Loading activity…
@brendanfoody
ceo @mercor | organizing human intelligence
AI
A post says Carbon scored 88.2% on SEC-bench Pro and is rolling out first to cybersecurity specialists.
AI
Kimi claims a 1-million-token context window for Pro and two to three times Pro's speed for Highspeed.
XAI
A post says Sam Altman linked open release to changes in internet cyber defenses and new protection against biological threats.
YouTubeAI
The @flappyairplanes account names Index and Lux as co-leads and says the team spent a year on GPU primitives.
AI
The Frontier's host says Databricks' Jonathan Frankle estimates AI has automated a very low percentage of enterprise tasks.
AI
A post argues that enterprise evals can reward just one correct answer while failing to penalize some wrong ones.
AI
Artificial Analysis says Exa's auto mode with Dynamic Highlights scored 80 on its Search Index, tying Perplexity Search (medium).
AI
A user says they don't believe in sunscreen and predicts AI will be able to fix our skin.
AI
A post says Mercor's entire company handwrote all the data in 24 hours after earning the pilot.
AI
One post argues that automating everything in the economy requires reinforcement-learning environments across every industry, job and task.
AI
On The Frontier, Liam Fedus explains why Periodic Labs is building physical labs to generate scientific data.
XAI
A post about The Frontier says Liam Fedus discussed what he learned at OpenAI as ChatGPT took off.
XAI
A post calls the spending share “surprisingly consistent” across labs and AI-native companies.
AI
Frontier host Brendan Foody says Harvey’s Gabe Pereyra sees building agent infrastructure as hard enough to be a company on its own.
AI
Reflection says Beam uses 23B active parameters and plans to publish its full weights later in October 2026.
AI
A 60 Minutes report shows experts teaching models while examining the unpredictable hours behind that work.
AI
A shared excerpt puts the average accountant’s score at roughly 37%, above the best AI models 18 months earlier.
AI
Google announces a 1 million-token output limit, up from 64,000, for its new model. Initial access goes to trusted cyber defenders through the Fairwind Program.
AI
A user argues frontier AI labs should offer this as a product, claiming it would create switching costs and provide more than 5× compute efficiency.
