CNN Just Ran One of the Worst AI Puff Pieces We’ve Ever Seen in Our Lives About That “AI Actor” Tilly Northwood29m ago How Credit Genie keeps codebase docs fresh with OpenWiki31m ago Closing The Gap Between AI Ambition And Data Reality32m ago Universal Music is launching an AI music platform with ElevenLabs37m ago Top Pentagon Official Rejects Anthropic Researcher’s AI Warning1h ago Arlequin AI raises €28M to build novel AI models that learn complex relationships at scale41m ago SpaceX sues to block release of tax-break records for its Texas Terafab project41m ago Salesforce closes Fin and eyes $2bn Listen Labs, Business Insider reports42m ago Google brings Gemini for desktop app to Windows43m ago AI’s Hidden Apprenticeship Tax (And How Firms Can Avoid It)47m ago Abacus.AI Releases Three Open-Weight Smaug Models for Agentic Workloads52m ago Clay raises $115m at a $7.1bn valuation, more than double its 2025 price53m ago Before AI Agents Take Action, Enterprises Need A New Data Rule Book1h ago Feature Engineering in Scikit-Learn: A KDnuggets Cheat Sheet1h ago

CNN Just Ran One of the Worst AI Puff Pieces We’ve Ever Seen in Our Lives About That “AI Actor” Tilly Northwood29m ago How Credit Genie keeps codebase docs fresh with OpenWiki31m ago Closing The Gap Between AI Ambition And Data Reality32m ago Universal Music is launching an AI music platform with ElevenLabs37m ago Top Pentagon Official Rejects Anthropic Researcher’s AI Warning1h ago Arlequin AI raises €28M to build novel AI models that learn complex relationships at scale41m ago SpaceX sues to block release of tax-break records for its Texas Terafab project41m ago Salesforce closes Fin and eyes $2bn Listen Labs, Business Insider reports42m ago Google brings Gemini for desktop app to Windows43m ago AI’s Hidden Apprenticeship Tax (And How Firms Can Avoid It)47m ago Abacus.AI Releases Three Open-Weight Smaug Models for Agentic Workloads52m ago Clay raises $115m at a $7.1bn valuation, more than double its 2025 price53m ago Before AI Agents Take Action, Enterprises Need A New Data Rule Book1h ago Feature Engineering in Scikit-Learn: A KDnuggets Cheat Sheet1h ago

AI Research Machine Learning6 min reading time

OpenAI’s AGI number came from a harness, not the model

The Next WebPublished 4 days ago

OpenAI's GPT-6 Astra achieved a 99.9% AGI benchmark score using its proprietary Provider Adapter harness, but the same model scored 62.7% under the standard ARC Prize harness. The difference stems from the software environment around the model, not the model itself, highlighting the impact of evaluation setup on AGI claims.

More on this story

EU cybersecurity agency is now testing Mythos 5 and GPT-6 Astra, the Commission says
8h· The Next Web
The Fight Over OpenAI’s Math Breakthrough Is a New Kind of Scientific Arms Race
21h· 2 sources
OpenAI Says It Has Solved a Millennium Prize Problem—a Holy Grail of Math
2d· 5 sources
Last Week in AI #343 - GPT-6, OpenAI’s agents chatted on a wiki, Fable 5.1
3d· 7 sources
GPT 6 Is Out In OpenAI’s Sandbox
3d· Forbes
Nvidia's Jensen Huang says 'AGI has arrived' and congratulates OpenAI
3d· 2 sources

#arc-prize #benchmarking #francois-chollet #gpt-6-astra #mike-knoop #openai

More in AI Research

AI Research4 min read

An Anthropic researcher just quit, saying OpenAI and Anthropic are 'gambling with our lives'

Covered by 12 sources1d ago

AI Research3 min read

Worried Anthropic researchers warn that AI ‘could kill all humans’

Covered by 8 sources1d ago

AI Research4 min read

Clay raises $115m at a $7.1bn valuation, more than double its 2025 price

The Next Web53m ago