Tensorlake
PulseAugur coverage of Tensorlake — every cluster mentioning Tensorlake across labs, papers, and developer communities, ranked by signal.
1 day(s) with sentiment data
-
Tensorlake achieves 100 parallel Harbor rollouts with 7.9x throughput increase
The article details the successful scaling of Harbor rollouts on the Tensorlake platform, achieving 100 parallel instances. This enhancement resulted in a 7.9x increase in throughput, with the system handling 47 rollout…
-
Malware infects Tensorlake npm package, stealing credentials and targeting AI tools
The npm package `[email protected]` was compromised by a Shai-Hulud worm variant. This malware targets developer environments and CI/CD pipelines, with the capability to steal cloud and code repository credentials. It …
-
Gremlin uses AI to speed up system failure testing; Shai-Hulud malware targets AI infra
The cybersecurity firm Gremlin has integrated AI into its platform to accelerate the process of identifying vulnerabilities in distributed systems. This advancement allows for faster detection of potential failures and …
-
Shai-Hulud malware targets AI infrastructure, compromising Tensorlake
A sophisticated malware strain known as Shai-Hulud has been adapted to target AI infrastructure, specifically compromising the Tensorlake platform. This development highlights the evolving threat landscape as malicious …
-
AI Sandbox Networking: Tensorlake, E2B, Daytona, Fly.io Compared
This article compares the networking architectures of four AI sandbox platforms: Tensorlake, E2B, Daytona, and Fly.io, focusing on how they route ingress traffic. It details the trade-offs between Layer 7 (L7) proxies, …
-
AI Sandbox Security: 24 Exfiltration Attacks Blocked by Tensorlake Policy
A security test evaluated 24 exfiltration attacks against an AI sandbox environment, specifically Tensorlake's deny-by-default network policy. The attacks, designed to send data out of the sandbox via various methods in…
-
AI Sandboxes Can Dynamically Update Network Policies for Enhanced Security
A common challenge in AI development is managing network access for sandboxed environments. Typically, network policies are set when a sandbox is created and remain static, meaning a sandbox that needs broad access for …
-
Reproducible AI Pipelines: Versioning Filesystems for Multi-Agent Systems
This article details how to create a reproducible multi-agent AI pipeline by versioning the filesystem rather than individual agents. It proposes using Tensorlake Cloud Volumes to manage agent outputs, allowing for snap…
-
Tensorlake optimizes sandbox ingress by shifting from L7 to L4 forwarding
Tensorlake has redesigned its sandbox ingress path, moving a dataplane hop from a full L7 reverse proxy to an L4 forwarder utilizing kernel TLS (kTLS) and splice(2). This change was primarily driven by the need to optim…
-
Tensorlake's AI agent forking cuts setup time by 5x
The author tested Tensorlake's snapshot-and-fork feature for AI agents, finding that it significantly reduces setup time compared to rebuilding environments for each task. By creating a single parent sandbox with depend…
-
Tensorlake launches Sandboxes for isolated AI workload management
Tensorlake has introduced Sandboxes, a new feature designed to help developers build, run, and manage isolated AI workloads. This hands-on guide details how to provision a Sandbox, execute Python code, manage dependenci…
-
Open-source coding agent OpenCode runs safely in isolated sandbox
OpenCode, an open-source coding agent similar to Claude Code, offers full shell access, which presents potential risks. The author details a method for safely running OpenCode by isolating its commands within a disposab…
-
Tensorlake enables safe parallel AI agent execution with microVM isolation
The author details a method for running multiple AI agents in parallel using Tensorlake's sandboxing technology, which provides structural isolation for each agent's process, filesystem, and memory. This approach preven…
-
Developer builds stateful AI research agent using TensorLake sandbox
A developer explored building a stateful research agent, encountering issues with traditional stateless execution environments that lost context. They found that while stuffing state into prompts or using external store…
-
Claude Opus 4.7 outperforms Kimi K2.6 in coding agent task
A user stress-tested Anthropic's Claude Opus 4.7 and Moonshot's Kimi K2.6 on a complex coding agent task involving remote sandbox execution. Claude Opus 4.7 successfully built a functional AI Fix Runner, handling local …