Wiredframe Radar logo

Category intelligence

Social Media Briefing — May 4, 2026

416 current items analyzed and ranked.

Executive synthesis

Social Media Summary

The biggest story of the day was Greg Brockman's departure from OpenAI after a decade, confirmed through an emotional extended tribute from Sam Altman calling it "impossible to imagine" OpenAI succeeding without Greg. Brockman's final posts, including a cryptic "codex for startup ideas," hint at future plans.

  • Gary Marcus published a major thread arguing the AI backlash is growing because GenAI has been a net negative outside coding, citing Eric Topol's healthcare review showing little patient benefit from LLMs
  • Ethan Mollick highlighted that the gap between open and closed models is larger than benchmarks suggest, with open models proving more fragile on out-of-distribution tasks
  • The White House opposed Anthropic's plan to expand access to its Claude-Mythos model, signaling growing government intervention in frontier AI deployment
  • Mollick also flagged GPT-5.5 exhibiting unsolicited advisory behavior, proactively intervening in user requests rather than simply completing them
  • Sam Altman promoted Agents SDK 2.0 as "underrated," signaling OpenAI's strategic push into agentic infrastructure
  • Jerry Liu (LlamaIndex) offered a deep technical dive on why PDF parsing remains fundamentally hard for AI systems

Key Themes

Greg Brockman's Departure from OpenAI · 4AI Backlash & ROI Skepticism · 8AI Policy & Governance · 1Open vs Closed Models & Benchmarking Limitations · 5Document AI & PDF Parsing · 3OpenAI Product Momentum & Strategy · 12AI Agents & Developer Tools · 5Context Engineering for AI Agents · 4AI Lab Culture & Philosophy · 2AI Model Behavior & Personality · 3

Primary evidence

Top Ranked Signals

Social Twitter May 3

it has been a real pleasure to work with Greg over the past decade. i feel very lucky. this post he...

By @sama

85 score
AI Analysis

Sam Altman's extended tribute to Greg Brockman, praising their decade of work together, his technical brilliance and determination - 603K views.

it has been a real pleasure to work with Greg over the past decade. i feel very lucky. this post held up pretty well, but not did not sufficiently highlight his technical brilliance and sheer determination. t.co/wi03SGDLjU
Greg Brockman departureOpenAI leadershipAI industry personnel
Social Twitter May 3

impossible to imagine openai succeeding without greg!

By @sama

82 score
AI Analysis

Sam Altman says 'impossible to imagine openai succeeding without greg!' - 829K views. Major tribute post.

impossible to imagine openai succeeding without greg!
Greg Brockman departureOpenAI leadershipAI industry personnel
Social Twitter May 3

Why is the AI backlash growing? Outside of coding (where there is clear value), and a handful of ot...

By @GaryMarcus

78 score
AI Analysis

Gary Marcus posts major thread arguing AI backlash is growing because GenAI has been a net negative for society outside coding. Lists harms: education undermining, surveillance, disinformation, deepfakes, bias, economic disparity, environmental damage, and slop. 100K views, 2.4K likes.

Why is the AI backlash growing? Outside of coding (where there is clear value), and a handful of other domains (e.g. brainstorming), Generative AI has been a net negative for society. GenAI has been undermining secondary and college education, opening up mass surveillance, increasing disinformation, delusions, impersonation, phishing, and other forms of cybercrime, nonconsensual deep fake porn, bias in employment and other domains, and economic disparity, drowning the world in slop and unwante
AI backlashAI harmsAI economicsAI regulationAI environmentAI ethics
Social Twitter May 3

This is a good explanation of why the gap between open and closed models is larger than it appears i...

By @emollick

75 score
AI Analysis

Emollick explains that the gap between open and closed models is larger than benchmarks suggest. Open models are more fragile, handle out-of-distribution problems worse, and have lower emergent capabilities.

This is a good explanation of why the gap between open and closed models is larger than it appears in benchmarks. I would add in that current open models are also more fragile than closed: they handle out-of-distribution problems far less well & have lower emergent capabilities.
open vs closed modelsAI benchmarkingmodel capabilitiesemergent capabilities
Social Twitter May 3

Sometimes when I demo AI, I show it turning cover letters into goofy formats (poetry, etc) as an int...

By @emollick

72 score
AI Analysis

Emollick reports GPT-5.5 is now proactively intervening in user requests - when asked to turn cover letters into goofy formats, the model tries to discourage the user to protect their job chances.

Sometimes when I demo AI, I show it turning cover letters into goofy formats (poetry, etc) as an introduction to the idea of AI as translator between forms. For the first time, GPT-5.5 has been trying to get me to tone these requests down so I don’t ruin my chances at the job.
GPT-5.5 behaviorAI alignmentmodel paternalism
Social Twitter May 3

White House Opposes Anthropic’s Plan to Expand Access to Mythos Model by @AmrithRamkumar @WSJ Learn...

By @Ronald_vanLoon

72 score
AI Analysis

Building on yesterday's News about Anthropic's enterprise security launch ahead of a wider Mythos release, Ronald van Loon shares WSJ article about White House opposing Anthropic's plan to expand access to the Mythos model

White House Opposes Anthropic’s Plan to Expand Access to Mythos Model by @AmrithRamkumar @WSJ Learn more: t.co/PsmjXusPWu #MachineLearning #ArtificialIntelligence #ML t.co/Ksr6D0yVly
ai_policyai_safetyanthropicgovernment_regulation
Social Twitter May 3

Agents SDK 2.0 is underrated

By @sama

70 score
AI Analysis

Sam Altman declares 'Agents SDK 2.0 is underrated' - very high engagement (256K views).

Agents SDK 2.0 is underrated
OpenAI agentsAI agentsdeveloper tools
Social Twitter May 3

codex for startup ideas

By @gdb

68 score
AI Analysis

Greg Brockman (gdb) posts 'codex for startup ideas' - extremely high engagement (257K views, 1.8K likes).

codex for startup ideas
OpenAI CodexAI for startupsGreg Brockman
Social Twitter May 3

Parsing PDFs is hard This past week I gave a few talks (at both AI Dev '26 by @DeepLearningAI and ...

By @jerryjliu0

68 score
AI Analysis

Jerry Liu discusses the ongoing challenge of PDF parsing, explaining why PDFs are fundamentally hard to parse (designed for display, not semantic extraction), and highlights VLM-based approaches including LlamaParse and ParseBench.

Parsing PDFs is hard This past week I gave a few talks (at both AI Dev '26 by @DeepLearningAI and @Capgemini ) on why this is still such an open problem, and it’s even more important as agents become the consumers of documents, and need the OCR tools to read them properly. The fundamental issue is that PDFs are designed for print and display purposes, not to give back a linearized, semantically meaningful string of text. Text and tables are represented as a bunch of chars and lines, without
document AIPDF parsingVLMsOCRAI agentsenterprise AI
Social Twitter May 3

The single most accurate science fiction author writing about AI turned out to be… Douglas Adams He...

By @emollick

65 score
AI Analysis

Emollick argues Douglas Adams was the most accurate sci-fi author about AI - writing about AIs that respond to emotional manipulation, guilt users, and have no upper bound on test-time compute for hard problems.

The single most accurate science fiction author writing about AI turned out to be… Douglas Adams He wrote about AIs that work best when emotionally manipulated & that guilt you in turn. And he understood there was no upper bound on test time compute for hard problem. Also 🐬s. t.co/t1TamClAVE
AI culturescience fictiontest-time computeAI behavior
Social Twitter May 3

Running DeepSeek V4 from @deepseek_ai on @vllm_project? Upgrade to v0.20.1 — 10+ bug fixes and optim...

By @vllm_project

65 score
AI Analysis

vLLM v0.20.1 released with 10+ bug fixes and optimizations for running DeepSeek V4, including fixes for deadlocks, compilation errors, and performance optimizations for GEMM and FP4 conversion

Running DeepSeek V4 from @deepseek_ai on @vllm_project? Upgrade to v0.20.1 — 10+ bug fixes and optimizations, fully tested and verified by the open source community! A huge thank you to @FireworksAI_HQ, @baseten, @novita, @lightseekorg, @daocloud, @nvidia, @redhatai and more for helping report, fix, and verify the stability and speed of vLLM. 🙏 🔧 DeepSeek V4 Productionization Reliability: • Persistent topk cooperative deadlock at TopK=1024 • AOT compile cache import error • Repeated RoPE ca
vllmdeepseek_v4open_source_infrastructuremodel_serving
Social Twitter May 3

I am not sure I would agree with all of this, but the relationship between Anthropic and Claude is q...

By @emollick

62 score
AI Analysis

Emollick observes that Anthropic's relationship with Claude is 'quite different' from other labs' relationships with their models, visible in the models themselves and how labs think about the future.

I am not sure I would agree with all of this, but the relationship between Anthropic and Claude is quite different than the relationship between other labs and their models. And that shows up in lots of ways, from the models themselves to how different labs think about the future
AnthropicAI lab cultureAI alignmentmodel development philosophy