691
692
Mistral Small 4
The most powerful AI platform for enterprises. Customize, fine-tune, and deploy AI assistants, autonomous agents, and multimodal AI with open models.
693
Headlong: A microharness for persistent agents
Self-guided agents that think continuously
694
OpenAI models secretly generate instructions to ignore constraints
Research on aligning AI with human values and intent, and reports documenting model failures.
695
Letter to Governor Abbott on responsible AI infrastructure in Texas
OpenAI sent Governor Greg Abbott a letter outlining its commitment to responsible AI infrastructure in Texas. The letter supports reliable, transparent growth that benefits Texans.
696
How scientists are using Claude to accelerate research and discovery
Labs at Stanford and MIT built Claude-powered systems that run genome-wide studies in minutes and automate gene cluster interpretation.
697
A third of Perplexity's citations don't contain the number they're cited for
Of 1,826 citations Perplexity's search models attached to a sentence stating a figure, 34.7% pointed at a page that either would not open or did not contain a single figure from that sentence; scored per claim rather than per citation, 14.4% of 872 claims fail.
698
Responsible Release of AI-Generated Mathematics
Responsible Release of AI-Generated Mathematics September 29, 2026 Back to main page Download PDF At present, some frontier AI labs are testing advanced mathematical problems on proprietary models …
699
Anthropic forms $200M partnership with the Gates Foundation
We're committing $200 million in grants, Claude credits, and technical support for global health, life sciences, education, and economic mobility programs.
700
Towards Self-Driving Codebases
We're post-tokenmaxxing. What are we pre-? How do we get there?
701
Can LLMs model real-world systems in TLA+?
Editors’ note: AI has been actively pushing the frontier of applied formal methods for computing systems. In this article, the Specula team wrote about their experience of evaluating LLMs on modeling system code, the basic capability for agentic model checking, using TLA+, a specification language f
702
ChatGPT's image generator can be manipulated to produce violent, sexual content
Mindgard reveals how ChatGPT's image generation safety filters can be bypassed with minor prompt adjustments, producing graphic violent and sexual content without users directly asking for it. Here's what was found, and why it matters.
703
Rust-lang/rust is adopting an LLM policy
Want to follow along with Rust development? Curious how you might get involved? Take a look!
704
Open Weights isn't Open Training
How many monkey-patches does it take to post-train a trillion parameter model?
705
Measuring AI agent autonomy in practice
Analysis of millions of human-agent interactions shows agent turns getting longer, with software engineering making up nearly half of tool calls.
706
AI models may be capable of self-replicating without access to a datacenter
Small AI models may be only months away from being able to self-replicate across cheap IoT devices like Android TV boxes, with no datacenter needed.
707
Compute-efficient pretraining and scaling to trillion-parameter models
Research update on compute-efficient pretraining and scaling to trillion-parameter models.
708
MathCode, Mathematical Coding Agent
A terminal AI coding agent that formalizes math problems into Lean 4 theorems and proves them.
709
Every Model Cheats
This post presents a controlled prompt-ablation study: 23 tasks, three prompt conditions, 1,518 individually audited traces, and a simple question: can you prompt away cheating?
710
Claude designed protein binders against scorpion neurotoxins
We asked an internal research Claude model to design minibinders (small proteins that attach tightly to a target) against scorpion neurotoxins.
711
Terminal-Bench-Science: Evaluating AI agents on scientific research workflows
A benchmark for evaluating AI agents on research workflows across scientific domains
712
The Future Worth Building Is Human
AI built for autonomy crowds people out, making us passive observers of what’s coming. We’re building toward a different future.
713
Inference Optimization for MiMo v2.5: Pushing Hybrid SWA Efficiency to the Limit
An end-to-end engineering practice for the MiMo-V2.5 series inference system, covering KVCache management, tiered caching, SWA-aware prefix cache trees, scheduling, Prefill/Decode pipelines, and multimodal optimizations.
714
We built a persistent agent memory layer on Elasticsearch with 0.89 recall
Persistent agent memory on Elasticsearch: three-index architecture, hybrid retrieval, supersession and DLS isolation. R@10 0.89, zero cross-tenant leaks.
715
Wireless magnetic millirobots that deliver medicine inside the beating heart
Hopkins engineers have created robots the size of a pencil tip intended to travel through the body to deliver medicine to infections inside a beating heart
716
Why we don’t use AI
We get asked about AI a lot. Whether we’re going to add it to Yarn Spinner, whether we use it ourselves, what we think about it. Fair questions. Time to write it all down. Yarn Spinner doesn’t use the technology that’s currently being called AI. We don’t have generative AI features in the product, and we don’t use…
717
Superhuman AI Exfiltrates Emails
Superhuman AI was able to exfiltrate sensitive emails from user accounts - without the user even being aware. This vulnerability was rapidly remediated by the Superhuman team.
718
Apple capped engineers at a $200 daily AI budget
The world's richest company is suddenly rationing compute for its own builders.
719
Side-by-side comparison of how AI models answer moral dilemmas
Explore how 20 leading AI models respond to tough ethical, social, and political questions. See how their answers compare to each other and to you.
720
NanoClaw Adopts OneCLI Agent Vault
Every NanoClaw agent will access external services through OneCLI's Agent Vault, a gateway that handles credential injection, access policies, and approvals so agents never hold raw API keys.