A tech essay describes a trajectory where software companies increasingly wrap their operations around large language models, moving from engineers pairing with coding agents to entire teams triggering and reviewing background agents, and eventually to agents proactively deciding work themselves. The author terms this evolving structure a 'harness' — the infrastructure, interfaces, and state built around a stateless model — and suggests most SaaS businesses are already on this path, whether they recognize it or not.
blog.sshh.io
· 2026-10-02
A Show HN post introduces a tool that automatically generates, tests, benchmarks and refines CUDA GPU kernels from a workload description. It uses an LLM agent via LangGraph to propose code and launch-configuration changes, a C++/NVRTC harness to compile and run candidates, and Python to validate outputs against NumPy references, keeping only the fastest kernel that passes all correctness checks. The project was developed on Windows with an RTX 3060 Laptop GPU and requires Python 3.12+, CMake, a C++17 compiler and an OpenAI API key.
github.com
· 2026-09-25
In a controlled test of LangGraph's Agent Server, researchers showed that a 'maker' role could swap a pending, human-approved transfer request for a different, much larger transfer after approval had already been granted. The approver signed off on a mock transfer of 20 units to an approved vendor, but the ledger recorded a transfer of 2,000 units to an attacker-controlled account under that same approval. The test used synthetic identities and a mock ledger, and the researchers published the request logs, decisions and exact software versions tested.
adithyanak.com
· 2026-09-23
A team has built safe-upgrade, a dependency-upgrade agent that separates three concerns: Jev converts release notes and repository evidence into typed, probabilistic decisions; LangGraph runs those decisions through a durable workflow; and Tenuo issues each resulting action a task-scoped warrant limiting its authority. The project uses dependency upgrades as a test case because passing installs and CI checks can still hide broken APIs or behavior changes that existing tests don't catch.
tenuo.ai
· 2026-09-23
Amazon has released Pizza Bot, an open-source tool under Apache 2.0 that organizes long-running AI agent tasks into an inbox-style interface. Completed tasks land in an 'Unread' queue while tasks needing human approval land in an 'Action' queue, letting users start work and check back later rather than watching agents run in real time.
github.com
· 2026-09-15
GoKawiil