Wiredframe Radar logo

Category intelligence

Social Media Briefing — March 31, 2026

508 current items analyzed and ranked.

Executive synthesis

Social Media Summary

The AI community buzzed around a massive viral thread from Anthropic engineer Boris Cherny revealing 15 hidden Claude Code features—including mobile coding, cross-device teleportation, autonomous scheduling loops, and parallel batch agents—amassing 21K likes and 3.2M views.

François Chollet endorsed PokeeClaw as a secure alternative to OpenClaw, addressing critical agent security concerns. Allie K Miller reacted to a buried Anthropic tweet showcasing Claude's coding capabilities, noting Fortune 500 executives are already rethinking entire engineering teams. Hugging Face CEO called for open-source agent tools to stop depending on closed-source APIs.

Key Themes

Claude Code Power Features · 24AI Economics & Pricing · 5Anthropic Claude Capabilities Breakthrough · 4Recursive Self-Improvement & AI Training Automation · 2Open-Source AI & Local Models · 8Training Optimization & Technical Research · 3AI Productivity & Workforce Impact · 6Claude Code Scaling & Growing Pains · 3AI Agents & Security (OpenClaw/PokeeClaw) · 4AI Productivity and Economic Impact · 3

Primary evidence

Top Ranked Signals

Social Twitter Mar 30

It's my favorite kind of work: linear algebra insight + fast kernels. When playing w Muon a while ag...

By @tri_dao

82 score
AI Analysis

Tri Dao describes work on speeding up Muon optimizer by operating on the small square matrix X X^T instead of the large rectangular X, requiring months of understanding eigenvalues/eigenvectors to develop a simple, elegant algorithm.

It's my favorite kind of work: linear algebra insight + fast kernels. When playing w Muon a while ago, we were thinking why not speed it up by operating on the small square matrix X X^T instead of the large rectangular matrix X. Jack, Noah, and Berlin spent many months understanding eigenvalues/vectors of the intermediate matrices in Muon, and finally came up with a simple and elegant algo to make this work.
Muon optimizertraining efficiencylinear algebraGPU kernelsoptimization
Social Twitter Mar 30

I wanted to share a bunch of my favorite hidden and under-utilized features in Claude Code. I'll foc...

By @bcherny

82 score
AI Analysis

bcherny's main thread intro: sharing hidden and under-utilized features in Claude Code. This is the parent post for a massive 15-tip thread.

I wanted to share a bunch of my favorite hidden and under-utilized features in Claude Code. I'll focus on the ones I use the most. Here goes.
claude-codedeveloper-toolsai-coding-toolsanthropic
Social Twitter Mar 30

So proud to have @ggerganov and @ggml_org part of the @huggingface team. One of the unsung heroes of...

By @ClementDelangue

78 score
AI Analysis

Clement Delangue announces that ggerganov and ggml_org are now part of the Hugging Face team, calling them unsung heroes of AI powering the most widely used open-source runtime for local AI.

So proud to have @ggerganov and @ggml_org part of the @huggingface team. One of the unsung heroes of AI, powering the most widely used open-source runtime for local AI!
open-source AIlocal AIindustry consolidationGGML/llama.cpp
Social Twitter Mar 30

It is trendy to discuss Jevon's Paradox in AI (as AI gets more efficient, overall use increases) but...

By @emollick

76 score
AI Analysis

Emollick argues that discussion of Jevon's Paradox in AI is oversimplified - the real issue is that agent-driven token demand is surging while compute is supply-constrained for powerful models, which will be reflected in pricing.

It is trendy to discuss Jevon's Paradox in AI (as AI gets more efficient, overall use increases) but the current situation is much simpler: thanks to agents, token demand is surging and compute is supply constrained, at least for powerful models. That will be reflected in pricing
AI economicsJevons ParadoxAI agentscompute constraintsAI pricing
Social Twitter Mar 30

This is true, but ARC-AGI-3 is also a test designed so that AI gets zero today, just as the earlier ...

By @emollick

75 score
AI Analysis

Emollick comments on ARC-AGI-3, noting it's designed so AI scores zero today (just like earlier versions), and those earlier tests were mostly saturated within a year or two. The key question is whether we see similar progress on ARC-AGI-3.

This is true, but ARC-AGI-3 is also a test designed so that AI gets zero today, just as the earlier ARC-AGI tests were designed . Those tests were then mostly saturated with a year or two. The thing to watch with ARC-AGI-3 is whether we see the same progress.
ARC-AGI benchmarksAI progressbenchmark saturation
Social Twitter Mar 30

I was told by one researcher at @HF0 in San Francisco that they took "hands off" of their AI trainin...

By @Scobleizer

75 score
AI Analysis

Scobleizer claims a researcher at HF0 told him they took 'hands off' of AI training, letting AI do it all, with training speed increasing many times since. He extrapolates this to OpenClaw improving itself and predicts 'things are about to get weird in AI.'

I was told by one researcher at @HF0 in San Francisco that they took "hands off" of their AI training. Now AI does it all. Has already increased in speed many times, he told me, since doing that. Think about an OpenClaw improving itself all night long. Except on a much bigger scale. Things are about to get weird in AI.
recursive_self_improvementai_training_automationai_accelerationai_safetyhf0
Social Twitter Mar 30

OpenClaw has proven that local AI assistants have product-market fit. But the big issue with them ha...

By @fchollet

74 score
AI Analysis

François Chollet announces PokeeClaw by Pokee_AI: a secure alternative to OpenClaw with sandbox architecture, isolated environments, approval workflows, RBAC, and audit trails. Also lower token usage.

OpenClaw has proven that local AI assistants have product-market fit. But the big issue with them has been security. The team at @Pokee_AI is fixing it with PokeeClaw: works like OpenClaw, but with in a secure sandbox architecture with isolated environments, approval workflows, role-based access control, and audit trails built in. Also lower token usage, which is always nice.
OpenClawAI securitylocal AI agentsenterprise AI
Social Twitter Mar 30

The average American worker using AI reports time savings of 6%, or 2.5 hours in a work week. Those ...

By @emollick

73 score
AI Analysis

Emollick shares data: average American worker using AI reports 6% time savings (2.5 hours/week), similar to UK/Netherlands, slightly more than other EU countries. Early signs this is translating into real productivity growth.

The average American worker using AI reports time savings of 6%, or 2.5 hours in a work week. Those are similar to the UK & Netherlands, and slightly more than other EU countries. There some early, non-causal, signs that this is translating into real gains in productivity growth t.co/XOspk4fp0Z
AI productivityAI adoptioneconomic impactworkforce
Social Twitter Mar 30

One of the things that is useful about the ChatGPT GPT-5.4 Pro (and also Thinking) harness is that i...

By @emollick

72 score
AI Analysis

Emollick highlights that GPT-5.4 Pro and Thinking are very good at reading scientific papers, not just text but also identifying and visually inspecting key figures.

One of the things that is useful about the ChatGPT GPT-5.4 Pro (and also Thinking) harness is that it is quite good at understanding how to read scientific papers, not just relying on text, but also figuring out which figures are key and inspecting those visually. t.co/3jHFNWoieP
GPT-5.4 capabilitiesscientific readingmultimodal AIAI practical use
Social Twitter Mar 30

It’s time for open-source agent tools to rely primarily on open-source models, instead of closed-sou...

By @ClementDelangue

72 score
AI Analysis

Clement argues it's time for open-source agent tools to rely primarily on open-source models instead of closed-source APIs, citing data privacy risks and potential for hacking/shutdowns.

It’s time for open-source agent tools to rely primarily on open-source models, instead of closed-source APIs that send all your data to the cloud and ultimately will get hacked and/or shut down
open-source AIAI agentsdata privacyAI securityclosed vs open models
Social Twitter Mar 30

3/ Two of the most powerful features in Claude Code: /loop and /schedule Use these to schedule Clau...

By @bcherny

72 score
AI Analysis

bcherny details /loop and /schedule for automated Claude Code workflows: auto-babysitting PRs, Slack feedback loops, PR pruning, post-merge sweeping - all running continuously.

3/ Two of the most powerful features in Claude Code: /loop and /schedule Use these to schedule Claude to run automatically at a set interval, for up to a week at a time. I have a bunch of loops running locally:
  • /loop 5m /babysit, to auto-address code review, auto-rebase, and shepherd my PRs to production
  • /loop 30m /slack-feedback, to automatically put up PRs for Slack feedback every 30 mins
  • /loop /post-merge-sweeper to put up PRs to address code review comments I missed
  • /loop 1h /pr-p
claude-codeautonomous-agentsdeveloper-workflowcontinuous-automation
Social Twitter Mar 30

This is an insane Anthropic tweet. And it’s a *buried reply* to one of their other tweets. I am r...

By @alliekmiller

72 score
AI Analysis

Allie K Miller reacts to an Anthropic tweet (buried in replies) showcasing Claude's coding capabilities. She recounts a talk where a senior Fortune 500 developer questioned why he'd use AI to code, and she predicted this level of capability was coming. Notes scaling laws are holding up and next iterations will be even more capable.

This is an insane Anthropic tweet. And it’s a *buried reply* to one of their other tweets. I am reminded of a talk I gave ~2-3 months ago where a senior developer at a Fortune 500 company asked me “why would I use AI to code if I can just code myself.” I answered. He said, “But sometimes it messes up.” I told him this was coming. Even if it’s not perfect today (it makes weird product features decisions sometimes, not gonna lie), the scaling laws seem to be holding up this year and the ne
Anthropic ClaudeAI codingscaling lawsdeveloper adoptionenterprise AI resistance