X

@cursor_ai

DeepSeek V4.1 Flash in Cursor: announces DeepSeek V4.1 Flash is available in Cursor, saying every effort level sits on the CursorBench cost frontier at $0.60 to $0.98 per task; enabled in Cursor Settings > Models. post (https://x.com/cursor_ai/status/2114839997568405608) · 3,116❤ · 162 RT

Agent SDK: announces the Cursor Agent SDK is available, defining agents as code with prompts in Markdown and tools in TypeScript, deployable on Cursor's hosted infrastructure or self-hosted, with built-in evals to measure each change. post (https://x.com/cursor_ai/status/2114793714531102901) · 3,366❤ · 235 RT

@poteto

Hire Grok Bot: quotes a post describing an hourly Grok Bot routine that checks email, DMs, mentions, and LinkedIn and drafts replies, with "hire grok bot!". post (https://x.com/poteto/status/2114818880548041189) · 716❤ · 23 RT

Shipping this week: asks what people will ship this week with Grok Bot. post (https://x.com/poteto/status/2114772732944728512) · 1,169❤ · 19 RT

@swyx

Anthropic valuation poll: asks whether people would buy $ANT stock at a $2T valuation. post (https://x.com/swyx/status/2114864999709753526) · 18❤ · 0 RT

AIE Code keynote: quotes a post about sharing a WoW shaman build at the AIE Code closing keynote, joking about moving from Claude Code to AIE Code. post (https://x.com/swyx/status/2114805687226376267) · 185❤ · 3 RT

@rauchg

cpu-chat: built a reproduction of OpenAI's findings on covert CPU communication across Firecracker microVMs with no internet access, chatting via CPU contention, cache eviction timing, and DRAM bandwidth contention despite Firecracker's production recommendations; says this differs from side-channel data theft, is harder to pull off and easier to detect in Vercel's environment, and that OpenAI and Anthropic models refused the work while open-weight models completed it. post (https://x.com/rauchg/status/2114770919436808572) · 2,186❤ · 198 RT

@HamelHusain

Benchmark review pushback: quotes Epoch AI's Benchmark Reviews launch (15 benchmarks: 4 Verified, 9 Flawed, 2 insufficient information) and rebuts common pushback, arguing benchmarks drive large token spend and are trusted blindly, listing checks for failing tasks, model configuration, harness realism, and scoring, and saying this motivates investing in private evals. post (https://x.com/HamelHusain/status/2114935968604299552) · 66❤ · 1 RT

Silent

@martinfowler, @karpathy, @hwchase17, @simonw: no original posts in window

Gaps

@xAI: account shows 0 posts, bio "Now at @SpaceXAI" (skipped)

GitHub trending

Nutlope/hallmark (https://github.com/Nutlope/hallmark): anti-AI-slop design skill for Claude Code, Cursor, and Codex (CSS, +2,244★ today)
TauricResearch/TradingAgents (https://github.com/TauricResearch/TradingAgents): multi-agent LLM financial trading framework (Python, +1,395★ today)
JetBrains/thinkrail (https://github.com/JetBrains/thinkrail): lightweight IDE for vibe coding with pi that customizes itself around how you work (TypeScript, +1,173★ today)
makecindy/cindy (https://github.com/makecindy/cindy): open-source AI agent that works out of the box (TypeScript, +1,296★ today)
affaan-m/ECC (https://github.com/affaan-m/ECC): agent harness performance optimization system with skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor, and others (JavaScript, +896★ today)