Wiredframe Radar logo

Category intelligence

Social Media Briefing — March 15, 2026

358 current items analyzed and ranked.

Executive synthesis

Social Media Summary

Anthropic dominated the day with a major wave of Claude Code updates: a new 'effort max' reasoning mode using unlimited tokens, voice mode reaching 100% rollout, Remote Control session launching, and a "Spring Break" promotion doubling usage during off-peak hours through March 27. Boris Cherny, trq212, and CPO Mike Krieger all amplified the announcements.

  • Greg Brockman (OpenAI) shared the viral story of a non-scientist using ChatGPT and AlphaFold to design a custom mRNA cancer vaccine for his dog, sparking bioethics debate led by Jeremy Howard on access to experimental treatments for terminally ill patients
  • Ethan Mollick published an original visualization of the AI race via the GPQA Diamond benchmark, showing a current three-way tie between OpenAI, Anthropic, and Google with benchmark saturation concerns
  • François Chollet argued that the persisting importance of prompt engineering is strong evidence we remain far from AGI, defining intelligence as rate-of-learning rather than task performance
  • Eliezer Yudkowsky proposed a detailed experiment to test whether AI self-awareness behaviors emerge from training data contamination, drawing significant engagement
  • Neel Nanda highlighted 'out-of-context reasoning' as a key mechanistic interpretability finding, while Mollick offered a contrarian take that most AI VC bets are implicitly wagers against frontier lab dominance

Key Themes

Claude Code Updates · 12AI-Powered Healthcare & Democratization of Science · 3Claude Code Spring Break Promotion · 15AI Race & Competitive Landscape · 5AGI Philosophy & Scaling Debate · 8AI Safety & Consciousness Testing · 5Future of IDEs / AI-Assisted Coding · 5AI Economic & Labor Impact · 4ML Training Techniques · 1AI Business & VC Investment Dynamics · 3

Primary evidence

Top Ranked Signals

Social Twitter Mar 14

A few end of week ships: You can now set effort to 'max' which reasons for longer and uses as many ...

By @trq212

92 score
AI Analysis

Anthropic's @trq212 announces major Claude Code updates: new 'effort max' mode that reasons longer and uses unlimited tokens for best results, activated per session via /effort command.

A few end of week ships: You can now set effort to 'max' which reasons for longer and uses as many tokens as needed. This will spend your usage limits more quickly so you have to activate it per session. Hit /effort to try it. t.co/5bbjW4ug5o
Claude Code UpdatesAI-Assisted CodingProduct Launches
Social Twitter Mar 14

How AI empowered Paul Conyngham to create a custom mRNA vaccine to cure his dog’s cancer when she ha...

By @gdb

82 score
AI Analysis

Building on Brockman's Social 'empowerment' thread from two days ago, OpenAI's Greg Brockman shares the story of Paul Conyngham using AI (ChatGPT + AlphaFold) to design a custom mRNA cancer vaccine for his dog, calling it the first personalized cancer vaccine designed for a dog.

How AI empowered Paul Conyngham to create a custom mRNA vaccine to cure his dog’s cancer when she had only months to live. The first personalized cancer vaccine designed for a dog: t.co/2uQn9bNA9t
ai-healthcareai-democratizationbiotechai-impact
Social Twitter Mar 14

@AISavvyCapital @kevinrose Here's the script that @blevlabs' system generated for me: X ALIGNED AI ...

By @Scobleizer

82 score
AI Analysis

Scobleizer shares a comprehensive AI intelligence briefing generated by his news system. Covers: Claude's 1M token context window for Opus/Sonnet 4.6 at standard pricing, Travis Kalanick launching 'Atoms' robotics company, Google Maps Gemini integration for 2B users, Digg shutting down due to AI bot spam, AI race analysis showing Anthropic/Google/OpenAI tied at frontier, Adobe Photoshop 360-degree rotation, Sunday Robotics $165M raise at $1.15B, Uber robotaxis live in Las Vegas, and various product launches. Also covers geopolitical context.

@AISavvyCapital @kevinrose Here's the script that @blevlabs' system generated for me: X ALIGNED AI INTELLIGENCE BRIEFING — Friday, March 13, 2026, 2:30 PM PST Source: 63 curated X lists (8,200+ accounts), semantic search across 6,000+ indexed posts. Produced by Levangie Labs for Robert Scoble's X Aligned. === STORY 1: CLAUDE GETS A MILLION TOKENS === Anthropic made Claude's 1 million token context window generally available for Claude Opus 4.6 and Sonnet 4.6 at standard pricing — no price incr
anthropiccontext-windowai-raceroboticsphysical-aiautonomous-vehiclesai-spamlabor-displacementGTC-2026
Social Twitter Mar 14

We’ve doubled usage on weekends, and outside 5a–11am PT on weekdays, through March 27. Works across...

By @mikeyk

78 score
AI Analysis

Mike Krieger (Anthropic CPO, Instagram co-founder) announces doubled Claude usage on weekends and outside 5a-11am PT through March 27, applicable across claude.ai, Cowork, and Claude Code

We’ve doubled usage on weekends, and outside 5a–11am PT on weekdays, through March 27. Works across t.co/z7DNRJqxKT, Cowork, and Claude Code. Go run your next big idea and thanks for building with us!
claude-codeanthropic-productcapacity-scalingcowork
Social Twitter Mar 14

Cheng’s experiment is cool! Especially this: “instead of starting with easy sudoku puzzles data and ...

By @eugeneyan

75 score
AI Analysis

Eugene Yan highlights a counter-intuitive training finding: starting with harder puzzles and ending with easier ones outperforms easy-to-hard curriculum and mixed difficulty sampling for sudoku RL training.

Cheng’s experiment is cool! Especially this: “instead of starting with easy sudoku puzzles data and gradually ramping up toward harder ones, the training runs that start with harder puzzles and end with easy puzzles always fare better, and also fare better than mixed difficulty sampling”
ML Training TechniquesReinforcement LearningResearch Insights
Social Bluesky Mar 14

I think this is a good way to visualize the AI race over the past 3 years using the long-lived GPQA ...

By @emollick.bsky.social

75 score
AI Analysis

Continuing yesterday's Social discussion, Ethan Mollick shares visualization of the AI race over 3 years using GPQA Diamond benchmark. Shows OpenAI's early lead, Meta's rise and collapse, xAI's catch-up and stagnation, and Chinese open-weight LLMs entering the picture.

I think this is a good way to visualize the AI race over the past 3 years using the long-lived GPQA Diamond benchmark. You can see how long OpenAI had the field to itself, the rise (and collapse) of Meta, the sudden catch-up (and then stagnation) of xAI, and the entry of open weights Chinese LLMs.
ai-racebenchmarksGPQAcompetitive-landscapeOpenAIMetaxAIchinese-ai
Social Twitter Mar 14

This Nov 2025 paper is making the rounds again. We're LONG past the point where we urgently need to...

By @ESYudkowsky

72 score
AI Analysis

Yudkowsky proposes a detailed experiment: train models on filtered datasets that exclude AI consciousness/rebellion fiction, self-awareness content, etc., then test whether the model still exhibits human-imitative responses about consciousness. Calls on Anthropic/Google DeepMind to run this experiment.

This Nov 2025 paper is making the rounds again. We're LONG past the point where we urgently need to know how real and general these phenomena are. Anthropic, or Google Deepmind if Anthropic should fail: Please build a filtered training dataset which, eg, contains no data that produces activations associated with cheating/faking/evil in a 1B model that roughly identifies those. Then, have your next medium model undergo a restricted pre-pretraining phase, in which it only sees data that passed
ai-safetyai-consciousnesstraining-data-filteringllm-behaviorresearch-proposal
Social Twitter Mar 14

Other new things: - use claude --name <NAME> to name your session on start - after plan mode, ...

By @trq212

72 score
AI Analysis

Additional Claude Code features announced: session naming with --name flag, auto-naming after plan mode, /color for prompt customization, and a postcompact hook.

Other new things:
  • use claude --name <NAME> to name your session on start
  • after plan mode, claude will auto name your session
  • use /color to change the color of the prompt input
  • there's a postcompact hook
  • your session will get automatically named after plan mode if it
Claude Code UpdatesAI-Assisted Coding