On X

On — Nov 7

  • 3-minute read
  • 7 topics
  • 21 posts

Coding-agent tooling leads the day: Claude Duet picks up Claude Code sessions, Google Antigravity ships remote agents and resets quotas, and the Claude Code email leak keeps drawing replies. Qwen's teaser draws the most attention, while the next frontier releases from OpenAI and Anthropic stay unannounced.

Read the Nov 7 brief →

Qwen 4 teaser X-led#

Qwen posts an image captioned "The becoming begins." No model card, sizes, or benchmarks are out yet. Posts point to Qwen4 model IDs already in the qwen-code repo; the expected Max, Plus, Flash, and 27B lineup is unconfirmed.

Why it spiked. People expect a new open model family within days, and a 27B size could run on a single 24GB GPU.

  • @Alibaba_Qwen Official teaser post: "The becoming begins."
  • @EamonMcKiernan5 Notes the teaser has no card, sizes, or benchmark table, and points to a merged qwen-code PR listing qwen4, qwen4-max, and qwen4-plus.
  • @Layton_Gott Explains the image as the opening of Zhuangzi's fish-to-bird story and lists the expected Qwen4-Max and Qwen4-Plus variants.

Claude Code put user emails in User-Agent headers#

A GitHub issue reports that Claude Code included the signed-in account email in the User-Agent header of web and curl requests. Claude Code 2.1.317 adds a privacy plugin that blocks it; the issue is still open.

Why it spiked. The reporter says the leaked address now gets dozens of scam emails a day.

  • @CR1337 Quotes the report: Claude Code "divulged my email to arbitrary 3rd parties in the User-Agent request header while conducting web searches."
  • @maikbuilds Says a Claude Code engineer explained on Hacker News that the email went into model context to stop hallucinated author tags.
  • @theDrewDag Issue #107757 is still open; the privacy plugin needs 2.1.317+ and its README calls it "a courtesy, not a boundary."

Claude Duet connects to Claude Code sessions X-led#

Anthropic's Claude Duet app now shows Claude Code sessions running locally or in the cloud. From a phone you can check progress, approve changes, and send updates.

Why it spiked. Users can approve agent actions without being at their computer.

  • @davidsuttonca Announces Claude Code sessions in Claude Duet: "Check what's running, approve changes and send updates - all from your phone."
  • @navbryce Claude Duet engineer; the team also shipped an in-app view of the cloud browser.
  • @daniel_mac8 Says cloud and local session access makes Duet his main personal agent.

Google Antigravity: remote agents, voice mode, quota reset X-led#

Antigravity 2.29 runs remote agents in cloud VMs on paid consumer plans and adds Gemini Live voice mode for local sessions. Varun Mohan resets everyone's Gemini and Argon quota for the weekend.

Why it spiked. Users get fresh quota to try the new cloud agents right away.

  • @_mohansolo "Resetting everyone's Gemini and Argon quota on Antigravity… Lots more to come next week!"
  • @antigravity Remote agents run tasks in independent cloud VMs; available on all paid consumer plans.
  • @antigravity Voice mode uses Gemini Live to plan, steer code changes, and approve permissions by voice.

Linear on shipping with cloud agents X-led#

Linear CEO Karri Saarinen says most of Linear's code (about 5M lines last quarter) is now written by its cloud coding agent. He argues the gap between teams comes from ambition, not tooling.

Why it spiked. A company publishes its own before-and-after output numbers from agent use.

  • @karrisaarinen Lists practices: standardize on cloud agents, auto-investigate bugs from triage, let PMs and designers submit PRs, run security as an automation.
  • @sthnavy Reads the chart: lines changed 1.37M to 5.82M (4.2x), merged PRs 4,964 to 11,577 (2.3x).
  • @amranihicham5 Notes the average PR nearly doubled in size, so review becomes the bottleneck.

T3 Code and the agent-app market after Conductor X-led#

Chatter: after Conductor's sale to Meta, developers debate whether standalone agent coding apps can last. Theo says T3 Code "already won."

Why it spiked. Conductor Cloud users are choosing a replacement before it shuts down.

  • @theo Replies to an "ADEs are a tarpit" post: "T3 Code already won."
  • @JSystemsX Disagrees; uses Orca, Herdr, or self-built tools more than T3 Code.
  • @ryanshipswell Calls T3 Code the closest thing to an ideal agent environment he has found.

GPT-6.2 Astra and Fable 5.5 still unreleased X-led#

Rumor: earlier leaks pointed to GPT-6.2 Astra and Claude Fable 5.5 this week, but neither OpenAI nor Anthropic has released or announced them. Posts joke that each lab waits for the other.

Why it spiked. Developers expected new frontier models this week and none arrived.

  • @StatsWire "GPT-6.2 Astra is waiting for Claude Fable 5.5. Claude Fable 5.5 is waiting for GPT-6.2 Astra."
  • @rezoundous "OpenAI doesn't want to release GPT-6.2 Astra before Fable 5.5… Seems like we're at an impasse."
  • @nimi_101 Suggests Fable 5.5 timing depends on Anthropic's IPO.