Category intelligence

Social Media Briefing

All 357 current items, analyzed and ranked. May 25, 2026

Executive synthesis

Social Media Summary

AI community chatter on 2026-05-24 centered on open-source friction, coding agent workflows, and renewed skepticism about the GenAI bubble.

vLLM publicly banned a contributor for submitting AI-slop PRs as part of a 'resume-building' workflow, sparking broader discussion about agents flooding OSS with low-value contributions.
Boris Cherny (Anthropic) drove heavy engagement recommending Claude Code's auto mode for parallel 'multi-clauding,' while also warning about the dangers of bypass permissions.
Greg Brockman highlighted that Codex is open source and shared a self-improvement prompt, both going viral.

Bubble/economics talk intensified: Gary Marcus likened LLM companies to airlines (thin margins, commoditization) and amplified a Princeton/Swarat neurosymbolic result beating OpenAI at the Erdős benchmark in three days. Andrej Karpathy praised a large company 'calling BS' on AI hype.
Ethan Mollick warned that as users learn AI tells, they'll realize how much social media, blogs, and scientific papers are AI-generated — echoed by complaints about bots impersonating humans on X.
Clément Delangue flagged that llama.cpp MTP support makes Qwen3.6-27B 78% faster on A10G, pushing local models toward daily-driver viability. hardmaru (Sakana) countered replacement fears with a Jevons-paradox take on engineer productivity.

Key Themes

Open-source and AI slop in OSS · 1Claude Code workflows & auto mode · 13GenAI Bubble & LLM Economics · 6Neurosymbolic AI vs OpenAI · 4AI Content Saturation & Authenticity · 3OpenAI/Codex Updates · 3Open science and lab culture critique · 5Local AI & Hardware · 3China AI dominance vs Japan · 3AI Authenticity & Saturation in Online Content · 2

Primary evidence

All Ranked Signals

Showing 10 of 357 signals 2 remaining

Social Twitter May 24

Thanks to the community report, we recently identified a PR https://t.co/QWboSmskkF that attempted t...

By @vllm_project

85 score

AI Analysis

vLLM project bans contributor for 'AI slop' PR submitted as part of resume-building 'PR training' workflow; announces formal channel for important contributions and warns about AI-generated low-quality OSS contributions.

Thanks to the community report, we recently identified a PR t.co/QWboSmskkF that attempted to solve a non-existent issue and was submitted as part of a “PR training” workflow for resume building. The contributor involved has been banned from the vLLM community. This kind of low-signal contribution increases maintainer review overhead and creates unnecessary operational costs for open-source projects. As AI coding agents make generating large volumes of small PRs increasingly cheap, op

open sourceAI slopvLLMOSS sustainability

Source details

open sourceAI slopvLLMOSS sustainability

Social Twitter May 24

People often ask what my biggest tip is for getting the most out of Claude Code. These days my #1 t...

By @bcherny

80 score

AI Analysis

Cherny's top tip for Claude Code: use auto mode (no permission prompts) to enable parallel multi-clauding

People often ask what my biggest tip is for getting the most out of Claude Code. These days my #1 tip is: use auto mode Auto mode means no more permission prompts. It is the key building block for multi-clauding: start a session, then while it runs, work on another session in parallel.

Claude Codeauto modeagentic workflowsproductivity

Source details

Claude Codeauto modeagentic workflowsproductivity

Social Twitter May 24

As more people come to recognize the tells of AI, which mostly happens as you start to work with AI ...

By @emollick

75 score

AI Analysis

Mollick predicts mass realization of how much online content is AI-generated.

As more people come to recognize the tells of AI, which mostly happens as you start to work with AI a lot, the scales are going to fall from their eyes and they are going to realize what some of us already see: how much of this site (and blog posts, articles, papers) are AI now.

AI contentAI detectiononline discourse

Source details

AI contentAI detectiononline discourse

Social Twitter May 24

llama.cpp with MTP support makes local models fast enough to use as daily drivers 🚀 Qwen3.6-27B de...

By @ClementDelangue

75 score

AI Analysis

Delangue: llama.cpp now supports MTP, accelerating Qwen3.6-27B by 78% on A10G.

llama.cpp with MTP support makes local models fast enough to use as daily drivers 🚀 Qwen3.6-27B dense generation below on A10G: From 25 tok/st to 45 tok/s (+78%)! t.co/rLjBVa3Yzh

local AIllama.cppQwen3.6performance

Source details

llama.cpp with MTP support makes local models fast enough to use as daily drivers 🚀 Qwen3.6-27B dense generation below on A10G: From 25 tok/st to 45 tok/s (+78%)! t.co/rLjBVa3Yzh

local AIllama.cppQwen3.6performance

Social Twitter May 24

under appreciated that codex is open source

By @gdb

75 score

AI Analysis

Brockman: 'under appreciated that codex is open source'.

under appreciated that codex is open source

Codexopen sourceOpenAI

Source details

under appreciated that codex is open source

Codexopen sourceOpenAI

Social Twitter May 24

⚠️⚠️⚠️i don’t think most people understand the implications of the mood shift below, so I will spell...

By @GaryMarcus

70 score

AI Analysis

Marcus argues GeoHotz acknowledging AI coding produces slop signals collapse of GenAI bubble.

⚠️⚠️⚠️i don’t think most people understand the implications of the mood shift below, so I will spell them out. they are serious, and eventually will affect the global economy. when a serious coder as hardcore AI-philic as GeoHotz dives deep into all the latest coding models and winds up echoing what I have been saying about the code turning out slop—and how that is going to mess over big companies—the whole genAI movement is a deadman walking. ‘cause coding was the tentpole. and its net impac

AI codingGenAI bubbleskepticism

Source details

AI codingGenAI bubbleskepticism

Social Twitter May 24

self improvement prompt for codex

By @gdb

70 score

AI Analysis

Brockman shares 'self improvement prompt for codex'.

self improvement prompt for codex

CodexOpenAIself-improvement

Source details

self improvement prompt for codex

CodexOpenAIself-improvement

Social Twitter May 24

@shreyansj It’s refreshing to see a company of this size successfully call bs on the whole thing to ...

By @karpathy

65 score

AI Analysis

Karpathy praises a company calling BS on AI hype.

@shreyansj It’s refreshing to see a company of this size successfully call bs on the whole thing to this extent. One group of MTS on a mission, clean.

AI hypeindustry critique

Source details

@shreyansj It’s refreshing to see a company of this size successfully call bs on the whole thing to this extent. One group of MTS on a mission, clean.

AI hypeindustry critique

Social Twitter May 24

Breaking: Young Princeton prof beats OpenAI at its own Erdos game, 3 days later. https://t.co/7gh3...

By @GaryMarcus

65 score

AI Analysis

Marcus: Princeton prof beats OpenAI at Erdos game in 3 days.

Breaking: Young Princeton prof beats OpenAI at its own Erdos game, 3 days later. t.co/7gh3KnxtZx

neurosymbolic AIOpenAIErdos problems

Source details

Breaking: Young Princeton prof beats OpenAI at its own Erdos game, 3 days later. t.co/7gh3KnxtZx

neurosymbolic AIOpenAIErdos problems

Social Twitter May 24

Classical AI planners can find optimal solutions to combinatorial problems like sliding-tile puzzles...

By @burkov

65 score

AI Analysis

Burkov explains Latplan: discrete autoencoder learning symbolic representations from images for classical planners.

Classical AI planners can find optimal solutions to combinatorial problems like sliding-tile puzzles in milliseconds, but only if a human first writes the problem out in a symbolic language such as PDDL that names every proposition and every action together with its preconditions and effects. This requirement has kept symbolic planning largely separate from raw sensor data such as photographs, where there are no pre-existing symbols, only pixels. The authors of this paper build a system called

neurosymbolic AIplanningautoencodersPDDL

Source details

neurosymbolic AIplanningautoencodersPDDL