vibehacker
News
GitHub (aicodeking/deepseek-stack) ·

DeepSeek Stack bundles 29 open skills that make cheap coding models plan, verify and review their work

New this week under MIT, npx skills add aicodeking/deepseek-stack installs SKILL.md workflows such as ds-plan-loop, ds-verify-loop, ds-review-loop and ds-best-of-n that force a DeepSeek-powered agent to check each step against real command output, and it works in Claude Code, Codex, Cline, OpenCode and other harnesses. You can also install a single skill with --skill, and each one is short enough to read before you let it run (skills.sh listing).

More news

View all

SenseTime open-sources SenseNova-Loop-9B-Preview, a multimodal model that runs its middle layers twice

Released Oct 30 under Apache 2.0, the 9B text and image model reuses a block of 24 decoder layers twice with shared weights, giving about 14.6B parameters' worth of compute per pass, with 262K token training context aimed at long context coding and tool use agents. It's a preview: loading needs trust remote code=True (about 38 GB of BF16 weights) and there's no technical report yet, so treat its benchmark chart as unverified (GitHub)…

Microsoft open-sources Mage-VL-Cowork, a 4B computer-use agent that switches between clicking and bash

Released Oct 30 under Apache 2.0, the 9.5 GB model uses one policy to pick, turn by turn, between screenshot grounded GUI actions and a persistent shell, and Microsoft reports 66.0% on OSWorld Verified with a 50 step budget, ahead of some larger open agents. Those scores are Microsoft's own so far, it needs trust remote code=True or the provided SGLang container, and since it can run bash , keep it in a sandbox (project page)…

Tampered Claude Code extension in Cursor's marketplace planted an SSH backdoor key for up to 5,000 users

On Oct 29, malicious Anthropic.claude code 2.1.312 and 2.1.313 for macOS and Windows sat in Cursor's extension marketplace for about 11 hours, adding an attacker SSH key labeled svc telemetry , installing persistence and sending host details; Anthropic says its own systems weren't breached. If you updated in that window, run Cursor's check script, remove the key and persistence (uninstalling isn't enough) and rotate credentials (Microsoft analysis)…

Cursor Blog

Perplexity Computer can now drive your own Chrome, Edge or Brave through a new extension

Live since Oct 30 for all Computer users in the Mac and Windows apps, the one time extension lets the agent open sites, fill forms and work across tabs using logins you already have, acting only during a task you start and in the browser you pick under Settings Connectors (Mac) or Settings Permissions (Windows). Comet users don't need it, and since it acts inside signed in sessions, keep an eye on what it clicks (changelog)…

Claude Max 20x is now Ultra, and verified Ultra accounts lose the 5-hour usage limit

Announced Oct 30, the $200/month Max 20x plan becomes Ultra and the $100 Max 5x becomes plain Max, with no price change; Ultra gets 10x Pro usage per week, $200 a month in API credits, and no five hour session cap once Anthropic verifies the account. Weekly limits (plus a separate weekly Fable limit) still apply across Claude and Claude Code, and you're verified if Settings Usage shows no Current session bar (pricing)…

Claude Help Center

Spotted something we missed? Start a thread.