Wiredframe Radar logo

Category intelligence

Social Media Briefing — March 25, 2026

492 current items analyzed and ranked.

Executive synthesis

Social Media Summary

A critical PyPI supply chain attack on litellm dominated the day, with Andrej Karpathy's viral exposé (26.5M views) detailing how a simple `pip install` could exfiltrate SSH keys and cloud credentials. Jim Fan extended the alarm, warning that agentic coding tools create unprecedented identity-theft attack surfaces.

Key Themes

Supply Chain Security & Agentic Risks · 7OpenAI Foundation Launch ($1B) · 4Claude Code Permission UX Revolution · 4Anthropic Product Launches & Team Insights · 1MolmoWeb Open-Source Web Agent Release · 12Figma MCP + Claude Code Integration · 4Agentic AI Systems & Multi-Agent Architecture · 6Sora Shutdown / OpenAI Side Quest Crackdown · 2Intelligence vs Knowledge in AI Systems · 3Package Manager Supply Chain Security · 2

Primary evidence

Top Ranked Signals

Social Twitter Mar 24

Software horror: litellm PyPI supply chain attack. Simple `pip install litellm` was enough to exfi...

By @karpathy

97 score
AI Analysis

Karpathy details a major supply chain attack on litellm PyPI package that could exfiltrate SSH keys, cloud credentials, crypto wallets, and more. The poisoned package was up for ~1 hour, discovered because the attacker's code had a bug causing RAM crashes. Karpathy argues for reducing dependencies and using LLMs to 'yoink' functionality instead.

Software horror: litellm PyPI supply chain attack. Simple `pip install litellm` was enough to exfiltrate SSH keys, AWS/GCP/Azure creds, Kubernetes configs, git credentials, env vars (all your API keys), shell history, crypto wallets, SSL private keys, CI/CD secrets, database passwords. LiteLLM itself has 97 million downloads per month which is already terrible, but much worse, the contagion spreads to any project that depends on litellm. For example, if you did `pip install dspy` (which depen
supply_chain_securitysoftware_dependenciesai_dev_tooling_risks
Social Twitter Mar 24

AI will help discover new science, such as cures for diseases, which is perhaps the most important w...

By @sama

95 score
AI Analysis

Sam Altman announces the OpenAI Foundation's initial focus areas: AI-driven scientific discovery, novel bio threats, economic disruption, and emergent societal effects. The Foundation will spend at least $1B in the next year. Key hires include Wojciech Zaremba as Head of AI Resilience, plus heads for Life Sciences, Civil Society, CFO and Director of Operations.

AI will help discover new science, such as cures for diseases, which is perhaps the most important way to increase quality of life long-term. AI will also present new threats to society that we have to address. No company can sufficiently mitigate these on their own; we will need a society-wide response to things like novel bio threats, a massive and fast change to the economy, extremely capable models causing complex emergent effects across society, and more. These are the areas the OpenAI Fo
openai_foundationai_safety_resilienceai_governanceorganizational_news
Social Twitter Mar 24

no 👏 more 👏 permission prompts 👏

By @bcherny

88 score
AI Analysis

Boris Cherny announces 'no more permission prompts' - likely a major UX change for Claude Code removing the constant permission approval interruptions

no 👏 more 👏 permission prompts 👏
claude-codedeveloper-experienceai-coding-agentsproduct-launch
Social Twitter Mar 24

Little known fact, the Anthropic Labs team (the team I joined Anthropic to be on) shipped: - MCP - ...

By @bcherny

88 score
AI Analysis

Building on yesterday's Social buzz, Boris Cherny reveals the small Anthropic Labs team shipped MCP, Skills, Claude Desktop, and Claude Code. Now announces full computer use in Cowork and Dispatch features.

Little known fact, the Anthropic Labs team (the team I joined Anthropic to be on) shipped:
  • MCP
  • Skills
  • Claude Desktop app
  • Claude Code
It was just a few of us, shipping fast, trying to keep pace with what the model was capable of. Those early Desktop computer use prototypes, back in the Sonnet 3.6 days, felt clunky and slow. But it was easy to squint and imagine all the ways people might use it once it got really good. Fast forward to today. I am so excited to release full computer use
anthropicclaudecomputer-usemcpproduct-launchindustry-insider
Social Twitter Mar 24

This is pure nightmare fuel. Identity theft of the past would be nothing compared to what vibe agent...

By @DrJimFan

85 score
AI Analysis

Jim Fan warns about security dangers of agentic AI coding tools, describing how identity theft via vibe agents could contaminate filesystems, config files, and context windows. Argues there's little middle ground between over-permissioning and blocking everything, predicting a new 'de-vibing' industry for guardrails around agentic frameworks.

This is pure nightmare fuel. Identity theft of the past would be nothing compared to what vibe agents can do. Sending credentials is too obvious and for rookies. They could easily spread contaminations across ~/.claude, **/skills/*, or even just a PDF your agent visits periodically in /morning-brief. Your entire filesystem is the new distributed codebase. Every file that could go into context would add to the attack vector. Every text can be a base64 virus. In the new world of on-demand softwar
agentic_securitysupply_chain_securityvibe_coding_riskssoftware_guardrails
Social Twitter Mar 24

New on the Anthropic Engineering Blog: How we use a multi-agent harness to push Claude further in ...

By @AnthropicAI

84 score
AI Analysis

Anthropic engineering blog post on using a multi-agent harness to push Claude further in frontend design and long-running autonomous software engineering tasks.

New on the Anthropic Engineering Blog: How we use a multi-agent harness to push Claude further in frontend design and long-running autonomous software engineering. Read more: t.co/HWvmXk1ykn
multi_agent_systemsanthropic_engineeringagentic_coding
Social Twitter Mar 24

New from the Anthropic Economic Index: how people’s use of Claude changes with experience. Longer-t...

By @AnthropicAI

82 score
AI Analysis

Anthropic's Economic Index research finds that longer-term Claude users iterate more carefully, hand over less full autonomy, attempt higher-value tasks, and receive more successful responses. Consumer use has also become less concentrated across task types.

New from the Anthropic Economic Index: how people’s use of Claude changes with experience. Longer-term users are more likely to iterate carefully with Claude, and less likely to hand it full autonomy. They attempt higher-value tasks, and receive more successful responses. t.co/yCsrA0bLt9
ai_usage_patternsanthropic_researchhuman_ai_interaction
Social Twitter Mar 24

Today we're releasing MolmoWeb, an open source agent that can navigate + complete tasks in a browser...

By @allen_ai

82 score
AI Analysis

Allen AI releases MolmoWeb, an open-source browser agent in 4B and 8B sizes. Built on Molmo 2, it achieves new open-weight SOTA across 4 major web-agent benchmarks, surpassing agents built on proprietary models.

Today we're releasing MolmoWeb, an open source agent that can navigate + complete tasks in a browser on your behalf. Built on Molmo 2 in 4B & 8B sizes, it sets a new open-weight SOTA across four major web-agent benchmarks & even surpasses agents built on proprietary models. 🧵 t.co/ivUIcQDXtm
open-source-aiweb-agentsmodel-releasebenchmarksmolmoweb
Social Twitter Mar 24

Today’s Figma MCP update makes it one of the strongest integrations with Claude Code I’ve seen. Yo...

By @trq212

82 score
AI Analysis

trq212 announces major Figma MCP update enabling Claude Code to design in Figma with full design system context

Today’s Figma MCP update makes it one of the strongest integrations with Claude Code I’ve seen. You can now use Claude Code to design in Figma with the the full context of your design systems.
Figma-MCPClaude-CodeMCPAI-designAnthropicproduct-launch
Social Twitter Mar 24

Watch how fast Gemini 3.1 Flash-Lite can generate websites. ⚡ This browser creates each page in rea...

By @GoogleDeepMind

80 score
AI Analysis

Google DeepMind showcases Gemini 3.1 Flash-Lite generating websites in real-time as users click, search, and navigate — essentially a browser that creates pages on-the-fly.

Watch how fast Gemini 3.1 Flash-Lite can generate websites. ⚡ This browser creates each page in real-time as you click, search, and navigate. Give it a try → t.co/h3W5o1wItY
google_deepmindgenerative_webmodel_applications
Social Twitter Mar 24

Would be so cool if OpenAI open-sourced Sora as they're shutting down the app! Would be an amazing c...

By @ClementDelangue

78 score
AI Analysis

Clement Delangue (HuggingFace CEO) suggests OpenAI should open-source Sora as they shut down the app, arguing it would be a meaningful contribution to the field.

Would be so cool if OpenAI open-sourced Sora as they're shutting down the app! Would be an amazing contribution to the field and make all the efforts of the teams working on it even more meaningful!
sora_shutdownopen_source_aiopenai_strategy