AI building AI · 1 Jan 2023 – 22 Nov 2026
Testing claims of AGI
What evidence do new models and tests offer that AI is approaching or falling short of AGI?
157 headlines since 1 January 2023; active. The log below is what changed each week, newest first: three lines a week, never more, each a story with the articles behind it.
-
Stop the AI Race: 150 Protesters Outside OpenAI Demand Trump-Xi Superintelligence Treaty
-
METR Finds Anthropic Benchmark Progress Accelerated 1.9-Fold Since October 2025
-
Arm's AGI CPU starts AI execution environments up to 45% faster than AMD's EPYC 9965 in sandbox testing
-
AI scientist is now running its own biology lab: It can design experiments, test ideas and make new discoveries
-
Trump's latest order: artificial intelligence becomes "superintelligence" · BNN
-
Week of 7 September
The week's stories are not public yet.
-
The Superintelligence Test: Build What Better AI Can't Commoditize
-
Many AI reasoning models burn through long chains of intermediate tokens before answering; Pathway's 150-million-parameter BDH-CQ instead reasons iteratively inside a continuous latent state. On ARC-AGI-1, it achieved 29.5% pass@2 at an estimated $0.0007 per puzzle—about 11 times cheaper than OpenAI's GPT-5.6 Luna at low reasoning effort, although Luna scored slightly higher.
-
The real U.S.-China AI contest is diffusion, not superintelligence — and China is already ahead
-
Vocalbeats.AI Joins AGI Playground Singapore as Strategic Partner
-
Subaru Wins CNCF End User Case Study Contest for Accelerating AI Development with Cloud Native Infrastructure
-
From digital to embodied intelligence: top Chinese AI scientist on China's contributions and urgent need to bridge global intelligence gaps
-
GPT-5.6 Release Nears: Ultra Mode Spawns Subagents, Terra Cuts Cost, METR Flags Risk
-
Sigmatic Sciences Powers AI Scientists with NVIDIA's Latest BioNeMo Agentic Capabilities
-
Moonshot AI's Kimi K2.5 scores 11.8% on ARC-AGI-2, leading Chinese models on the benchmark · Digg
-
Epoch AI updates FrontierMath benchmark after correcting errors in 42% of problems, boosting scores for GPT-5.5 and Google · Digg
-
AIs like ChatGPT fall apart in classic 'Stroop' psychological test — and that could stand in the way of achieving artificial general intelligence
-
Arvind Narayanan and DeepMind's Séb Krier argue shifting AGI goalposts represent a 'false summit' of inadequate benchmarks · Digg
-
DeepMind CEO: 'AGI' Is Coming Soon, But Here's the Test It Must Pass First
-
Protesters shut down Berkeley Forum event hosting Google AI scientist
-
Sea sets up AI centre of excellence in Singapore to accelerate AI-native ambitions
-
Week of 6 April
The week's stories are not public yet.
-
Microsoft's latest AI hire frees Mustafa Suleyman to chase superintelligence and build in‑house AI models
-
Universal Robots and Scale AI Launch Imitation Learning System to Accelerate AI Model Training, Bridging the 'Lab-to-Factory' Gap
-
New AGI benchmark: Demis Hassabis proposes 'Einstein test'—Ultimate challenge to prove true intelligence
-
The AGI illusion: Why the world's smartest AI is only as good as its data architects
-
Exeter Probes AI Scientists' Reasoning Abilities
-
Amodei's AI Reckoning: Superintelligence Tests Humanity's Grip
-
Logical Intelligence Introduces First Energy-Based Reasoning AI Model, Signals Early Steps Toward AGI, Adds Yann LeCun and Patrick Hillmann to Leadership
-
Andrew Ng proposes new 'Turing AGI' test to cut through AI hype | Technology News
Part of Argus Overwatch's investigations: a topic followed over months, as the record wrote it. Stories link to their pages and the articles behind them; names link to their own pages; claims link to the fact-checkers who rated them. Nothing on these pages is projected; a quiet week is drawn as a quiet week.