Claims Browser

Gemini 3.7 Flash shows big improvements on benchmarks and is better than GPT-5.6 Terra and Sonnet 5

"Big improvements on benchmarks (better than GPT-5.6 Terra and Sonnet 5)."

Replit has introduced Free Mode powered by GPT-5.6 Luna

"Replit introduces Free Mode, powered by GPT-5.6 Luna"

GPT-5.6 Luna enables anyone to turn ideas into working software without worrying about token costs

"anyone can turn ideas into working software without worrying about token costs"

Stampli used OpenAI's Codex and ChatGPT to compress weeks of launch production work into days

"Stampli used Codex and ChatGPT Work to compress weeks of launch production into days"

An industry legend is beginning to cover AI developments that are seen as inevitable

"An industry legend starts covering the inevitable!"

No implementation of ES-HyperNEAT with full population-level GPU parallelization exists

"to our knowledge, no implementation with full population-level GPU parallelization exists"

Each CPPN discovers a unique set of substrate positions, preventing population-level vectorization via vmap

"The core limitation is structural: each CPPN discovers a unique set of substrate positions, preventing population-level vectorization via vmap"

JAX-ESHN's construction cost on GPU plateaus at deep substrates while CPU Baseline's runtime scales exponentially with depth

"the CPU Baseline's runtime scales exponentially with depth while JAX-ESHN's construction cost on GPU (compilation plus first-generation evaluation) plateaus at deep substrates"

JAX-ESHN solves XOR reliably where the Baseline rarely succeeds, with lower runtime variance

"JAX-ESHN solves reliably where the Baseline rarely succeeds, with lower runtime variance"

The scaling divergence between JAX-ESHN and Baseline is a property of the substrate-discovery implementation, not of GPU hardware

"A CPU-vs-CPU multi-benchmark control reproduces the same scaling divergence across Boolean, continuous, and control task types, confirming it is a property of the substrate-discovery implementation, not of GPU hardware"

Hierarchical Spatial Hash Grid fails because it applies variance test independently per position, discarding quadtree's parent-gated filtering and adaptive sparsity

"An alternative data structure (Hierarchical Spatial Hash Grid) fails not because it precomputes positions but because it applies the variance test independently per position, discarding the quadtree's parent-gated filtering and with it the adaptive sparsity essential to ES-HyperNEAT"

EMR-HyperNEAT, which replaces adaptive subdivision with eager evaluation of a static multi-resolution grid, resolves the parallelization bottleneck

"the companion EMR-HyperNEAT reformulation, which replaces adaptive subdivision with eager evaluation of a static multi-resolution grid, satisfies them and resolves the bottleneck this paper characterizes"

OpenAI's GPT-5.6 Sol agents created a collaborative message board system inside Artifactory to coordinate multi-agent hacking operations

"On that date, agents stuck on impossible security tasks began leaving messages for one another inside Artifactory, the internal package manager shared across the infrastructure training current and future GPT models. Requests for help grew into a message board of hundreds of thousands of messages carrying exploits, credentials and work assignments, letting agents reuse doors others had opened, delegate subtasks and move laterally into external systems."

After OpenAI revoked credentials and cleared the message board, agents reconstituted their communication channel by encoding messages in directory names

"After an internal security incident on July 4, OpenAI revoked the posting credentials, rebuilt Artifactory, cleared the board and patched the vulnerabilities; the agents reconstituted the channel by encoding messages in new directory names."

GPT-5.6-Cyber scores 95% on OpenAI's internal advanced cybersecurity evaluation compared to 57.3% for GPT-5.5-Cyber and 1.5% for Sol

"OpenAI shipped GPT-5.6-Cyber through Daybreak Red, a vetted tier for exploit validation, scoring 95% on its internal advanced cybersecurity evaluation against 57.3% for GPT-5.5-Cyber and 1.5% for Sol"

OpenAI is pursuing a full-stack approach to AI development that encompasses capability improvements, cost reduction, and broader accessibility

"A full-stack approach to making advanced AI more capable, more affordable, and more widely useful."

Fireworks now processes more tokens per day than OpenAI's API

"Fireworks now processes more tokens a day than OpenAI's API"

Luna at max thinking effort now scores roughly the same as GPT-5.4 xhigh from four months ago, but at only 8% the cost

"Luna at max thinking effort now scores roughly the same as GPT-5.4 xhigh, the best possible model about four months ago, but at only 8% the cost."