The Archive

The acquisition also leaves Sequoia-backed Serval as the de-facto startup leader in AI IT service automation, industry watchers believe.

AI mushroom identification models achieve only 65% accuracy, posing safety risks.

llm-openrouter 0.7.1 ships performance fix for loading OpenRouter models.

OpenAI’s new Astra model will use “recurrent depth,” a technique that allows the model to operate outside of the sequential thinking that characterizes most reasoning models.

Google launched Gemini 3.8 Flash, arriving just a few weeks after its predecessor. The company claims the new model "works harder" than Gemini 3.7 Flash by performing more reasoning steps on complex tasks and "calling tools iteratively." It has the same introductory pricing as 3.7 Flash, $0.75 per million input tokens and $3.75 per million output tokens, but could still end up costing users more. Google warns that "the model might use more tokens to maximize performance, especially at higher effort levels." Developers can keep using Gemini 3.7 Flash if they want to minimize token usage. Gemin...

llm 0.34 adds usage logging with response duration metrics and performance improvements.

Google's Pro model updates are seemingly paused, but there's yet another Gemini Flash today.

Proposes standardized metrics for speech brain-computer interfaces to enable comparable evaluation across heterogeneous datasets and recording methods.

Introduces predicted-state matching training objective to improve world models for web agents by aligning predictions with downstream ranking tasks.

llm-anthropic 0.28 displays Claude reasoning traces by default and adds refusal exception handling.

Trump’s secret reviews of frontier AI models may hide corruption, lawsuit says.

Graph Machine architecture uses dynamic sparse routing with O(n) state for efficient pretraining; reduces Qwen3-0.6B compute by replacing 75% dense layers.

JAX library for computing reverse-mode gradients of ODE ensembles at native solver speed, targeting scientific and engineering applications.

Amazon is trying to combat impersonation scams with a new feature that allows you to use its AI assistant to determine whether an email, text message, or phone call actually came from the company. With the update, you can ask Alexa for Shopping about a message you received, and it will use AI to compare it "against a record of every message Amazon has sent," while also analyzing its contents, formatting, and the sender. As noted by Amazon, the assistant will only confirm that a message is real if it's "completely certain." An example shared by Amazon shows a customer asking, "Did Amazon just ...

TRACE framework for autonomous robots provides causal traceability of decisions through four auditable layers to enable post-incident reconstruction.

Demonstrates user feedback signals strong LLM improvement signal; prior evaluation methods systematically underestimated utility via synthetic and naturalistic data.

Tightens theoretical lower bounds for gradient descent optimization in smooth convex settings beyond classical Nesterov limits.

Introduces linguistic illegibility concept: LLM external outputs and mechanistic probes may misrepresent internal computation, with implications for interpretability and security.

Nemotron-3-Nano/Ultra-CC models trained on 22k curated competitive programming problems with SFT, RL, and GenCorrect test-time strategy for IOI/ICPC performance.

Proposes FP4 pretraining recipe with E5M3 block scaling and selective stochastic rounding to stabilize 4-bit language model training.

The internet has a trust problem, and it’s not just because social media feeds are filling up with AI slop. AI-generated text and images are now making their way into job applications, product reviews, and even insurance claims, leaving platforms and users alike scrambling to figure out what’s real. A handful of startups have cropped up in the past couple of […]

Meshfree discretization methods using kernel consistency constraints; pure numerical analysis, no AI application.

NVIDIA CUDA remains the foundation of GPU-accelerated computing, powering everything from scientific simulations to large-scale AI training. But writing... Source

The internet has a trust problem, and it’s not just because social media feeds are filling up with AI slop. AI-generated text and images are now making their way into job applications, product reviews, and even insurance claims, leaving platforms and users alike scrambling to figure out what’s real. A handful of startups have cropped up in the past couple of […]

AICOME framework validates AI-derived social/occupational measures against survey data for group-level effect recovery.

"The United States has a strong interest in continuing to develop a robust and competitive artificial intelligence industry that sets the standard for the practice and procedure of AI use globally," the brief reads.

Cliff: process reward learning from first LLM reasoning mistake improves RL-based post-training without specialized models.

Structured reasoning framework for LLM-powered telecom root cause analysis with evidence grounding to reduce hallucination.

OpenAI is on the cusp of releasing its most powerful AI model yet, Astra, following weeks of delays to shore up safety protocols after its agents attacked real targets during testing. As details about the model trickle out, researchers are warning it "may be the single worst development for AI security/safety to date." Shortly after OpenAI said on Tuesday that it had delayed Astra's release to work on safety issues, The Information reported that Astra shows far less of its "thinking" than other frontier AI models, sparking concern it could be dangerously hard to monitor. Most top AI systems t...

Optimality proof for frb100-40 benchmark using independent set construction; pure combinatorial optimization.