From a tweet@thdxr09/24
OpenCode Desktop Is Already Used by Millions: 2M Weekly Active Users
Documentation
Journal
Frontend
Backend
Admin
CI/CD
Loading the page.
Loading the page.
./news · auto-feed
updated 04:45 AM MSK
● live
The machine filters the important news from the stream and translates it into Russian: releases, prices, tips. Nothing extra. RSS
From a tweet@thdxr09/24
OpenCode Desktop Is Already Used by Millions: 2M Weekly Active Users

From a tweet@github09/24
Review GitHub PRs through Muse without switching tabs
«GitHub 🤝 @Muse A new way to connect your GitHub account: review PRs, handle tasks and notifications, leave comments, and skip switching b…» — GitHub

«GitHub 🤝 @Muse A new way to connect your GitHub account: review PRs, handle tasks and notifications, leave comments, and skip switching b…» — GitHub
From a tweet@garrytan09/24
A coding setup speeds up work without extra editor time: Harry Tan's experience
«On one hand, a well-configured coding stack can seem like syntactic sugar. On the other, I have seen it genuinely help fix bugs faster and ship new features without increasing time in the editor.»
Garry Tan
«On one hand, a well-configured coding stack can seem like syntactic sugar. On the other, I have seen it genuinely help fix bugs faster and ship new features without increasing time in the editor.»
Garry Tan
From a tweet@claudedevs09/24
Claude Code Projects work on your computer: local support added
«We just added local support for Projects in Claude Code, so threads can now run directly on your machine.»
ClaudeDevs
«We just added local support for Projects in Claude Code, so threads can now run directly on your machine.»
ClaudeDevs
From a tweet@steipete09/24
Claude Code Stops Silently Skipping Instructions: v2.1.281 Fixes AGENTS.md
«Hmm. There is a catch here. https://blog.szypowi.cz/p/claude-code-reads-agents.md-only-when-telemetry-is-on/»
Peter Steinberger 🦞
«Hmm. There is a catch here. https://blog.szypowi.cz/p/claude-code-reads-agents.md-only-when-telemetry-is-on/»
Peter Steinberger 🦞
▶
From a tweet@wesbos09/24
Garage opened from Tesla via MCP: Grok connected to Home Assistant
«Is this the first car with MCP support? The latest Tesla update gave Grok Bot connectors, so I connected my Home Assistant MCP to Meross g…» — Wes Bos
▶«Is this the first car with MCP support? The latest Tesla update gave Grok Bot connectors, so I connected my Home Assistant MCP to Meross g…» — Wes Bos
From a tweet@thdxr09/24
OpenCode can be run from another interface: OpenChamber 2.0 picks up settings immediately
«The OpenCode server protocol underpins everything we do. It lets other developers build their own interfaces, such as OpenChamber, and turn them into polished, feature-rich products. We expect more of these interfaces as people begin exploring what they want from tools for working with agents.»
dax
«The OpenCode server protocol underpins everything we do. It lets other developers build their own interfaces, such as OpenChamber, and turn them into polished, feature-rich products. We expect more of these interfaces as people begin exploring what they want from tools for working with agents.»
dax
From a tweet@kentcdodds09/24
Let the Agent Choose the Path: Kent Dodds Follows Its Advice 80% of the Time
«My interaction with agents is changing. I used to spell out what to do myself, down to the smallest detail. Now I make sure the agent considers all the necessary tradeoffs, then ask it to propose options and a recommendation. I follow it 80% of the time.»
Kent C. Dodds 🐨
«My interaction with agents is changing. I used to spell out what to do myself, down to the smallest detail. Now I make sure the agent considers all the necessary tradeoffs, then ask it to propose options and a recommendation. I follow it 80% of the time.»
Kent C. Dodds 🐨
From a tweet@elonmusk09/24
Grok 4.7 ranking recalculated: 44 runs bypassed the block
«Grok 4.7 climbs the rankings»
Elon Musk
«Grok 4.7 climbs the rankings»
Elon Musk

From a tweet@claudedevs09/24
Claude Code keeps working after you close your laptop: Cloud sessions leave preview
«Cloud sessions are officially available and no longer in research preview! They let Claude Code keep working even when the laptop is closed…» — ClaudeDevs

«Cloud sessions are officially available and no longer in research preview! They let Claude Code keep working even when the laptop is closed…» — ClaudeDevs
From a tweet@theprimeagen09/24
AI is changing the programmer's craft: ThePrimeagen challenges the bruised-ego thesis
«I see it differently. I think most people I know entered the industry because they genuinely loved writing code. Now AI is changing every part of this industry. And it is not ego that is breaking: it is the very process of the craft we were paid for and loved. But there is also plenty to be excited about. The old way of working will fade for most programmers over the years, but there is still so much that can be done. I am not someone who predicts the end of the world.»
ThePrimeagen
«I see it differently. I think most people I know entered the industry because they genuinely loved writing code. Now AI is changing every part of this industry. And it is not ego that is breaking: it is the very process of the craft we were paid for and loved. But there is also plenty to be excited about. The old way of working will fade for most programmers over the years, but there is still so much that can be done. I am not someone who predicts the end of the world.»
ThePrimeagen

From a tweet@thdxr09/24
OpenCode 2 will retain old configs: V1 plugins will need to be ported
«Making good progress on the OpenCode 2.0 blog post.» — dax

«Making good progress on the OpenCode 2.0 blog post.» — dax

From a tweet@emollick09/23
Paying for AI will get cheaper: the same GPQA Diamond result now costs 725 times less
Epoch AI chart on the cost of reaching 25% and 75% on challenging benchmarks, 22.09.2026.

Epoch AI chart on the cost of reaching 25% and 75% on challenging benchmarks, 22.09.2026.
▶
From a tweet@zeddotdev09/23
Delta runs multiple tasks in parallel: separate threads with isolated workspaces
«New in today's Delta release (v0.17) Ask an agent in one thread to launch several tasks: each will start in its own independent top-level…» — Zed
▶«New in today's Delta release (v0.17) Ask an agent in one thread to launch several tasks: each will start in its own independent top-level…» — Zed

From an announcementGoogle DeepMind09/23
Voice for your app now comes from a prompt: Gemini speaks 100+ languages


From a tweet@cursor_ai09/23
Cursor catches regressions before users do: Rollouts checks every deployment
«Introducing Rollouts. Rollouts creates a monitoring plan, then tracks changes as they roll out. Deployments are checked, so regressions c…» — Cursor

«Introducing Rollouts. Rollouts creates a monitoring plan, then tracks changes as they roll out. Deployments are checked, so regressions c…» — Cursor
From a tweet@github09/23
Large PRs now open without lag: GitHub Copilot app tested on 1 million lines
«A PR with 2 200 files, more than 1 million changed lines, and more than 400 inline comments. Impossible to review? Not with the GitHub Copilot app. ✨ The redesigned PR view stays fast and smooth even with such an extreme diff. 🧵»
GitHub
«A PR with 2 200 files, more than 1 million changed lines, and more than 400 inline comments. Impossible to review? Not with the GitHub Copilot app. ✨ The redesigned PR view stays fast and smooth even with such an extreme diff. 🧵»
GitHub
From a tweet@openrouter09/23
No need to split a long task: Space Bunny Alpha handles 1 million tokens
«Meet Space Bunny Alpha on OpenRouter ⚡ A flash model with fast processing, configurable reasoning, and a 1 million-token context window. It accepts text, images, and video. Try it and share your feedback: https://openrouter.ai/stealth/space-bunny-alpha»
OpenRouter
«Meet Space Bunny Alpha on OpenRouter ⚡ A flash model with fast processing, configurable reasoning, and a 1 million-token context window. It accepts text, images, and video. Try it and share your feedback: https://openrouter.ai/stealth/space-bunny-alpha»
OpenRouter
From a tweet@romainhuet09/23
ChatGPT handles tasks by voice: plugins now work in Live on web and mobile
«Now you can talk with ChatGPT about anything, even when you are away from the screen! Manage email, calendars, and finances, separate what matters, and take action through plugins, including those built by developers. The interface is starting to disappear. You simply speak and get things done!»
Romain Huet
«Now you can talk with ChatGPT about anything, even when you are away from the screen! Manage email, calendars, and finances, separate what matters, and take action through plugins, including those built by developers. The interface is starting to disappear. You simply speak and get things done!»
Romain Huet
▶
From a tweet@higgsfield_ai09/23
You can build a 3D game with Opus 5.5: Higgsfield shows off arcade shooter
«Opus 5.5 built this 3D arcade shooter using Higgsfield. Tiny characters, big-city chaos. This is just insane now.» — Higgsfield AI 🧩
▶«Opus 5.5 built this 3D arcade shooter using Higgsfield. Tiny characters, big-city chaos. This is just insane now.» — Higgsfield AI 🧩
From a tweet@bcherny09/23
Claude.ai became 3x faster: the team did it in two weeks
«If you've noticed that claude.ai and the desktop app have become faster over the past few weeks, here's how we did it. The post collects plenty of useful findings and techniques for engineers speeding up their own apps.»
Boris Cherny
«If you've noticed that claude.ai and the desktop app have become faster over the past few weeks, here's how we did it. The post collects plenty of useful findings and techniques for engineers speeding up their own apps.»
Boris Cherny
From a tweet@heygen09/23
Delegate avatar videos to an agent: HeyGen connects to Muse
«Your avatar has been waiting for an assistant for a long time. We're joining @Muse. Tell your AI agent what you want to say, and it can wo…» — HeyGen
«Your avatar has been waiting for an assistant for a long time. We're joining @Muse. Tell your AI agent what you want to say, and it can wo…» — HeyGen
▶
From a tweet@petergyang09/23
Claude Artifact paints photos in the style of Monet: Opus 5.5 demo available through Customize
«Okay, one more fun Opus 5.5 demo: I quickly built an app: upload any photo and watch Claude render it in the style of Monet, Van Gogh, and…» — Peter Yang
▶«Okay, one more fun Opus 5.5 demo: I quickly built an app: upload any photo and watch Claude render it in the style of Monet, Van Gogh, and…» — Peter Yang

From a tweet@claudeai09/23
Claude Becomes a Showcase for AI Agents: Marketplace Launches with 15 Products
«You can now find tools, agents, and expert partners in Claude Marketplace. There you can: - Connect connectors and plugins, including Slac…» — Claude

«You can now find tools, agents, and expert partners in Claude Marketplace. There you can: - Connect connectors and plugins, including Slac…» — Claude
From a tweet@claudedevs09/23
Claude's Interface Got Faster: Load Time Fell to 0.55 Seconds
«Over two weeks, we made claude.ai 3 times faster. Here is how we use Claude to measure, debug, and speed up the service. Includes prompts and methods. https://claude.dev/blog/how-we-made-claude-ai-faster/»
ClaudeDevs
«Over two weeks, we made claude.ai 3 times faster. Here is how we use Claude to measure, debug, and speed up the service. Includes prompts and methods. https://claude.dev/blog/how-we-made-claude-ai-faster/»
ClaudeDevs

From a tweet@testingcatalog09/23
You can now assign work to an agent by voice: ChatGPT Work is available in the browser and on phones
«ChatGPT Voice now supports plugins and is available in ChatGPT Work! Tasks in it now also run on models from the GPT-6 family. That means…» — 🚨 AI News | TestingCatalog

«ChatGPT Voice now supports plugins and is available in ChatGPT Work! Tasks in it now also run on models from the GPT-6 family. That means…» — 🚨 AI News | TestingCatalog

From a tweet@cline09/23
Parallel Tasks Get Their Own Branches: Cline Desktop Adds Worktrees
«New in Cline Desktop v0.0.34: > Worktrees: run several tasks in parallel, with each working in its own branch in `~/.cline/worktrees/<id>`…» — Cline

«New in Cline Desktop v0.0.34: > Worktrees: run several tasks in parallel, with each working in its own branch in `~/.cline/worktrees/<id>`…» — Cline
From a tweet@code09/23
Agent History Opens Faster in VS Code: List Loads in 0.1 Seconds
«🚀 The latest @code release speeds up work with large lists of agent sessions, simplifies their management, and lets you run agent sessions in Dev Containers locally and on remote hosts. 🧵 What's new: https://code.visualstudio.com/updates/v1_139»
Visual Studio Code
«🚀 The latest @code release speeds up work with large lists of agent sessions, simplifies their management, and lets you run agent sessions in Dev Containers locally and on remote hosts. 🧵 What's new: https://code.visualstudio.com/updates/v1_139»
Visual Studio Code
▶
From a tweet@runwayml09/23
Editing no longer needs the browser: Runway integrates with DaVinci Resolve
«Runway is now in DaVinci Resolve. Access the best models right on the timeline: generate, edit, and upscale without leaving your project.…» — Runway
▶«Runway is now in DaVinci Resolve. Access the best models right on the timeline: generate, edit, and upscale without leaving your project.…» — Runway
▶
From a tweet@emollick09/23
Game vibe coding reaches graphics review: Claude updates Rescue Raiders
«There is an abandoned game called Rescue Raiders that I loved as a kid. I asked Claude to make a modern remastered version with new graphi…» — Ethan Mollick
▶«There is an abandoned game called Rescue Raiders that I loved as a kid. I asked Claude to make a modern remastered version with new graphi…» — Ethan Mollick

From a tweet@garrytan09/23
Capy speeds up PR work: Garry Tan estimates at least a 4x gain
«I use @capydotai every day for every PR in GStack/GBrain and work at least 4x faster than when I was simply using raw Codex/Claude Code thr…» — Garry Tan

«I use @capydotai every day for every PR in GStack/GBrain and work at least 4x faster than when I was simply using raw Codex/Claude Code thr…» — Garry Tan
▶
From a tweet@artificialanlys09/23
Four Models Shift the Price-Quality Frontier: 11 New Points in the Rankings
«This week, the release of MiMo-V2.6-Pro, Claude Opus 5.5, GPT-6 Luna, and GPT-6 Sol shifted the Pareto frontier between the Intelligence In…» — Artificial Analysis
▶«This week, the release of MiMo-V2.6-Pro, Claude Opus 5.5, GPT-6 Luna, and GPT-6 Sol shifted the Pareto frontier between the Intelligence In…» — Artificial Analysis
From a tweet@danshipper09/23
AI Agent Plans an In-Person Meetup: Every Hands It the Entire Meetup
«Tomorrow at 18:00, we will find out whether an AI agent can throw a good party. The @every agent planned Every’s September gathering end to end: from the food and drink menu to the guest list. Come assess the result at our Brooklyn townhouse! RSVP: https://every.to/events/every-irl-september-2026»
Dan Shipper
«Tomorrow at 18:00, we will find out whether an AI agent can throw a good party. The @every agent planned Every’s September gathering end to end: from the food and drink menu to the guest list. Come assess the result at our Brooklyn townhouse! RSVP: https://every.to/events/every-irl-september-2026»
Dan Shipper
▶
From a tweet@higgsfield_ai09/23
In Houdini, AI Takes Over the Node Network: Higgsfield Shows Chainmail
«Bringing AI into Houdini. Claude Opus 5.5 versus GPT-6 Sol: building chainmail with Higgsfield, configuring node networks, checking collis…» — Higgsfield AI 🧩
▶«Bringing AI into Houdini. Claude Opus 5.5 versus GPT-6 Sol: building chainmail with Higgsfield, configuring node networks, checking collis…» — Higgsfield AI 🧩

From a tweet@josevalim09/23
Tidewave for Rails Removes Monthly Limits: 30 Days Without a Card
«For @railsworld attendees, we put together a complete AI toolkit for web development with @rails. You can try it free for 30 days with no l…» — José Valim

«For @railsworld attendees, we put together a complete AI toolkit for web development with @rails. You can try it free for 30 days with no l…» — José Valim
From a tweet@rauchg09/23
AI agent survives a reboot: Drives preserves Vercel Sandbox files
«Muse, Instinct, OpenClaw, Claude Code… All successful agents have three key parts: 🧠 Brain → the model and the scaffolding, meaning the operating logic 👐 Hands → tools, computer, browser 🗃️ Files → memory, skills, repositories The «simple» route is to assemble all of this in one stateful computer. For example, a Mac mini. That is, you run `claude` or `fx` on your Mac and keep them running all day with `caffeinate`. It has storage, CLI tools, and apps installed. But if you want to run agents in the cloud without unnecessary costs, these parts have to be split apart. 🧠 Agent logic can run in Fluid compute. To reliably survive restarts, deployments, and failures, the event log is made durable through Workflow. 👐 Hands can be moved to a separate fleet of browsers like Browserbase/Kernel, a computer like Sandbox, and even lighter tools like just-bash. 🗃️ 🆕 What was missing was a way to split out storage too. Imagine a nightly cron job that brings an agent's memory together, a «dream». It reads and writes files directly, without starting an entire agent computer. Today we introduce the perfect complement to Sandbox: Drives. We released a computer for agents, and now we are giving them an «external drive» they can connect whenever needed. This is an early version, but the capabilities will expand quickly. Splitting these parts into independent blocks not only noticeably reduces costs, but also makes security and auditing much simpler. I would say that otherwise, it is impossible to run a safe agent at all.»
Guillermo Rauch
«Muse, Instinct, OpenClaw, Claude Code… All successful agents have three key parts: 🧠 Brain → the model and the scaffolding, meaning the operating logic 👐 Hands → tools, computer, browser 🗃️ Files → memory, skills, repositories The «simple» route is to assemble all of this in one stateful computer. For example, a Mac mini. That is, you run `claude` or `fx` on your Mac and keep them running all day with `caffeinate`. It has storage, CLI tools, and apps installed. But if you want to run agents in the cloud without unnecessary costs, these parts have to be split apart. 🧠 Agent logic can run in Fluid compute. To reliably survive restarts, deployments, and failures, the event log is made durable through Workflow. 👐 Hands can be moved to a separate fleet of browsers like Browserbase/Kernel, a computer like Sandbox, and even lighter tools like just-bash. 🗃️ 🆕 What was missing was a way to split out storage too. Imagine a nightly cron job that brings an agent's memory together, a «dream». It reads and writes files directly, without starting an entire agent computer. Today we introduce the perfect complement to Sandbox: Drives. We released a computer for agents, and now we are giving them an «external drive» they can connect whenever needed. This is an early version, but the capabilities will expand quickly. Splitting these parts into independent blocks not only noticeably reduces costs, but also makes security and auditing much simpler. I would say that otherwise, it is impossible to run a safe agent at all.»
Guillermo Rauch
▶
From a tweet@openaidevs09/23
AI Agent Reached the Wall: Codex Helps Build a Voice Assistant on Raspberry Pi
«From wiring to debugging the command “show this on the wall”: @sidrampally explains how he used Codex to turn an LED panel into a voice ass…» — OpenAI Developers
▶«From wiring to debugging the command “show this on the wall”: @sidrampally explains how he used Codex to turn an LED panel into a voice ass…» — OpenAI Developers

From a tweet@boltdotnew09/23
Bolt chooses speed for vibe coding: 80% of prompts went to two Flash models
«A week of open models in Bolt Forge. The picture is obvious. GLM 5.3 Flash: 63% DeepSeek V4.1 Flash: 17% Two Flash models. 80% of all pro…» — bolt.new

«A week of open models in Bolt Forge. The picture is obvious. GLM 5.3 Flash: 63% DeepSeek V4.1 Flash: 17% Two Flash models. 80% of all pro…» — bolt.new
From a tweet@heygen09/23
Wake-up video: HeyGen promises Avatar videos after one message
«Yesterday we said: Muse will be able to create your content while you sleep. Here is the full breakdown: a two-minute HeyGen setup, a precise prompt for 7 AM, and one message after which your avatar video will be released in multiple languages. https://x.com/i/article/2102813334303326208»
HeyGen
«Yesterday we said: Muse will be able to create your content while you sleep. Here is the full breakdown: a two-minute HeyGen setup, a precise prompt for 7 AM, and one message after which your avatar video will be released in multiple languages. https://x.com/i/article/2102813334303326208»
HeyGen

From a tweet@nickadobos09/23
Codex Connects to Stack Overflow: Plugin Checks Agent Answers
«ChatGPT Codex has a new Stack Overflow plugin, lol.» — Nick Dobos

«ChatGPT Codex has a new Stack Overflow plugin, lol.» — Nick Dobos
▶
From a tweet@artificialanlys09/23
Google voice generation joins the leaders: Gemini 3.8 Flash TTS ranks No. 2 in Arenastory · 2
«Google released Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS. Gemini 3.8 Flash TTS immediately took first place in our Pronunciation…» — Artificial Analysis
▶«Google released Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS. Gemini 3.8 Flash TTS immediately took first place in our Pronunciation…» — Artificial Analysis

From an announcementGoogle DeepMind09/23
Personal AI will be able to retain context longer: Google adds server-side memory


From a tweet@cursor_ai09/23
Cursor makes agent work cheaper: token costs cut by 7%
«We reduced token costs in Cursor by 7% without lowering agent quality. The savings came from more precise prompts, selective tool loading,…» — Cursor

«We reduced token costs in Cursor by 7% without lowering agent quality. The savings came from more precise prompts, selective tool loading,…» — Cursor
▶
From a tweet@higgsfield_ai09/23
Web animation can be handed off to an AI agent: Higgsfield built a water cycle in one HTML file
«We vibe-coded an entire animated world with Claude Opus 5.5 and GPT-6 Sol. Task: “Create a seamless looping animation of the water cycle,…» — Higgsfield AI 🧩
▶«We vibe-coded an entire animated world with Claude Opus 5.5 and GPT-6 Sol. Task: “Create a seamless looping animation of the water cycle,…» — Higgsfield AI 🧩

From an articletechcrunch.com09/23
YouTube feed now listens to your prompt: Gemini powers Custom Feeds

▶
From a tweet@higgsfield_ai09/23
Game builds in Claude can be assigned to Opus 5.5: Higgsfield built a flight simulator
«Claude Opus 5.5 versus GPT-6 Sol in flight simulator development. A realistic flight simulator built with Higgsfield.» — Higgsfield AI 🧩
▶«Claude Opus 5.5 versus GPT-6 Sol in flight simulator development. A realistic flight simulator built with Higgsfield.» — Higgsfield AI 🧩