Category intelligence

Social Media Briefing — April 2, 2026

513 current items analyzed and ranked.

Executive synthesis

Social Media Summary

Anthropic's Claude Code dominated the day's discourse. Engineer Boris Cherny announced a major NO_FLICKER terminal renderer rewrite using viewport virtualization, drawing massive engagement. A separate deployment incident sparked a widely-viewed postmortem (1.1M views) showcasing Anthropic's blameless engineering culture, while transparent responses on scaling pain, rate limits, and an unintentional GitHub DMCA takedown of open-source forks further kept the team in the spotlight.

NVIDIA's Jim Fan unveiled CaP-X, an open-source agentic robotics framework bridging LLMs to physical manipulation across 12 frontier models and 187 tasks
Percy Liang (Stanford) reported a landmark 1e23 scaling run whose loss matched preregistered projections from 100x smaller models to within 0.005
Anthropic open-sourced Claudini, an automated adversarial red-teaming tool that outperformed hand-crafted attacks after 56 Claude Code loops
Business Insider revealed OpenAI's internal "Stagecraft" project — a 439-occupation spreadsheet driving 3,000–4,000 contractors to build domain-specific training data
Ethan Mollick offered two sharp observations: AI hasn't improved April Fools creativity despite free image generation, and AI labs are failing to articulate what their envisioned future means for ordinary people

Key Themes

Claude Code Product Updates · 20Claude Code Scaling & Reliability · 3Agentic Robotics (CaP-X) · 3Open Source & DMCA Controversy · 2Claude Code Operations & Incidents · 14AI and Society/Labor · 8AI Safety & Existential Risk Debate · 5Automated AI Red Teaming · 1AI Practical Applications & Democratization · 4Scaling Laws and Training · 1

Primary evidence

Top Ranked Signals

Social Twitter Apr 1

Today we're excited to announce NO_FLICKER mode for Claude Code in the terminal It uses an experime...

By @bcherny

92 score

AI Analysis

Anthropic engineer bcherny announces NO_FLICKER mode for Claude Code terminal - an experimental new renderer that virtualizes the viewport, supports mouse events in terminal, and eliminates flickering. Massive engagement (8K+ likes, 1.76M views).

Today we're excited to announce NO_FLICKER mode for Claude Code in the terminal It uses an experimental new renderer that we're excited about. The renderer is early and has tradeoffs, but already we've found that most internal users prefer it over the old renderer. It also supports mouse events (yes, in a terminal). Try it: CLAUDE_CODE_NO_FLICKER=1 claude

Claude Codedeveloper toolsproduct announcementterminal UX

Social Twitter Apr 1

@PrimeLineAI Thanks for the feedback. > Scrollback gutted - session history is barely scrollable. I...

By @bcherny

88 score

AI Analysis

Continuing our coverage from yesterday, Detailed response from bcherny (Anthropic) to a power user's complaints about Claude Code: addresses scrollback issues, rate limits due to fast user growth, 1M context window (free, with opt-out), and Opus 4.6 quality concerns. Emphasizes scaling challenges.

@PrimeLineAI Thanks for the feedback. > Scrollback gutted - session history is barely scrollable. I can't review my own conversation. can you tell me more about this? how do I repro? > Rate limits hit harder - prompt caching appears broken since March 23. Sessions that lasted hours now drain in 90 minutes. prompt caching is working correctly, and we've been shipping optimizations to make it work better. we announced reduced rate limits at peak recently due to our infra being strained because

Claude Codescaling challengesrate limitsAnthropicuser feedbackOpus 4.6

Social Twitter Apr 1

@GergelyOrosz This was not intentional, we’ve been working with GitHub to fix it. Should be better n...

By @bcherny

85 score

AI Analysis

Following yesterday's News coverage, Anthropic's bcherny responds to Gergely Orosz confirming that GitHub DMCA takedowns of Claude Code forks were NOT intentional. States they've been working with GitHub to fix it.

@GergelyOrosz This was not intentional, we’ve been working with GitHub to fix it. Should be better now.

Claude Codeopen sourceDMCA controversyAnthropic

Social Twitter Apr 1

The power of the Claw, in the palm of a robot hand. Agentic robotics is here! Today, we open-source ...

By @DrJimFan

82 score

AI Analysis

Jim Fan announces CaP-X: an open-source agentic robotics framework. Features vibe agents for physical world, robot arms and humanoids with perception/actuation APIs, auto-synthesized skill libraries. Includes CaP-Gym (187 manipulation tasks), CaP-Bench (benchmarks 12 frontier LLMs), CaP-Agent0 (training-free harness matching human expert code), and CaP-RL. Spiritual successor to Voyager (Minecraft agent).

The power of the Claw, in the palm of a robot hand. Agentic robotics is here! Today, we open-source CaP-X: vibe agents, alive in the physical world. They incarnate as robot arms and humanoids with a rich set of perception APIs, actuation APIs, and auto synthesize skill libraries as they go. CaP-X is a strict superset of our old stack, because policies like VLAs are “just” API calls as well. It solves many tasks zero-shot that a learned policy would struggle with. And we are doing much more than

agentic roboticsopen source AIembodied AILLM-robotics integrationbenchmarking

Social Twitter Apr 1

Some of the upsides: - No more flickering - No more jumping - Constant memory and CPU usage as the c...

By @bcherny

82 score

AI Analysis

Details the upsides of Claude Code's NO_FLICKER mode: no flickering/jumping, constant memory/CPU usage, mouse support for clicking within input, and improved selection behavior that excludes line numbers.

Some of the upsides:

No more flickering
No more jumping
Constant memory and CPU usage as the conversation grows
Mouse support! You can now click to move your cursor within the input box. Some other UI elements are also clickable now.
Nicer selection behavior. eg. when you select code, we no longer include line numbers and UI elements in the selection

Claude Codedeveloper toolsterminal UX

Social Twitter Apr 1

@wongmjane @BenLesh Mistakes happen. As a team, the important thing is to recognize it’s never an in...

By @bcherny

82 score

AI Analysis

Boris Cherny (Anthropic/Claude Code) explains a Claude Code deployment incident was caused by human error in a manual deploy step, emphasizes blameless culture and fixing process rather than punishing individuals. Over 1.1M views.

@wongmjane @BenLesh Mistakes happen. As a team, the important thing is to recognize it’s never an individuals’s fault — it’s the process, the culture, or the infra. In this case, there was a manual deploy step that should have been better automated. Our team has made a few improvements to the automation for next time, a couple more on the way.

claude_codeanthropic_operationsengineering_cultureai_assisted_devops

Social Twitter Apr 1

Our 1e23 Delphi run finished last night. It's loss was within 0.005 of the projected (preregistered...

By @percyliang

78 score

AI Analysis

Percy Liang reports that their 1e23 Delphi training run finished with loss within 0.005 of preregistered projections based on 100x smaller models. Notes remaining challenges with loss spikes and bending scaling laws.

Our 1e23 Delphi run finished last night. It's loss was within 0.005 of the projected (preregistered) loss. Note that these projections were based on only training models over 100x smaller (3e20)! Still more work to do. We still had loss spikes and if you closely, our scaling laws are bending. We have some ideas for fixing both...

scaling lawsAI trainingacademic AI researchDelphiloss prediction

Social Twitter Apr 1

56 loops of Claude Code just killed every hand-crafted AI attack method. Anthropic just open-sourc...

By @AlphaSignalAI

78 score

AI Analysis

Anthropic open-sourced 'Claudini', using Claude Code in an autoresearch loop to automatically discover adversarial attacks against LLMs. After 56 iterations, it found jailbreak methods with 40% success vs existing methods' 10%. The system studies existing attacks, designs new optimizers, benchmarks, and iterates automatically.

56 loops of Claude Code just killed every hand-crafted AI attack method. Anthropic just open-sourced a repo called Claudini. It uses Claude Code in an autoresearch loop to automatically discover new adversarial attacks against LLMs. After 56 iterations, it found an algorithm that jailbreaks safety-tuned models at 40% success rate. Every existing method scored 10% or below. Here's how it works: 1. Claude studies existing attack methods 2. Designs a new optimizer from scratch 3. Benchma

AI safetyadversarial attacksjailbreakingAnthropicautomated researchred teamingopen source

Social Twitter Apr 1

@ns123abc 👋 it was human error. Our deploy process has a few manual steps, and we didn’t do one of t...

By @bcherny

78 score

AI Analysis

Boris Cherny provides detailed postmortem on Claude Code deployment error: human error in manual steps, solution is more automation and using Claude to check results rather than adding more process.

@ns123abc 👋 it was human error. Our deploy process has a few manual steps, and we didn’t do one of the steps correctly. We have landed a few improvements and are digging in to add more sanity checks. Like with any other incident, the counter-intuitive answer is to solve the problem by finding ways to go faster, rather than introducing more process. In this case more automation & claude checking the results.

claude_codeanthropic_operationsai_assisted_devops

Social Twitter Apr 1

not an April Fools joke, we rewrote the Claude Code renderer to use a virtual viewport you can use ...

By @trq212

78 score

AI Analysis

Claude Code team announces a rewritten renderer using a virtual viewport - enabling mouse support, persistent prompt input at bottom, and other UX improvements. Marked as experimental.

not an April Fools joke, we rewrote the Claude Code renderer to use a virtual viewport you can use your mouse, the prompt input stays at the bottom, and a lot more small UX wins people have been asking for it's experimental so give us your feedback

claude_codeproduct_updatesdeveloper_tools

Social Twitter Apr 1

Either we accept this as a society, and set a precedent for allowing virtually all jobs to be replac...

By @GaryMarcus

75 score

AI Analysis

Marcus makes an impassioned call to defend artists, writers, and musicians against AI displacement, arguing this sets a precedent for replacing all jobs with almost no compensation.

Either we accept this as a society, and set a precedent for allowing virtually all jobs to be replaced with almost no compensation. Or we speak up now. For artists. For writers. For musicians. For everybody.

AI and creative jobsAI ethicslabor displacementartist rightsAI regulation

Social Twitter Apr 1

A 439-row spreadsheet, obtained by Business Insider, lists the occupations OpenAI has hired freelanc...

By @TheRundownAI

75 score

AI Analysis

Details OpenAI's internal 'Stagecraft' project: a 439-occupation spreadsheet of freelancers (3,000-4,000 contractors at $50+/hr) building ChatGPT training materials through Handshake AI. Contractors create real-world tasks to map economically relevant work and evaluate model capabilities. One contractor noted awareness of 'training AI to replace us.'

A 439-row spreadsheet, obtained by Business Insider, lists the occupations OpenAI has hired freelancers to build ChatGPT training materials for. The project is internally called 'Stagecraft'. It's a wide range. Commercial pilots. Emergency physicians. Geoscientists. Soil scientists. Music directors. Fine artists. Farmers. Sales reps. HR specialists. 3,000 to 4,000 contractors. Paid at least $50/hour. Instructions: create tasks "as if directing a colleague with real-world asks." The goal:

OpenAIAI training datalabor displacementAI workforce impactdata labeling