Wiredframe Radar logo

Category intelligence

Social Media Briefing — April 23, 2026

559 current items analyzed and ranked.

Executive synthesis

Social Media Summary

The AI community was dominated by dueling enterprise agent platform launches. OpenAI unveiled workspace agents built on Codex, with Greg Brockman detailing cloud-hosted agents connected to Slack and recurring tasks, while Sam Altman endorsed the product to massive engagement. Simultaneously, Google DeepMind launched the Gemini Enterprise Agent Platform with Google Cloud, signaling agentic AI as the primary enterprise battleground.

  • Ethan Mollick delivered the day's sharpest insight: every system implicitly regulated by human effort—recommendation letters, lawsuits, government filings—will break as AI removes those effort constraints
  • Aravind Srinivas revealed Perplexity post-trained a model on Qwen achieving Pareto-optimal accuracy-cost, already serving production traffic and outperforming GPT and Sonnet on efficiency
  • Jeremy Howard sharply criticized Anthropic for quietly removing Claude Code mentions from Pro plan documentation, calling it a collapse of integrity
  • Sony AI published a Nature paper on the first autonomous robot to beat elite humans at competitive table tennis
  • A developer reverse-engineered Claude Mythos, revealing a novel adaptive-depth architecture where a single block loops up to 16 times per forward pass
  • NVIDIA and Google Cloud expanded their partnership around Vera Rubin A5X instances scaling toward ~1M GPUs

Key Themes

OpenAI Workspace Agents & Codex · 5Enterprise Agentic AI Launches · 8AI's Societal & Institutional Impact · 2Anthropic Pricing/Transparency Controversy · 4Perplexity Custom Post-Trained Model · 1GPT-ImageGen-2 Capabilities & Limitations · 12NVIDIA-Google Cloud Infrastructure Partnership · 12New Model Releases and Architecture Insights · 4AI Pricing Sustainability Crisis · 9AGI Definition & Evaluation · 1

Primary evidence

Top Ranked Signals

Social Twitter Apr 22

Build workspace agents for your team, on top of a cloud-hosted Codex harness. Hook them up to tools,...

By @gdb

88 score
AI Analysis

Greg Brockman announces OpenAI's workspace agents: cloud-hosted Codex-based agents that teams can build, connect to tools, give recurring tasks, and interact with via Slack.

Build workspace agents for your team, on top of a cloud-hosted Codex harness. Hook them up to tools, give them recurring tasks, and talk to them from surfaces like Slack. Easier than ever to bring the power of agents to your computer work.
openai-productsai-agentsenterprise-aicodexworkplace-automation
Social Twitter Apr 22

Every system that was regulated, either explicitly or implicitly, by the fact that they were effortf...

By @emollick

82 score
AI Analysis

Mollick argues that every system implicitly regulated by being effortful for humans (letters of recommendation, lawsuits, government filings, essays) will break due to AI automation of effort.

Every system that was regulated, either explicitly or implicitly, by the fact that they were effortful for humans (letters of recommendation, lawsuits, government filings, essays) will break.
societal-impactai-disruptioninstitutional-designautomation
Social Twitter Apr 22

Introducing workspace agents in ChatGPT—shared agents that can handle complex tasks and long-running...

By @OpenAI

82 score
AI Analysis

OpenAI introduces workspace agents in ChatGPT - shared agents for complex tasks and long-running workflows across tools and teams.

Introducing workspace agents in ChatGPT—shared agents that can handle complex tasks and long-running workflows across tools and teams. t.co/eHplfXCWlk
OpenAIworkspace agentsenterprise AIproduct launchagentic AI
Social Twitter Apr 22

We’ve post trained a model on top of Qwen that achieves Pareto optimality on accuracy-cost curves. ...

By @AravSrinivas

82 score
AI Analysis

Arav Srinivas (Perplexity CEO) announces they've post-trained a model on Qwen that achieves Pareto-optimal accuracy-cost curves, unifying tool-call routing and summarization. It outperforms GPT and Sonnet in cost efficiency and is already serving significant production traffic.

We’ve post trained a model on top of Qwen that achieves Pareto optimality on accuracy-cost curves. Unlike our previous post trained models, this model has been trained to be good at search and tool calls simultaneously, allowing us to unify the tool call router and summarization together in one model. The resulting model performs better than GPT and Sonnet in terms of cost efficiency to serve daily Perplexity queries in production. The production model runs on our own inference platform. We
Perplexitypost-trainingQwenproduction AIcost optimizationtool callingopen-source models
Social Twitter Apr 22

.@GoogleCloud and NVIDIA are expanding their partnership across agentic and physical AI.   At #Googl...

By @nvidia

80 score
AI Analysis

Major NVIDIA-Google Cloud partnership expansion at Google Cloud Next: Vera Rubin-powered A5X instances scaling to ~1M Rubin GPUs, Gemini on Google Distributed Cloud with Blackwell, first confidential computing Blackwell GPUs in cloud, agentic AI on Gemini Enterprise Agent Platform with Nemotron/NeMo.

.@GoogleCloud and NVIDIA are expanding their partnership across agentic and physical AI. At #GoogleCloudNext, the companies made several announcements, including: ✅ NVIDIA Vera Rubin-powered A5X instances, scaling up to nearly 1M Rubin GPUs ✅ Gemini on Google Distributed Cloud, powered by the NVIDIA Blackwell platform ✅ First Confidential Computing NVIDIA Blackwell GPUs in the cloud ✅ Agentic AI built on Gemini Enterprise Agent Platform with NVIDIA Nemotron and NeMo
AI infrastructureNVIDIA-Google Cloud partnershipVera Rubin GPUsconfidential computingagentic AIGoogle Cloud Next
Social Twitter Apr 22

Sony AI just published the first autonomous robot to beat elite humans at a competitive physical spo...

By @TheRundownAI

78 score
AI Analysis

Sony AI published a Nature paper on 'Ace', the first autonomous robot to beat elite humans at table tennis. Uses 9 cameras, 20ms reaction time (10x faster than humans), trained via 3000 hours of self-play simulation. By December 2025 it beat a professional player.

Sony AI just published the first autonomous robot to beat elite humans at a competitive physical sport. Its name is Ace. The sport is table tennis. The paper dropped in Nature today. 9 cameras triangulate the ball in 3D. Three systems zoom in on the ball's logo mid-flight to read its spin axis. End-to-end reaction time is around 20 milliseconds, roughly 10x faster than a human. It learned by playing itself for 3,000 hours in simulation. No human demonstrations. In April 2025, Ace beat 3 of
roboticsreinforcement learningsim-to-real transferphysical AIscientific publication
Social Twitter Apr 22

For the "small test" they've modified their docs to remove mention of Claude Code in Claude Pro: htt...

By @jeremyphoward

78 score
AI Analysis

Following yesterday's Reddit discussion, Jeremy Howard criticizes Anthropic for modifying docs to remove Claude Code from Claude Pro mentions, calling it a collapse of integrity under commercial pressure.

For the "small test" they've modified their docs to remove mention of Claude Code in Claude Pro: t.co/cG75PWlZyj It's been a shock to see Anthropic's integrity collapse in the face of commercial pressure. Would love a renewed commitment to straightforward honesty.
Anthropic controversyClaude Codepricing transparencyAI company trust
Social Twitter Apr 22

This wasn't the case with previous image generators, but the LLM you select has a huge effect on GPT...

By @emollick

75 score
AI Analysis

Building on yesterday's Reddit discussion, Mollick reveals that GPT-ImageGen-2 output quality varies significantly based on which LLM is selected. GPT-5.4 Thinking and GPT-5.4 Pro produce much better images, especially for complex tasks. Notes this isn't intuitive or documented.

This wasn't the case with previous image generators, but the LLM you select has a huge effect on GPT-imagegen-2 output. GPT-5.4 Thinking and GPT-5.4 Pro will produce much better images, especially for complex things. This is, of course, not intuitive or explained anywhere.
image-generationopenai-productsgpt-imagegen-2model-behaviorpractical-tips
Social Twitter Apr 22

We’re launching Gemini Enterprise Agent Platform with @GoogleCloud: a platform for businesses to dev...

By @GoogleDeepMind

75 score
AI Analysis

Google DeepMind launches Gemini Enterprise Agent Platform with Google Cloud - a platform for businesses to develop, scale, govern and optimize agents, evolving from Vertex AI.

We’re launching Gemini Enterprise Agent Platform with @GoogleCloud: a platform for businesses to develop, scale, govern and optimize agents. It’s the evolution of Vertex AI, bringing together model selection and agent building with new features for integration, security and more.
Google Cloudenterprise AIagentic AIproduct launchVertex AI
Social Twitter Apr 22

A developer just reverse engineered Claude Mythos and open sourced it. It's called OpenMythos and ...

By @AlphaSignalAI

75 score
AI Analysis

Developer reverse-engineered Claude Mythos and open-sourced it as 'OpenMythos' in PyTorch. Key architecture: single block runs up to 16 times per forward pass (looped transformer). 770M params matches 1.3B standard transformer. Reasoning in latent space without CoT tokens. Adaptive compute with early halting.

A developer just reverse engineered Claude Mythos and open sourced it. It's called OpenMythos and implemented in PyTorch. The core idea: instead of stacking hundreds of unique layers, a single block runs up to 16 times per forward pass. Same weights, more loops, deeper thinking. A 770M parameter version matches a 1.3B standard transformer in quality. Reasoning happens silently in continuous latent space, with no chain-of-thought tokens emitted between steps. Each iteration activates a dif
model architectureClaude Mythoslooped transformersadaptive computeopen sourcelatent reasoning