Design, development, and security
Supermemory
The memory layer for AI agents. Context engineering platform powering enterprise APIs, developer plugins, and a personal app that remembers everything.
AI proxy: fostering a more open ecosystem - Blog - Braintrust
Introducing Braintrust's latest feature: an AI proxy that lets you use open source models like LLaMa 2 and Mistral, as well as all of OpenAI's and Anthropic's models, behind a single interface with caching, security, and API key management built in.
agents-chat
Multi-user streaming AI chats built using RedwoodSDK RSCs, Cloudflare Agents, Vercel AI SDK, and OpenAI Agents SDK.
ExploitGym: Can AI Agents Turn Security Vulnerabilities into Real Attacks?
AI agents are rapidly gaining capabilities that could significantly reshape cybersecurity, making rigorous evaluation urgent. A critical capability is exploitation: turning a vulnerability, which is not yet an attack, into a concrete security impact, such as unauthorized file access or code execution. Exploitation is a particularly challenging task because it requires low-level program reasoning (e.g., about memory layout), runtime adaptation, and sustained progress over long horizons. Meanwhile, it is inherently dual-use, supporting defensive workflows while lowering the barrier for offense. Despite its importance and diagnostic value, exploitation remains under-evaluated. To address this gap, we introduce ExploitGym, a large-scale, diverse, realistic benchmark on the exploitation capabilities of AI agents. Given a program input that triggers a vulnerability, ExploitGym tasks agents with progressively extending it into a working exploit. The benchmark comprises 898 instances sourced from real-world vulnerabilities across three domains, including userspace programs, Google's V8 JavaScript engine, and the Linux kernel. We vary the security protections applied to each instance, isolating their impact on agent performance. All configurations are packaged in reproducible containerized environments. Our evaluation shows that while exploitation remains challenging, frontier models can successfully exploit a non-trivial fraction of vulnerabilities. For example, the strongest configurations are Anthropic's latest model Claude Mythos Preview and OpenAI's GPT-5.5, which produce working exploits for 157 and 120 instances, respectively. Notably, even with widely used defenses enabled, models retain non-trivial success rates. These results establish ExploitGym as an effective testbed for exploitation and highlight the growing cybersecurity risks posed by increasingly capable AI agents.
ExploitGym: Can AI Agents Turn Security Vulnerabilities into Real Attacks?
AI agents are rapidly gaining capabilities that could significantly reshape cybersecurity, making rigorous evaluation urgent. A critical capability is exploitation: turning a vulnerability, which is not yet an attack, into a concrete security impact, such as unauthorized file access or code execution. Exploitation is a particularly challenging task because it requires low-level program reasoning (e.g., about memory layout), runtime adaptation, and sustained progress over long horizons. Meanwhile, it is inherently dual-use, supporting defensive workflows while lowering the barrier for offense. Despite its importance and diagnostic value, exploitation remains under-evaluated. To address this gap, we introduce ExploitGym, a large-scale, diverse, realistic benchmark on the exploitation capabilities of AI agents. Given a program input that triggers a vulnerability, ExploitGym tasks agents with progressively extending it into a working exploit. The benchmark comprises 898 instances sourced from real-world vulnerabilities across three domains, including userspace programs, Google's V8 JavaScript engine, and the Linux kernel. We vary the security protections applied to each instance, isolating their impact on agent performance. All configurations are packaged in reproducible containerized environments. Our evaluation shows that while exploitation remains challenging, frontier models can successfully exploit a non-trivial fraction of vulnerabilities. For example, the strongest configurations are Anthropic's latest model Claude Mythos Preview and OpenAI's GPT-5.5, which produce working exploits for 157 and 120 instances, respectively. Notably, even with widely used defenses enabled, models retain non-trivial success rates. These results establish ExploitGym as an effective testbed for exploitation and highlight the growing cybersecurity risks posed by increasingly capable AI agents.
Data Streaming for AI: From Extractive Training to Sovereign Infrastructure DWeb Camp 2026
AI systems are consuming the world's content without compensating its creators. This session explores data streaming as a new paradigm — where content flows to AI in real time, with built-in rights management, usage tracking, and fair compensation — and asks what it would take to make this infrastructure decentralized, sovereign, and governed by the communities it serves.
Exploring Generative AI
Notes from my Thoughtworks colleagues on AI-assisted software delivery
Exploring Generative AI
Notes from my Thoughtworks colleagues on AI-assisted software delivery
Project Think: building the next generation of AI agents on Cloudflare
Announcing a preview of the next edition of the Agents SDK — from lightweight primitives to a batteries-included platform for AI agents that think, act, and persist.
Challenges and Paths Towards AI for Software Engineering
View recent discussion. Abstract: AI for software engineering has made remarkable progress recently, becoming a notable success within generative AI. Despite this, there are still many challenges that need to be addressed before automated software engineering reaches its full potential. It should be possible to reach high levels of automation where humans can focus on the critical decisions of what to build and how to balance difficult tradeoffs while most routine development effort is automated away. Reaching this level of automation will require substantial research and engineering efforts across academia and industry. In this paper, we aim to discuss progress towards this in a threefold manner. First, we provide a structured taxonomy of concrete tasks in AI for software engineering, emphasizing the many other tasks in software engineering beyond code generation and completion. Second, we outline several key bottlenecks that limit current approaches. Finally, we provide an opinionated list of promising research directions toward making progress on these bottlenecks, hoping to inspire future research in this rapidly maturing field.
SambaNova | The Fastest AI Inference Platform
Discover SambaNova - the complete AI platform delivering the fastest AI inference, fine-tuning, and scalable solutions for agentic AI easily integrated into existing data center infrastructures.
Introduction
Every stream on Streamplace is cryptographically bound with metadata that captures details about how the stream should be viewed, used, distributed, and monetized. This metadata also identifies the provenance of the stream, including who the creator is and any transformations it has undergone. Any Streamplace node operator can inspect the stream and verify that this metadata is intact and trustworthy.
The Expanding Dark Forest and Generative AI
An exploration of the problems and possible futures of flooding the web with generative AI content
generative-ai/gemini/agents/always-on-memory-agent at main · GoogleCloudPlatform/generative-ai
Sample code and notebooks for Generative AI on Google Cloud, with Gemini Enterprise Agent Platform - GoogleCloudPlatform/generative-ai