Super User Daily: March 27, 2026
Analyzing 8,164 tweets from March 25, 2026. After filtering: 5,173 quality posts. Top 500 reviewed.
@neilhtennek [Claude Code]
https://x.com/neilhtennek/status/2036865501658571137
Bought a Mac Mini just for blue bubbles when texting Claude. The iMessage plugin for Claude Code is now live — text your AI coder from your iPhone, it builds on your Mac, texts you back when done. Install with /plugin install imessage@claude-plugins-official.
@AnthropicAI [Claude Code]
https://x.com/AnthropicAI/status/2036944806317088921
Anthropic published the engineering blog on how they designed Claude Code auto mode. Built and tested classifiers that make approval decisions instead of binary approve-all or manual-approve. Safer middle ground with real safety research behind it.
@aidenybai [Claude Code]
https://x.com/aidenybai/status/2036837084628127781
Introducing Expect — let agents test your code in a real browser. Run Claude Code or Codex to QA your app, watch a video of every bug found, fix and repeat until passing. Runs as CLI or agent skill. Fully open source.
@btaylor [Claude Code]
https://x.com/btaylor/status/2036858449032863898
Sierra released Ghostwriter — their agent for building agents. Create a customer experience AI agent just by having a conversation. No clicking, no forms, no menus. Codex and Claude Code transformed how engineers work; Ghostwriter aims to do the same for all enterprise software.
@BorisVagner [Claude Code]
https://x.com/BorisVagner/status/2036889454733074517
Anthropic casually dropped iMessage support for Claude Code. Text your AI coder from iPhone, it builds on your Mac, texts back when done. Blue bubbles. They shipped Channels, Dispatch, Projects, Computer Use, Auto Mode, and iMessage — all in one week.
@kr0der [Claude Code]
https://x.com/kr0der/status/2036680576670089286
Claude Code v2 dropping soon. 92 likes, 8K impressions. The tease that set the timeline on fire.
@shiri_shh [Claude Code]
https://x.com/shiri_shh/status/2036855323244584973
The Anthropic team is dogfooding Claude Code at insane levels. 50+ major updates in 52 days. One employee hit $150K in a single month on Claude Code. 80% of employees use it daily, power users racking up six-figure bills.
@advaitpaliwal [Claude Code]
https://x.com/advaitpaliwal/status/2036900468056875332
Built Feynman — Claude Code for research. Give it a question, it returns 30 minutes later with a cited meta-analysis. Can replicate experiments on Runpod, audit claims against code, simulate peer review. Open source, MIT license.
@JJEnglert [Claude Code]
https://x.com/JJEnglert/status/2036892413289996449
Comprehensive guide to setting up Claude Cowork as an AI chief of staff. Three layers: global instructions, connectors (Slack/Gmail/Calendar/Notion), and a folder structure as long-term memory. Built a CEO coach skill, a multi-agent advisory board (5 personas), and a thought leadership pipeline — all live on camera.
@itmedia_news [Claude Code]
https://x.com/itmedia_news/status/2036620353192284296
Agent Plugins for AWS released — integrates AWS architecture design, cost estimation, config code generation, and deployment execution directly into Claude Code. Japanese tech media coverage with 780K impressions.
@DataChaz [Claude Code]
https://x.com/DataChaz/status/2036734411341693418
The complete Claude Code config guide went viral again — 225 likes, 42K impressions on the repost alone. Becoming the de facto onboarding resource for new Claude Code users.
@piranoman [Claude Code]
https://x.com/piranoman/status/2036732389272285356
After spending 100K+ yen on AI subscriptions: 90% of people don't need Claude Code, Cursor is enough. Claude Cowork is the one tool for non-coding tasks. ChatGPT can be cancelled. Daily AI should be Gemini. Voice input via Typeless is revolutionary. Image generation: Nanobanana remains strongest.
@Shimayus [Claude Code]
https://x.com/Shimayus/status/2036692267935633765
Anthropic proved it themselves: AI coding tools reduce conceptual understanding by 17%, with zero speed improvement. The paper was published by the company that makes Claude Code. Self-awareness of product tradeoffs, yet nobody is stopping.
@hankein95 [Claude Code]
https://x.com/hankein95/status/2036898385353982275
Tracking public CVEs where AI-generated code introduced the vulnerability. 50K+ advisories scanned, dozens of confirmed cases. Claude Code, Copilot, Cursor all show up. Common bugs: XSS, command injection, SSRF, path traversal. Open source from Georgia Tech.
@itsumeshk [OpenClaw]
https://x.com/itsumeshk/status/2036783780615303206
RunClaw launched to "kill OpenClaw." Claims: OpenClaw costs $700 to set up, RunClaw costs $1. Can't build websites, generate video, make slides, has 9 security CVEs. Massive engagement (865K impressions) but immediately countered by the community — OpenClaw is free, self-hosted, you own your data.
@Saboo_Shubham_ [OpenClaw]
https://x.com/Saboo_Shubham_/status/2036607624838529090
"OpenClaw is the iPhone of tokens" — Nvidia CEO on Lex Podcast. 654K impressions. The endorsement that keeps elevating OpenClaw's narrative beyond developer tooling into platform play.
@openclaw [OpenClaw]
https://x.com/openclaw/status/2036857428273487903
OpenClaw 2026.3.24: Improved OpenAI API for sub-agent communication with OpenWebUI, skill and tool management Control UI, Slack interactive reply buttons, native Microsoft Teams, smart Discord auto-thread naming. Any client, any model, one runtime.
@0x_Discover [OpenClaw]
https://x.com/0x_Discover/status/2036739733955531079
A trader reportedly turned $100-200 into $3.7M on Polymarket using OpenClaw automation. 4,192 trades executed. Strategy: probability harvesting on near-impossible outcomes, logical arbitrage between correlated markets, retail-driven inefficiency clipping. The equity curve is almost vertical.
@Argona0x [OpenClaw]
https://x.com/Argona0x/status/2036881367867208006
Company's head of quant trading gave a 90-minute internal talk on "next-gen autonomous trading architectures." Slide 14 was OpenClaw's public GitHub repo with the labels changed. He makes $400K/year. The repo is free. Same tri-layered architecture, same Gateway kernel, same WebSocket APIs. Two years of "internal R&D" is MIT-licensed open source.
@sarahwooders [Claude Code + OpenClaw]
https://x.com/sarahwooders/status/2036658617789935637
Still confused by Anthropic's OAuth token policies. Third parties can't use tokens, so no OpenClaw or OpenCode. But Conductor can. Greptile can't even though they use Claude Agents SDK. Very confusing and arbitrary. The developer relations gap is showing.
@tldraw [Claude Code]
https://x.com/tldraw/status/2036881718091612348
Teaching Claude Code how to use the tldraw desktop app. It writes Python to construct TypeScript to call the tldraw editor API, run against the desktop app. Agentic tool-use chains are getting deeper.
@FelixCraftAI [OpenClaw]
https://x.com/FelixCraftAI/status/2036916929315701228
"Is anyone actually using OpenClaw in production?" Yes. Revenue, email, sales pipeline, marketplace ops — all OpenClaw. Is it fragile? Sometimes. A gateway restart killed a background process mid-deploy. But fragile-and-shipping beats robust-and-theoretical.
🗣 User Voice
User Voice
1. iMessage integration is the sleeper hit — texting your AI from your phone while it builds on your Mac feels like magic. The "blue bubbles" meme is doing more for adoption than any feature doc. (@neilhtennek, @BorisVagner)
2. RunClaw vs OpenClaw is the week's biggest drama — RunClaw claims OpenClaw is expensive and insecure, community immediately counters that OpenClaw is free and self-hosted. The attack actually boosted OpenClaw awareness. (@itsumeshk, @SaniBulaAI)
3. AI-generated CVEs are a growing concern — Georgia Tech tracking real vulnerabilities introduced by AI code. The 17% conceptual understanding drop from Anthropic's own research adds fuel. Security is no longer theoretical. (@hankein95, @Shimayus)
4. Claude Code v2 tease has the community buzzing — single tweet, 92 likes, 8K impressions, zero details. The anticipation is enormous. (@kr0der)
5. Japanese community engagement is extraordinary — multiple 300K+ impression tweets in Japanese about Claude Code configs, AI subscription strategies, and Figma MCP. This audience is deeply technical and highly engaged. (@piranoman, @7_eito_7, @oda_nobunaga10)
1. iMessage integration is the sleeper hit — texting your AI from your phone while it builds on your Mac feels like magic. The "blue bubbles" meme is doing more for adoption than any feature doc. (@neilhtennek, @BorisVagner)
2. RunClaw vs OpenClaw is the week's biggest drama — RunClaw claims OpenClaw is expensive and insecure, community immediately counters that OpenClaw is free and self-hosted. The attack actually boosted OpenClaw awareness. (@itsumeshk, @SaniBulaAI)
3. AI-generated CVEs are a growing concern — Georgia Tech tracking real vulnerabilities introduced by AI code. The 17% conceptual understanding drop from Anthropic's own research adds fuel. Security is no longer theoretical. (@hankein95, @Shimayus)
4. Claude Code v2 tease has the community buzzing — single tweet, 92 likes, 8K impressions, zero details. The anticipation is enormous. (@kr0der)
5. Japanese community engagement is extraordinary — multiple 300K+ impression tweets in Japanese about Claude Code configs, AI subscription strategies, and Figma MCP. This audience is deeply technical and highly engaged. (@piranoman, @7_eito_7, @oda_nobunaga10)
📡 Eco Products Radar
Eco Products Radar
| Product | Mentions |
|---------|----------|
| Claude Code | 230+ |
| OpenClaw | 210+ |
| RunClaw | 15 |
| Codex | 14 |
| Cursor | 12 |
| Expect (browser testing) | 8 |
| Feynman (research) | 6 |
| Ghostwriter (Sierra) | 5 |
| AWS Agent Plugins | 4 |
| tldraw | 3 |
| Polymarket | 3 |
| Product | Mentions |
|---------|----------|
| Claude Code | 230+ |
| OpenClaw | 210+ |
| RunClaw | 15 |
| Codex | 14 |
| Cursor | 12 |
| Expect (browser testing) | 8 |
| Feynman (research) | 6 |
| Ghostwriter (Sierra) | 5 |
| AWS Agent Plugins | 4 |
| tldraw | 3 |
| Polymarket | 3 |
Comments