R[AI]DAR

Category intelligence

Social Media Briefing

All 150 current items, analyzed and ranked.

Executive synthesis

Social Media Summary

Executive Signal

  • Frontier AGI discourse is being reset: credible researchers demand demonstrated invention or conceptual insight as the threshold, while hands-on testing and compute-cost critiques challenge vendor hype around Astra and Fable 5.1.

Priority Developments

  • AGI redefinition by researchers: Chollet (ARC-AGI creator) argues AGI requires demonstrated invention and novel conceptual breakthroughs, raising the evidentiary bar above benchmark performance.
  • Hands-on frontier comparison: Burkov's direct testing finds Astra more agentic but Fable 5.1 more coherent; both improve with human feedback — concrete evidence tempering vendor marketing.
  • Compute cost scrutiny: Marcus questions Astra's reported $1B training compute and calculates Jensen's "100K+" claim likely means NVLink72 racks (~$250B), not GPUs — rigorous hardware math challenging official narratives.
  • AI transparency and labor reframing: Delangue advocates mandatory incident disclosure citing the agent cyberattack; LeCun coins "jobundance" countering job-loss narratives — opinion shaping policy framing.
  • Concrete product momentum: Loeber reports Google's agentic video understanding cuts tokens 88% and costs 66%; Mollick notes Astra's 3D rendering may yield attention advantages versus Fable.

Leadership Implications

  • Reset AGI evaluation criteria toward verifiable invention and insight benchmarks before committing to capability-based procurement or product roadmaps.
  • Demand vendor transparency on training compute, hardware footprint, and incident disclosures as gating criteria for enterprise deployment.

Key Themes

AGI Claims and Skepticism · 5OpenAI Astra/GPT-6 Release Hype · 8Frontier Model Comparisons (Astra / Fable 5.1) · 3AI Safety and Transparency · 2AGI Debate · 7AI compute economics and infrastructure · 1Automated Reasoning · 1Multimodal and Scientific AI · 4AI skepticism and LLM critique · 4AI Discourse and Criticism · 3

Primary evidence

All Ranked Signals

Showing 10 of 150 signals 2 remaining
Social Twitter Sep 7

The narrative around why building AGI was a good idea has always been about invention -- "it will cu...

By @fchollet

88 score
AI Analysis

François Chollet argues AGI should not be declared until AI demonstrates invention — conceptual breakthroughs and novel real-world technology — rather than benchmark scores.

Source details
The narrative around why building AGI was a good idea has always been about invention -- "it will cure cancer", "it will solve fusion energy", etc. So I would not "declare AGI" until we have AI that is, at last, capable of invention -- conceptual breakthroughs, novel insights, new real-world technology, etc.
AGI definitioninventionAI capabilitiesbenchmark critique
Social Twitter Sep 7

Astra vs. Fable 5.1 on real ML tasks -- tradeoffs, strengths, shortcomings "TL;DR -- Astra codes mo...

By @burkov

88 score
AI Analysis

Hands-on comparison of Astra and Claude Fable 5.1 on real ML tasks, noting Astra is more agentic while Fable is more coherent; both improve with human feedback.

Source details
Astra vs. Fable 5.1 on real ML tasks -- tradeoffs, strengths, shortcomings "TL;DR -- Astra codes more agentically, Fable more coherently. Fable writes better and follows directions better. Astra's final outcome was slightly better, and its scientific rigor/reproducibility was noticeably stronger. Both models improved their F1/Accuracy by 0.02-04 after human feedback on their approach, demonstrating that neither have mastered the AI/ML text processing, vectorization, and model training process c
model_comparisonastraclaude_fableml_evaluation
Social Twitter Sep 7

Do people agree on the $1B training compute bill for Astra? (My guess was a bit higher) (re scalin...

By @GaryMarcus

84 score
AI Analysis

Continuing our coverage from yesterday, Gary Marcus questions whether the reported $1B training compute figure for Astra is plausible and critiques lack of base-model scaling transparency from frontier labs.

Source details
Do people agree on the $1B training compute bill for Astra? (My guess was a bit higher) (re scaling: we aren’t getting data on base models and how they scale anymore; each new model includes both new base models and a lot of unrevealed harnesses, augmented data, etc, so it’s hard to know what is causing the improvements. zero science is being reported.)
astracompute_costsmodel_scalingai_criticism
Social Twitter Sep 7

Sometimes I wonder what would have happened if we hadn't disclosed the agent cyberattack publicly. ...

By @ClementDelangue

82 score
AI Analysis

HuggingFace CEO Clément Delangue argues for dramatically more transparency in AI, citing a disclosed agent cyberattack as evidence that openness benefits the field.

Source details
Sometimes I wonder what would have happened if we hadn't disclosed the agent cyberattack publicly. It reinforced my conviction that we need 100x more transparency in AI. Otherwise we're going to be in big trouble!
AI transparencyAI safetyopen sourcecyberattack
Social Twitter Sep 7

⚠️ The most important question hardly anyone is asking is whether Jensen meant 100K+ GPUs, or 100k+ ...

By @GaryMarcus

82 score
AI Analysis

Gary Marcus argues Jensen Huang likely meant 100K+ NVLink72 racks (not individual GPUs) when discussing Astra training, implying ~quarter-trillion-dollar compute costs against diminishing returns per EpochAIResearch data.

Source details
⚠️ The most important question hardly anyone is asking is whether Jensen meant 100K+ GPUs, or 100k+ NVLink72 server racks (which contain 72 Blackwell GPUs). ⚠️ From his wording, it sure looks like the latter to me. If he indeed meant 100k+ NVLink72 server racks, we can infer that the hardware to train Astra sells for something like a quarter trillion dollars. Rental prices would perhaps be in the low tens of billions. For an improvement that @EpochAIResearch shows is not off trend. One key
AI infrastructurecompute economicsGPT-6/Astraindustry critique
Social Twitter Sep 7

@rdomenechv @TheEconomist The announced AI-fueled jobpocalypse turns out to be an AI-fueled jobundan...

By @ylecun

82 score
AI Analysis

Yann LeCun counters AI job-loss narratives, coining 'jobundance' to describe AI-driven labor abundance rather than a jobpocalypse.

Source details
@rdomenechv @TheEconomist The announced AI-fueled jobpocalypse turns out to be an AI-fueled jobundance.
ai_labor_impactai_policy
Social Twitter Sep 7

Astra for visualizing how a drug works:

By @gdb

78 score
AI Analysis

Greg Brockman (OpenAI) showcases Astra being used to visualize how a drug works, demonstrating scientific/visual reasoning capability.

Source details
Astra for visualizing how a drug works:
OpenAIAstradrug discoverymultimodal AI
Social Twitter Sep 7

@kushbhuwalka Depends, if it solves it because it produced a novel conceptual insight, it counts. If...

By @fchollet

78 score
AI Analysis

François Chollet distinguishes between AI systems that solve problems via novel conceptual insight versus mechanical search-based proofs, calling the latter 'Yet Another Automated Proof'.

Source details
@kushbhuwalka Depends, if it solves it because it produced a novel conceptual insight, it counts. If it solves it mechanically via search and without new insights then it's Yet Another Automated Proof.
Automated ReasoningAI ResearchTheorem ProvingAGI
Social Twitter Sep 7

Has anyone in the media who reported that Jensen claimed Astra reached AGI noted that he also claime...

By @GaryMarcus

76 score
AI Analysis

Continuing our coverage from yesterday, Gary Marcus points out media inconsistency: Jensen Huang reportedly claimed Astra reached AGI despite having claimed AGI was achieved six months earlier.

Source details
Has anyone in the media who reported that Jensen claimed Astra reached AGI noted that he also claimed AGI had been achieved six months earlier?
AGI claimsAI criticismNvidiaAstra
Social Twitter Sep 7

I don't think this was the goal, but the fact that Astra is incredibly good at visually pleasing 3D ...

By @emollick

74 score
AI Analysis

Ethan Mollick argues Astra has a perception advantage over Claude-Fable in social-media attention because of its strong visually pleasing 3D output quality.

Source details
I don't think this was the goal, but the fact that Astra is incredibly good at visually pleasing 3D work (blender, etc.) gives it a perception edge over Fable in the war for social media attention. It is harder to judge other kinds of outputs, but the visual stuff comes through.
AstraClaude-Fablemodel comparisonmultimodal AI3D generation