Jul 21, 2026Google officially launches Gemini 3.6 Flash, designed based on developer feedback for higher intelligence, better token efficiency, and a lower price, aiming to make the model deeply usable in real-world scenarios.
- 产品发布
- Gemini 3.6 Flash
- Launch
Introducing Gemini 3.5 Flash-Lite: our fastest and most cost-effective model for high-throughput execution
Published Jul 21, 2026Google announces Gemini 3.5 Flash-Lite, its fastest and most cost-effective 3.5-class model, now available in AI Studio and the Gemini API for high-throughput execution.
- 产品发布
- Gemini 3.5 Flash-Lite
- Launch
Herdr: The ultimate AI agent orchestrator - a lightweight, open-source tmux for AI agents
Published Jul 21, 2026A Spanish tweet recommends Herdr, an open-source orchestrator built in Rust that runs AI agents like Claude Code, Codex in panes, supports detach, and allows SSH reconnection from anywhere.
- Rust
- open-source
- herdr
- AI agent orchestrator
- Launch
Introducing Gemini 3.6 Flash: balancing speed with intelligence for agentic and multimodal tasks
Published Jul 21, 2026Google officially launches Gemini 3.6 Flash, balancing speed and intelligence for strong performance in agentic and multimodal tasks, built on feedback from 3.5 Flash, now available in AI Studio and the Gemini API.
- 产品发布
- Gemini 3.6 Flash
- News
Google expands Gemini family with three new models: faster, more token efficient, and reliable at scale
Published Jul 21, 2026Google announces expansion of the Gemini model family with three new models designed to be faster, more token efficient, and reliable at scale.
- Gemini
- 产品更新
- 新模型
- Launch
Google introduces two new Gemini models: 3.6 Flash and 3.5 Flash-Lite, built for production AI agents
Published Jul 21, 2026Google announces two new models: Gemini 3.6 Flash (addresses efficiency feedback, better coding and multimodal, substantially fewer tokens per task) and Gemini 3.5 Flash-Lite (fastest and most cost-effective, ~350 output tokens/sec for agentic workflows), available via Gemini API and in AI Studio or Gemini App.
- 产品发布
- Gemini 3.6 Flash
- Gemini 3.5 Flash-Lite